Jump to ContentJump to Main Navigation
How Do You Feel?An Interoceptive Moment with Your Neurobiological Self$

A. D. (Bud) Craig

Print publication date: 2014

Print ISBN-13: 9780691156767

Published to Princeton Scholarship Online: October 2017

DOI: 10.23943/princeton/9780691156767.001.0001

Show Summary Details
Page of

PRINTED FROM PRINCETON SCHOLARSHIP ONLINE (www.princeton.universitypressscholarship.com). (c) Copyright Princeton University Press, 2022. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in PRSO for personal use. Subscriber: null; date: 16 May 2022

The Interoceptive Pathway to the Insular Cortex

The Interoceptive Pathway to the Insular Cortex

Lamina I Spinothalamic Input to the Thalamus and Cortex in Primates

(p.130) 5 The Interoceptive Pathway to the Insular Cortex
How Do You Feel?

A. D. (Bud) Craig

Princeton University Press

Abstract and Keywords

This chapter looks at the experiments that demonstrated in monkeys and humans the unforeseen lamina I pathway to the thalamus and its subsequent projection to the interoceptive cortex. The ascending interoceptive thalamocortical pathway is phylogenetically unique to primates; it most likely arose in conjunction with the enormous encephalization associated with the emergence of the primate lineage. The existence of this pathway was a surprise to most investigators in the field of somatosensory neurobiology. As mentioned in chapter 1, a sensory representation of general feelings from the body had been envisioned by the German natural philosophers of the nineteenth century. However, that concept was superseded by the heuristic codification of nociception and the assignment of pain and temperature sensations to the somatosensory cortex. The chapter's findings rectify that misconception and substantiate the fundamental neurobiological distinction between interoception and exteroception at the thalamocortical level in the monkey and human.

Keywords:   lamina I pathway, interoceptive cortex, interoceptive thalamocortical pathway, primates, encephalization, nociception, somatosensory cortex, interoception, exteroception

In chapters 1 and 2, I introduced the idea of an interoceptive pathway that provides a sense of the physiological condition of the body, and I outlined the progression of ideas that led me from that ascending pathway to a model of feelings as the coinage of homeostatic valuation. Chapters 3 and 4 presented the functional and anatomical evidence regarding lamina I neurons and their projections at spinal and brainstem levels. That evidence identifies lamina I as the origin of the central interoceptive pathway that serves homeostasis, and the origin of ascending sensory channels for bodily affective feelings of sharpness, pricking pain, burning pain, cool, warm, itch, muscle ache, and so on that are virtual labeled lines.

Here I describe the experiments that demonstrated in monkeys and humans the unforeseen lamina I pathway to the thalamus and its subsequent projection to the interoceptive cortex. The evidence is described in this chapter in considerable detail, using more specialized terminology than in the prior chapters and particular words that had been omitted. In case you are more interested in an overview, there is a thorough summary at the end of the chapter.

The ascending interoceptive thalamocortical pathway is phylogenetically unique to primates; it most likely arose in conjunction with the enormous encephalization associated with the emergence of the primate lineage. The existence of this pathway was a surprise to most investigators in the field of somato-sensory neurobiology. As mentioned in chapter 1, a sensory representation of general feelings from the body had been envisioned by the German natural philosophers of the nineteenth century (which they called Gemeingefühl, or "common sensation"). However, that concept was superseded by the heuristic codification of nociception and the assignment of pain and temperature sensations to the somatosensory cortex, the primary cortical representation of cutaneous mechanoreceptive sensation that is portrayed by Penfield's homunculus. The (p.131) findings described here rectify that misconception and substantiate the fundamental neurobiological distinction between interoception and exteroception at the thalamocortical level in the monkey and human.

The ascending lamina I spinothalamic axons are small-diameter fibers, and their terminations could not be distinguished with the silver-stained degeneration method, the autoradiographic method, or the wheat-germ agglutinin / horseradish peroxidase method. The high resolution and sensitivity needed to identify lamina I spinothalamic projections in monkey was finally provided by the introduction of anterograde tracing methods based on a plant lectin (PHA-L) in 1984 and label-conjugated dextrans in 1990.

I begin the description of these functional neuroanatomical studies of the thalamus by relating how I was introduced to this scientific endeavor and by recounting how I learned to find "Waldo." Many of the findings described in this chapter were obtained in my laboratory, and again I relate several episodes in sufficient detail for you to gain a feel for what we experienced and our thoughts and decisions as the experiments progressed. Some of the physiological experiments were performed in collaboration with Jonathan Dostrovsky (University of Toronto, Canada), and some of the neuroanatomical analyses were performed in collaboration with Anders Blomqvist (Linköping University, Sweden).

Box 2 provides an explanation of the experimental methods we used, and boxes 4–6 address ancillary issues on these topics: details on lamina I projections in cat and rat, projections to MDvc, and projections to VPI and VPL. Much, but not all, of the evidence described in this chapter has already been documented in peer-reviewed publications.

My introduction to functional neuroanatomy.

I first became interested in studying the brain while I was an undergraduate student majoring in mathematics at Michigan State University during 1968–1973. After the first two years of my undergraduate studies, I took a year off to consider the direction of my life; the bachelor's degree in mathematics I would receive could lead to a job as an actuary for an insurance company or as a math teacher, but those were not jobs I dreamed about.

When I was a young boy, I had dreamed of being a scientist. Between the ages of 7 and 12, I enjoyed reading books about every kind of science. Later, as my proficiency in mathematics led to advanced classes and to scholarships, my educational focus narrowed; I even opted out of ninth-grade biology class by attesting that I could not stomach dissecting frogs! At that time, space exploration was the frontier, and it generated enormous excitement. That was the era in which humans made the first footprints on the moon and sent the Voyager spacecrafts off toward other star systems. However, in my mind, astronomy and physics were not the most exciting fields of science. To me, the most fascinating scientific frontier (p.132) was the mystery inside my own head, and everyone else's. Study of the brain was just beginning to blossom; new departments of neurobiology were being organized at major universities, and the Society for Neuroscience had been formed just two years earlier, in 1969. During my break from college, I decided to explore the possibility of becoming a neuroscientist.

The door opened for me when, after an interview, I was granted permission to register for a graduate level course in functional neuroanatomy given by Professor J. I. Johnson of the biophysics department at Michigan State. This was a hands-on course. Working in small groups, we first made our own glass-coated tungsten microelectrodes by etching the wire to a fine point and coating all but the very tip with molten glass. Then, on one extraordinary day, we used our microelectrodes to record the responses of neurons in the somatosensory thalamus of an anesthetized raccoon. Our task was to map the somatotopic distribution of clusters of neurons, an experiment that Professor Johnson had in fact first reported in an elegant paper published only ten years earlier. After the grad assistant inserted a microelectrode in the somatosensory thalamus, we began mapping. Each time we touched the animal's body in just the right location or flicked the right hairs at the right speed, we heard over the loudspeaker and saw on the oscilloscope the spiking activity we had elicited from the neurons at that particular location in the somatosensory thalamus. [Welker and Johnson, 1965]

"Wow! That's cool!"

When we moved the microelectrode further into the brain or to a different spatial coordinate, we found that the receptive fields (RFs) of the neighboring clusters of neurons were arranged on neighboring parts of the body in an orderly, topographic manner that defined the somatotopic map of the raccoon's contralateral body surface. After we found activation from adjacent fingers on the raccoon's hand at adjacent coordinates, we plotted all of the response coordinates and RFs together, and there was the raccoon's hand in our three-dimensional graph!

Later we did the histology with a sliding microtome, which means that we moved a very sharp knife on oiled guides across the surface of the hardened brain, in order to cut a series of sections through the thalamus at a thickness (35 micrometers, or in laboratory jargon, "microns") appropriate for examination on the microscope. We stained every other section for cells and the alternate set of sections for fibers. We located the tracks of our recording electrodes in the serial sections, which were revealed by thin, straight lines of cell damage (and "reactive gliosis"), and we identified the locations where we had recorded neuron clusters with particular RFs. Gradually, we reconstructed our recordings and visualized the anatomical map, which confirmed Prof. Johnson's published results. We saw in the raccoon's somatosensory thalamus an anatomical structure formed by parallel, vertical fiber bundles with long intervening regions, where stained neurons were visible in the adjacent histological sections. Our plots showed that (p.133) the functional organization of the regions representing the individual fingers corresponded exactly with that anatomical structure. The histological fingerlike pattern that we could see in the stained section, even without a microscope, was perfectly aligned with the map of the responses we had obtained with our recording microelectrodes by touching each of the adjacent fingers.

"Oh, wow!"

For me, that was an enlightening moment, an epiphany. We could see the outline of the raccoon's hand and fingers in the histological structure of its somato-sensory thalamus with our own eyes! We could actually see the anatomical basis of the functional organization of this part of the brain, and it made sense! And with that, I was hooked.

Basically, I have been doing similar experiments using that same scientific paradigm during my entire career. The mindset of a functional neuroanatomist has guided all of my research. As a neurophysiologist, I have placed recording electrodes in the right locations based on an understanding of anatomical structure; conversely, as a neuroanatomist, I looked in the right place with the right stain and experimental tracing techniques to identify the underlying structural features and connections based on an understanding of functional characteristics. These convergent perspectives gave me a powerful set of complementary tools for studying the fundamental organization of the brain.

The significance of somatotopic organization.

A definitive characteristic of the representations of mechanoreceptive input to the somatosensory thalamus and the somatosensory cortical areas is their somatotopic organization, the property that we demonstrated in the raccoon. The term somatotopy signifies a topologically well-organized neural mapping of the body, which provides a basis for both stimulus localization and modality identification. The somatotopic map of the racoon's hand that we examined in the thalamus as students and the homunculus that was famously mapped by Penfield and his colleagues by stimulating the postcentral gyrus of human cortex both exemplify this organizational principle (see Schott's erudite article on the conceptual and historical significance of Penfield's homunculus). Neurons throughout the map respond to the same type(s) of stimulation within a limited RF, and neurons at neighboring locations respond to the same type(s) of stimulation in neighboring regions of the skin. Each region of the body is represented within a smaller or larger territory (and thus has a different "magnification factor") in accordance with the varying innervation density of that part of the body (and its ecological significance for the animal). Each region is in the right place topologically, because each has the right neighbors. [Schott, 1993]

If there is a global somatic topography without the local, point-to-point precision that underlies correct neighborhood relationships, then we say it has a somatic (p.134) topography but that it is not somatotopic. A somatotopic map is required for independent, modality-specific processing that is capable of indicating stimulus localization by maintaining the neighborhood relationships on the body of the real-world stimuli. In the same way, a retinotopic map is present in primary visual cortex, and a tonotopic map is found in primary auditory cortex. Such topographic order is an efficient basis for the developmental construction of the brain because it can be anatomically generated with simple trophic (or, chemical) gradients and naturally correlated activity between neighbors. [Kaas, 1997]

Thus, if the neural representations of feelings from the body in the interoceptive pathway to the thalamus and cortex do indeed provide the basis for both modality identification and stimulus localization, as expected of primary sensory representations, then there should be a somatotopic map for each distinct feeling, for each modality, in which clusters of neurons with similar response properties maintain appropriate neighborhood relationships. The demonstration of a coherent somatotopic map for a specific feeling from the body thus provides strong support for the independence of that particular representation.

The conceptual significance of somatotopy is emphasized by the fact that the S1, the primary somatosensory cortex in the postcentral gyrus, was for years imputed to contain the cortical representations of pain and temperature sensations (and still is by some). That presumption was based primarily on the fact that it was the only known cortical area that had a somatotopic map of the body (and similarly for S2, the secondary somatosensory cortex). As explained in this chapter, that is no longer true.

The lateral spinothalamic tract.

The first neuroanatomical studies that identified ascending fibers from the spinal cord to the forebrain at the beginning of the twentieth century were made with a heavy-metal precipitation technique (the "Marchi method," which uses osmium tetroxide) that stained degenerating my-elinated fibers subsequent to damage, or "ablative lesions." Two ascending fiber bundles were followed in human and monkey spinal cord, one in the middle of the lateral white matter ("funiculus") and one in the lower (most ventral, or in humans, anterior) part of the white matter; the two bundles merged in the brain-stem and could be followed to the level of the thalamus, where the stained fibers disappeared. The two fiber bundles were designated the "lateral spinothalamic tract" and the "ventral spinothalamic tract," respectively. The clinical investigators associated them with pain and temperature and crude touch and movement, based on the effects of severing those fibers in animals and human patients. In fact, cutting the lateral tract in human patients interrupts temperature, pain, itch, sensual touch ("slow brush"), and visceral pain sensations on the contralateral body. That fits well with the experimental evidence I obtained in monkeys showing that it consists nearly entirely of ascending lamina I spinothalamic axons. It (p.135) also contains a few axons of lamina V spinothalamic neurons, but otherwise all laminae V and VII spinothalamic axons ascend in the ventral tract. [Craig, 2004a]

I saw individual PHA-L–labeled contralateral ascending lamina I fibers in the middle of the lateral white matter in obliquely cut spinal cord sections in cat and monkey, and I saw lamina V fibers in the ventral tract in cases with injections in the deep dorsal horn. My observations fit very well with the precise localization of ascending contralateral thermosensory fibers by Ulf Norrsell in behaving cats, as I mentioned earlier. By comparing the location of ascending fibers subsequent to lamina I injections at cervical and lumbosacral levels, I demonstrated the same topographic gradient of lamina I axons in the lateral spinothalamic tract at upper cervical segments that neurosurgeons had reported based on their experience with spinothalamic tractotomies (medial to lateral, hand to foot). In addition, we found that we could selectively highlight the lateral spinothalamic tract in both monkey and human spinal cord using a stain for the neurochemical calbindin, which labels about two-thirds of the lamina I spinothalamic neurons in the monkey but no other neurons with ascending fibers. The ventral tract can be visualized selectively using a stain for a different neurochemical (parvalbumin). [Craig, 1991; Craig et al., 2002]

Finding Waldo.

After I had mapped the locations of the ascending axons of lamina I neurons and their termination sites in the spinal cord and brainstem, I began a systematic examination of their termination sites in the thalamus, first in the cat (see box 4, page 146) and then in the monkey. It quickly became clear that lamina I spinothalamic terminations in the macaque monkey are concentrated in two regions: a particular portion of the medial thalamus and a portion of the postero-lateral thalamus that extended further longitudinally. The region in the posterior thalamus had variously been called the posterior group (Po) or the medial posterior region (POm) or the posterior/suprageniculate complex (Po-SG), and it had been described in print as "murky," because it is difficult to delineate specific cell groups ("nuclei") in that region. Injections in lamina I in the trigeminal dorsal horn (which receives input from face and head) produced dense terminal labeling that continued from the posterior region into a region of darkly stained cells that was usually interpreted as the posterior aspect of the trigeminal portion of the main somatosensory nucleus, the ventral posterior medial nucleus (VPM). In contrast, in cases with injections in lamina I in the spinal cord, the labeling I saw extended anteriorly from posterior thalamus into the region ventral to the so-matosensory thalamus known as the ventral posterior inferior nucleus (VPI).

In the earliest cases, I was unable to specify the location of the densest terminations in the posterior thalamus; I was seeing brightly fluorescent lamina I terminations against a uniformly dark background, and then in an adjacent section stained for cells, I was looking for a consistent structure in the "murky" sea of (p.136) neurons, guided only by the locations of small vessels, which change across sections and between brains. Like nearly all prior investigators, I had taken only a one-in-four or one-in-three, or at most, a one-in-two alternate series of sections to stain for cells (with a Nissl stain called thionin); unfortunately, the sections with the fluorescent labeling could not be Nissl-stained, because that obscured the visibility of the fluorescent-labeled terminations. (Fluorescent cell stains were not yet available.) Every section potentially contained the best fluorescent labeling that would make the best photograph, so it made sense to use as few sections as possible for the Nissl stain. But the neurons in the "murky" posterolateral thalamus looked different in different sections, and even when we Nissl-stained every other section, a consistent structure did not "pop out."

Thus, in my notes for the earliest cases, I began using the name Waldo for the lamina I terminations in the posterior thalamus, after the popular book Where's Waldo?, which my children liked. The book contained pages with large, highly detailed, color drawings, and the task was to find and recognize the small drawing of the young man Waldo with his conspicuously striped shirt in the large and very busy scenes. That captured how I was feeling. I was looking for a consistent visual pattern in a confusing scene, and I couldn't find it. I felt a childlike sense of awe and wonder seeing the beautiful labeling (see the photomicrograph in figure 8E) and considering its potential significance, and like a child, I needed to learn how to identify its location.

The labeled lamina I terminations in Waldo were distinctive, comprising compact clusters of profuse terminal arbors with grapelike complexes of large synaptic enlargements ("boutons"). Eventually, I obtained several well-labeled cases with injections in lamina I at the trigeminal, cervical, and lumbosacral levels, and by comparing them in detail, I was able to recognize a consistent pattern of anterior-to-posterior topographic organization. The pattern encompassed the lumbosacral and cervical labeling in the posterior region of the thalamus as well as the trigeminal labeling that extended from the posterior region into the posterior VPM. As I discussed above, the presence of a topographic map signified the presence of a coherent entity, and the labeling now circumscribed its entire structure. I had an outline of Waldo. I needed to understand why its structure in Nissl-stained sections had such a variable appearance.

First, I compared different views. We usually sectioned the thalamus in the coronal plane on the side with the labeling, so to get a different view, we sectioned a few cases in either of the two orthogonal planes (sagittal or horizontal), and then sectioned the thalamus on the other side in the coronal plane. The dense (p.137)

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Figure 8. The identification of PHA-L-labeled lamina I spinothalamic terminations in the VMpo.

The drawings in (A) show the locations of labeled terminals in the posterior thalamus of a monkey with a cervical injection, plotted in transverse sections (anterior at bottom) with reference to the cytoarchitectonic borders of the well-recognized nuclei indicated with numerous abbreviations (for definitions of these abbreviations, see the original paper, Craig 2004a). The pho-tomicrographs (B), (C), and (D) show the areas indicated by the dashed rectangles in (A); they demonstrate how the dense lamina I terminations guided the identification of the VMpo (arrow-heads at borders) in Nissl-stained sections within the posterior region previously described as "murky." The photomicrograph in (E) shows a region of dense fluorescent-PHA-L-labeled lamina I terminations in VMpo. The scale bar at the lower right in (D) equals 1.0 mm for (B), (C), and (D).

(p.138) terminations identified Waldo in the sections cut in the unfamiliar plane, and I could directly compare the two different Nissl-stained views, because the two sides of one brain are always fairly similar. I saw that Waldo was round and encircled by a fiber bundle in sagittal sections, and that convinced me that it is a unitary structure, but the cells looked different across sagittal sections, too.

Next, I added a drawing attachment to the fluorescent microscope (a "camera lucida"), and I plotted the labeling in each section on an accurately scaled color print made from a photograph of the adjacent Nissl-stained section, using small blood vessel holes for alignment. That enabled me to specify exactly the cells targeted by each cluster of labeled terminations, and immediately I saw a direct correspondence; each group of terminals matched precisely one distinct group of neurons. The neurons within each single group looked similar in a section, but the cells in one group didn't look exactly like the cells in the neighboring groups in the same section that also received lamina I input, nor did they look like the cells at the same location that received the labeled terminations in the next consecutive Nissl-stained sections.

Finally, because we had cut the opposite sides of those brains in the coronal plane, we had complete sets of serial Nissl-stained sections to examine in each of the three planes. After repeated close examinations going forward and backward and by comparing serial photographs, I saw that the groups of neurons are three-dimensional structures that consist of neurons with homogeneous shapes that have a quickly rolling orientation within each group, like a school of fish swimming in unison in a tight circle. Because of their rolling orientation, the cells in each group can appear thin and lightly stained in side profile in one section, but large and dark in full profile in the adjacent or next consecutive section. Notably, the groups of large, darkly stained cells anteriorly can easily be mistaken for VPM cells, which is quite near, because in one-in-four sections, it's almost impossible to distinguish their rolling orientation. However, upon close examination, their homogeneity contrasts starkly with the inhomogeneity of VPM neurons. The variegated appearance of Waldo that results from the quickly rolling orientation and the varied appearance of neighboring groups separated by small bundles of fibers are the main features that confound the cytoarchitectonic identification of this region as a coherent entity. Nevertheless, guided by the lamina I terminations, I learned to distinguish these groups of neurons as a cytoarchitectonically distinct nucleus (as shown in figure 8). Finally, I had found Waldo!

As explained in chapter 2, I named this structure VMpo (or, the posterior part of the ventral medial nucleus), because it forms a complete representation of all homeostatic sensory activity in conjunction with the VMb (the basal part of the ventral medial nucleus), which it adjoins anteriorly.1 This nomenclature is consonant (p.139) with the fact that they project to a distinct subregion of insular cortex in parallel and in topographic continuity, as I describe in more detail below.

The functional anatomical characteristics of the VMpo in the macaque monkey.

The functional and anatomical characteristics of the VMpo support the assertion that it is a modality-specific lamina I spino-thalamo-cortical relay nucleus. The detailed anatomical characteristics are as follows: the lamina I terminations consist of dense clusters of large synaptic boutons on multiple short, terminal axonal branches. The ultrastructural evidence that we obtained with the electron microscope showed convincingly that the large lamina I terminal boutons are glutamatergic; that is, at synapses they release the excitatory amino acid glutamate as a neurotransmitter. Three-dimensional reconstructions from serial ultrathin sections showed that they terminate on the proximal dendrites of VMpo relay cells in triadic arrangements, with inhibitory ("GABAergic") presynaptic dendrites having the characteristics of branches from local interneurons. Retrograde labeling studies showed that the VMpo receives spinal input almost exclusively from lamina I neurons, with the same anterior-to-posterior (face to foot) topographic order displayed by the anterograde labeling. Anterograde double-labeling experiments with simultaneous injections in lamina I of different-colored fluorescent dextrans at different levels of the spinal cord produced definitive evidence that the lamina I terminations in the VMpo are topographically organized. Its organization is visible both at the global level, with trigeminal input most anteriorly and lumbosacral input most posteriorly, and also at the local level, since single, closely spaced trigeminal lamina I injections produced closely spaced patches of terminations with the same anterior-to-posterior gradient. Those observations support the conclusion that the VMpo is somatotopically organized. These properties are all typical anatomical characteristics of a primary thalamic sensory relay nucleus. [Blomqvist et al., 1996; Beggs et al., 2003; Craig, 2004a; Craig and Zhang, 2006; Craig et al., 1999]

Obtaining microelectrode recordings to support that conclusion was considerably more difficult than recording from the raccoon's somatosensory thalamus. The somatosensory map of cutaneous mechanoreceptors in the monkey's somato-sensory thalamus, referred to here as VPM + VPL (i.e., face plus body representations), spans almost 5 millimeters in each dimension. In contrast, the VMpo is tiny, and its size varies; in the long-tailed (cynomolgus) macaque monkeys that I (p.140) studied, it extended approximately 0.6–1.4 millimeters anteroposteriorly, and about the same in each of the other two dimensions. And, it lies approximately 25 millimeters deep in the monkey brain. That made systematic recordings nearly impossible. First, I had to find the VMpo with the microelectrode tip. I began by locating the representations of the fingers, lips, and tongue (in succession) in the somatosensory map in VPM + VPL, then I made closely spaced (0.25 mm) micro-electrode penetrations successively further posterior and medial until I found, at the expected depths, units and clusters selectively responsive to cool or to noxious stimuli applied somewhere on the contralateral body. While that strategy seems straightforward, it had several practical complications.

The first neurons recorded in the VMpo in each experiment were at a serendipitous initial point of encounter. The neurons in most of the VMpo are selectively nociceptive, responding only to noxious stimulation (pinch and/ or noxious heat and/or cold) within relatively small RFs that can be located anywhere on the contralateral body. Most cells recorded together in a cluster or in a short vertical sequence have overlapping RFs, as expected for a somatotopic map, but the RFs of adjacent clusters or single units in any microelectrode penetration can be widely separated, as the electrode tip moves through different cell groups in the VMpo. The glass-coated, platinum-plated, tungsten microelectrodes that we constructed in the laboratory (tip size ~ 10 x 20 μ‎m) provided some advantage; a good microelectrode was small enough to isolate single VMpo units yet large enough so that activated neurons more than 0.1 millimeter away from the tip could still be distinguished in the background on the audio monitor. However, identifying selectively nociceptive units requires cautious application of noxious stimuli, because the stress from repeated stimulation applied too frequently eventually produces abnormal physiological conditions, even in a deeply anesthetized animal, that are signaled by excessive spindling and unresponsive neurons. Also, such closely spaced penetrations inevitably produce tissue distortion as the microelectrode travels through 25 millimeters of brain tissue to reach the VMpo, and then the tracks can be misaligned, complicating the identification of each recording site in the histological reconstruction.

Thus, over a period of almost twenty years, I recorded units in the VMpo, VPI, and surrounding regions in dozens of monkeys, both in sterile ("survival") experiments, prior to making anatomical tracer injections, and in single ("acute") recording experiments, performed to make quantitative response analyses. In some experiments, I found few or no VMpo units and clusters, whereas in others, I found one or more dozen. Altogether, I obtained recordings from 208 well-isolated single units and 680 clusters of neurons in the VMpo.

In five particular experiments, I recorded a total of 71 units and clusters, and complete three-dimensional histological reconstructions showed that nearly the entire anteroposterior extent of VMpo had been sampled. The collated data demonstrated graphically that the 38 selectively nociceptive units and clusters with (p.141) trigeminal RFs (on the face or head) were located in the anterior half of the VMpo, and the 15 units and clusters with RFs on the hindlimb or foot were located in the posterior half of the VMpo, while the 18 recordings with RFs on the forelimb or hand were recorded in between. That topography matches the anteroposterior gradient seen in the anterograde and the retrograde anatomical studies. (These charts can be seen in Craig, 2014.)

Thermoreceptive-specific cells form an additional map in the VMpo. They are located in the most medial portion of the VMpo and are most numerous at its most anterior level, just posterior and medial to the representation of ipsilateral intraoral mechanoreceptors in VPM, where they have RFs on the tongue, lips, and nose. I recorded a few warm-responsive units in this region, as well. In three propitious experiments, I obtained recordings exclusively from units and clusters with thermoreceptive-specific cooling responses in three or more microelectrode penetrations spaced 0.25 millimeters anteroposteriorly, and in every case there was an anteroposterior somatic topography of cool-sensitive units and clusters with RFs on the face, hand, and foot. In one of those, I successfully obtained anterograde tracing evidence from microinjections made at all three sites (see below).

The largest portion of VMpo is the anterior portion that contains neurons with RFs on the face. At this level, three major groups of neurons are visible histologically that are arranged mediolaterally and contain differently shaped neurons. Anterograde labeling from a single tracer injection in trigeminal lamina I usually showed three distinct bursts of terminations that were aligned with these three cell groups; that evidence suggested that they might correspond to the three major physiological classes of lamina I trigemino- and spinothalamic neurons, the thermoreceptive-specific (COOL), nociceptive-specific (NS), and polymodal nociceptive (HPC, "heat, pinch and cold") cells described in chapter 3. The unpublished cases that I am still documenting in the laboratory include several cases in which recordings were obtained in at least two of these cell groups. The evidence indicates that the units in the most medial group behave like COOL cells, the units in the middle cluster are NS-like, and the recordings in the most lateral group are HPC-like. Since the quantitative analyses showed that the response characteristics of these three types of units in the VMpo in fact match the characteristics of these lamina I spinothalamic response categories, we can conclude that these are indeed labeled lines and call them COOL, NS, and HPC neurons, too. These results provide strong support for the conclusion that the VMpo contains distinct somatotopic maps for these three labeled lines, and perhaps for each type of modality-specific lamina I input it receives.

The retrograde-labeling evidence indicates that almost all lamina I trigemino-and spinothalamic neurons (85%) terminate only in the VMpo. Based on the shapes of the retrogradely labeled cells (see photomicrograph in figure 3D), the three major categories of lamina I neurons (COOL, NS, HPC) are almost equally (p.142) represented, with each category contributing close to one-third of the entire population, and with less than 10% of unknown character. We recorded from all three types as well when we used antidromic activation to physiologically identify spinothalamic lamina I neurons that project to the VMpo. All of those findings suggest that equal portions of the ascending pathway represent each major response category; however, the recordings in the VMpo indicate an unequal representation. The recording evidence indicates that NS and HPC clusters are present throughout the anteroposterior extent of the VMpo, whereas COOL cells are found in the most medial aspect over a much smaller anteroposterior portion of the VMpo. Similarly, in the Nissl-stained histological sections, all three cell groups are readily identifiable at the most anterior level, but proceeding posteriorly, the COOL cell group essentially disappears within 0.5 millimeters, whereas the other two groups seem to intertwine and extend the full length of the VMpo. At present, I have no explanation for the disproportionately smaller relative volume of the COOL portion of the VMpo. The RF sizes of COOL units and clusters in the VMpo are not noticeably larger. The COOL cells in the VMpo are somewhat smaller and more densely packed than the other two types, but that does not seem to be a sufficient explanation. [Craig and Zhang, 2006; Zhang and Craig, 1997; Dostrovsky and Craig, 1996; Craig, 2014]

The smaller extent of the COOL map in the VMpo is also revealed by the anterograde double-labeling evidence, which showed that the most medial portion of the VMpo receives trigeminal, cervical, and lumbosacral lamina I input that is confined to its most anterior one-third (or less), whereas the topographic terminal distribution in the remainder of the VMpo extended over its entire anteroposterior length. The anterograde double-labeling evidence also identified one more patch that receives lamina I input from all three levels, located in the most anterior and dorsolateral aspect of the VMpo (at its junction with the VPM and the anterior pulvinar nucleus). It seems reasonable to suggest that the very small group of neurons in that location respond to a different type of modality-specific lamina I spinothalamic neurons, perhaps itch-specific neurons. Nevertheless, there is no further evidence to suggest that other types of modality-selective lamina I neurons are represented in the VMpo; thus, the possibility must be considered that the ascending lamina I thalamocortical pathway in the macaque monkey does not include all possible interoceptive modalities. In contrast, there is positive evidence that several additional modalities are represented in this pathway in humans (see below).

The cytoarchitectonic identification of the VMpo in the monkey was supported by immunohistochemical labeling for the neurochemical calbindin-28kD (a calcium-binding protein). As I described above, the cell bodies of approximately two-thirds of lamina I spinothalamic neurons in the macaque are calbindin immunoreactive, and so are their axons that ascend in the lateral spinothalamic tract. In fluorescent double-labeled material, I found that many clusters of anterogradely (p.143) PHA-L–labeled lamina I terminations in the VMpo were coextensive with a moderately dense plexus of calbindin-positive terminal fibers. At high magnification, I was able to capture a few photographs of PHA-L–labeled lamina I terminals in the VMpo that were also immunoreactive for calbindin (figure 9C and D); I had to be quick, because at high magnification the intense illumination usually bleached the fluorochrome completely within 5–10 seconds. Such double-labeling documentation validated the identification of clusters of calbindin-positive terminal bursts as lamina I terminations. However, within the cytoarchi-tectonic borders of the VMpo, not all cell groups contain calbindin-positive fiber labeling. Notably, the immunoreactive terminal clusters in the VMpo are easily distinguishable from a contiguous, extremely dense calbindin-positive fiber plexus that is coextensive with the VMb, the other half of the interoceptive pathway (see below). [Craig, 2004a, 2011b; Craig et al., 1994]

These neuroanatomical and neurophysiological findings support the conclusion that the VMpo constitutes a dedicated lamina I spino-thalamo-cortical relay nucleus. Based on the functional characterization of lamina I spinothalamic neurons as "pain and temperature" neurons and on our own recordings of VMpo neurons, we first suggested that the VMpo specifically represents pain and temperature. However, it makes more sense to view this as an interoceptive pathway, as discussed in more detail further below.

Finally, whereas the VMpo is relatively tiny compared to the somatosensory nuclei VPM + VPL in the macaque monkey thalamus, and only primordially represented in nonprimate mammals, if at all (box 4), it is proportionately approximately two times larger in the human thalamus (figure 9A). On average, it is roughly l × l × l millimeter in the macaque monkey, and 4 × 4 × 2 millimeters in the human. Its volume varies over the range 5–10% as a proportion of the volume of VPM + VPL in the macaque, and over the range 14–20% in the human. In work that we presented at a large meeting, we measured that volume ratio in eighteen brains representing seven different anthropoid (catharrhine) primates (Old World monkeys, great apes, and humans), and we found a highly significant linear relationship (R2 = 0.98) with a log-log ("allometric") scaling factor of 1.14. These data support the idea that as humans evolved, the VMpo enlarged relative to VPM + VPL. In fact, there are two obvious disjunctions in the cytology of the VMpo across all primates. First, in strepsirhine primates (lemurs, lorises, bushbabies) and tarsiers, there is only a very small region that could be the VMpo based on neighborhood relations, and it contains only plump, round cells.2 That contrasts with the haplorhine long-tailed macaque and rhesus macaque, as well as the orangutan, in which the VMpo is clearly visible and contains at least three distinct cell types. In turn, that pattern contrasts with the structure of the (p.144)

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Figure 9 Calbindin-labeled terminations in the VMpo.

The photomicrographs in (A) show the location of the human VMpo (arrowheads) in adjacent Nissl- (left) and calbindin-stained (right) transverse sections; the photo in (B) shows the calbindin terminal-fiber-labeling at the boxed region in (A) at higher magnification. Photomicrographs in (C) and (D) show a double-labeled (C, calbindin; D, PHA-L; green and red, respectively, in plate 5) lamina I spinothalamic terminal in the monkey VMpo at high magnification.

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Plate 5 Calbindin-labeled terminations in the VMpo.

The photomicrographs in (A) show the location of the human VMpo (arrowheads) in adjacent Nissl- (left) and calbindin-stained (right) transverse sections; the photo in (B) shows the calbindin terminal-fiber-labeling at the boxed region in (A) at higher magnification. Photomicrographs in (C) and (D) show a double-labeled (C, calbindin; D, PHA-L; green and red, respectively, in plate 5) lamina I spinothalamic terminal in the monkey VMpo at high magnification.

(p.145) VMpo in the common chimpanzee, pygmy chimpanzee (bonobo), gorilla, and human. In all of these hominid species, the VMpo is clearly enlarged, it has a more variegated and complex appearance, and it is better delimited in all dimensions. In all of these species, the VMpo visibly enwraps the ascending spinal fibers at their entry into thalamus (figure 9A and B; see also figure 6A and B in Blomqvist et al., 2000). These observations support the idea that the VMpo in humans is qualitatively significantly more developed than it is in the macaque monkey.

The projection from the VMpo to the dorsal posterior insula in the macaque monkey.

We obtained anterograde labeling evidence from tracer injections in VMpo in some of the very first recording experiments. It projects to the center of the sharp fold of cortex which forms the upper morphological margin of the insula; in precise neuroscientific terms, that is called the functus of the superior limiting (peri-insular) sulcus (SLS). A year later, we even had the initial double-labeling evidence that revealed the anterior-to-posterior (face to foot) topography of this projection, as I showed in preliminary illustrations. That was lucky; it took much longer to acquire comprehensive evidence. [Craig, 1995, 2000]

The entire set of tracing experiments performed over a period of almost twenty years provide much greater insight on the projection of the VMpo to the dorsal posterior insular cortex in the long-tailed macaque monkey. The cortical projections of the VMpo occupy an anatomically distinct granular cortical area centered in the fundus of the SLS. The limits of the projections match the medial and lateral borders of this area, and they span the entire posterior half of the SLS in the macaque, from its origin at the posterior limit of the insula to its approximate middle (near the lateral end of the central sulcus). The anterior half of this distinct area extends forward to the anterior end of the SLS, where it disappears into orbitofrontal cortex. That anterior portion receives the terminations of the VMb (see next section). This strip of cortex had not been recognized as a distinct area of cortex by prior investigators, and unfortunately, part of it had been included in S2.

Studies of connections with somatosensory areas prior to the identification of the VMpo had included the fundus of the SLS in the S2, which extends across the wide bank of cortex ("parietal operculum") of the deep fold that covers the insula (the "Sylvian fissure," or lateral sulcus). That was based in part on tracing evidence from injections in the thalamus that had been mislocalized to that dark-celled posterior portion of the VPM that is actually the anterior portion of the VMpo. As a result, the fundus of the posterior SLS is erroneously labeled as part of the S2 in many current atlases of both monkey and human brain, and that designation has confounded the interpretation of modern functional imaging studies in humans. Nevertheless, modern evidence is consistent with our

findings; the twin S2 fields occupy the full width of the parietal operculum but do not extend into the fundus in both the macaque monkey (called S2/PV) and human (called OP1/OP4). Furthermore, the face in that map, which is labeled from the VPM, is located most laterally, approximately 20 millimeters from the fundus of the SLS in humans. Similarly, the cortex in the fundus of the anterior SLS was erroneously divided into gustatory cortex and premotor areas by prior investigators. [Krubitzer et al., 1995; Eickhoff et al., 2007; for a detailed explanation and more references, see Craig, 2014]

A reexamination of the anatomical characteristics of this strip and the entire insular cortex in the macaque monkey is described in a study that was just reported, which I performed in collaboration with Henry Evrard (who was for several years a postdoctoral scientist working in my lab and is now a professor at the University of Tübingen in Germany). We made a rigorous cytoarchitectonic analysis, which means that we compared the cellular structure in Nissl-stained sections and the neuropil structure in fiber-stained sections, and we identified consistent features of the laminar architecture across eleven monkeys. We documented the presence of 15 distinct areas, of which there are 4 granular, 4 dysgranular and 7 agranular areas (see figure 10C). We recognize the entire fundus of the SLS (p.149)

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Figure 10 The organization of the insular cortex in the monkey and the human.

The left insula is exposed in a macaque (A; MRI reconstruction) and a human (B; formalin-fixed) brain. The drawings in (C) and (D) show the current architectural maps in the monkey (C) and the human (D). The shaded regions that surround the SLS in (C) denote the interoceptive cortex; the functional subdivisions shown represent the preliminary evidence described in the text. The interoceptive cortex in the human architectural map (D) of the posterior insula probably lies within the pale area that surrounds the circular sulcus at the top ("OP3"). The first resting-state functional connectivity maps reported in humans (E) differentiate the posterior and anterior portions of the left and right insulae, but only partially the mid-insula. The more complex maps in the partially unfolded three-dimensional renderings in (F) show the density of connectivity foci in the left and right insulae reported in a recent meta-analysis of 1305 fMRI studies. The small boxes indicate regions of the highest density ("hubs"), which have interconnections throughout the brain on each side.

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Plate 6 The organization of the insular cortex in the monkey and the human.

The left insula is exposed in a macaque (A; MRI reconstruction) and a human (B; formalin-fixed) brain. The drawings in (C) and (D) show the current architectural maps in the monkey (C) and the human (D). The shaded regions that surround the SLS in (C) denote the interoceptive cortex; the functional subdivisions shown represent the preliminary evidence described in the text. The interoceptive cortex in the human architectural map (D) of the posterior insula probably lies within the pale area that surrounds the circular sulcus at the top ("OP3"). The first resting-state functional connectivity maps reported in humans (E) differentiate the posterior and anterior portions of the left and right insulae, but only partially the mid-insula. The more complex maps in the partially unfolded three-dimensional renderings in (F) show the density of connectivity foci in the left and right insulae reported in a recent meta-analysis of 1305 fMRI studies. The small boxes indicate regions of the highest density ("hubs"), which have interconnections throughout the brain on each side.

(p.150) as a continuous strip of granular cortex that we call the dorsal functus of the insula, which has distinct posterior and anterior halves (Idfp and Idfa). The coherent, topographic projections of VMpo + VMb to this longitudinal cortical strip identify it as interoceptive cortex (see below). Based on a synthesis of prior functional and connectional evidence, we suggest that it serves as the primary organizational axis for most of insular cortex, which contains parallel longitudinal dysgranular strips that could constitute successive levels of interoceptive integration with different types of ongoing emotional behavior (see next section).3 [Evrard et al., 2014]

The anteroposterior topography of the VMpo projection to the Idfp is documented in another article using evidence from six particular experiments selected from the entire data set. In each of these six experiments, a single, precise injection of tracer was made in the VMpo, guided by microelectrode recordings of units and clusters responsive selectively to noxious mechanical and/ or heat stimulation. Each injection was made at a different level of the anteroposterior (face to foot) topographic map in the VMpo. Each anterograde tracer injection in the VMpo produced a single coherent field of dense, patchy terminal labeling across the entire width of the Idfp. In all cases, the labeling was found at an anteroposterior location consonant with the anteroposterior location of the microinjection in the VMpo. Thus, together these cases show that the VMpo projection to the Idfp is organized along an anteroposterior (face to foot) topographic gradient. Since a single coherent field was observed for each single injection, these data also support the presence of a point-to-point topological order (a "homotopy"), consistent with the conclusion that the entire projection is somatotopically organized. In sum, these data strongly suggest that a somatotopic representation of selectively nociceptive lamina I spinothalamic activity exists in the dorsal posterior insular cortex of the macaque monkey and that it is organized along an anteroposterior gradient. [Craig, 2014]

The organization of the dorsal posterior insula in the macaque monkey.

Jonathan Dostrovsky (University of Toronto, Canada) and I reported our first microelectrode recordings from nociceptive-specific neurons in the fundus of the SLS of the barbiturate-anesthetized macaque monkey at the annual meeting of the Society for Neuroscience in 1996. Those experiments were far more difficult than recording from VMpo neurons. We drove the microelectrode deep into the cortex, crossing folds that we could not see, with almost no physiological guideposts. (Figure 10A shows the location of insular cortex in a cutaway magnetic resonance image, (p.151) MRI, view of the macaque cortex.) The neurons at most recording sites were unresponsive to every kind of stimulation we tried. We isolated and characterized only nineteen neurons and five clusters at histologically verified recording locations in the Idfp. Most were nociceptive-specific units responsive to pinch, or to pinch and noxious heat, with medium-sized RFs. They were clearly distributed with an anteroposterior topography, as documented in a recently published theoretical article. [Dostrovsky and Craig, 1996; Craig, 2010]

Ten years later, I recorded laser-evoked potentials (LEPs) from the scalp of deeply anesthetized macaque monkeys with two collaborators from Germany, Rolf-Detlef Treede and Ulf Baumgärtner (University of Mainz), which provided much more convincing evidence.4 We used a carbon dioxide laser, which my collaborators had shown is a selective noxious heat stimulus in humans that elicits simultaneous, time-locked activation of several AS nociceptive fibers without any confounding activation of nonnoxious sensory fibers at short latencies. We used an electrode placement scheme on the monkey scalp with a pattern similar to the standard electrode locations used in humans (see figure 11A). The LEPs we recorded had several detailed characteristics that match the characteristics of LEPs recorded in awake humans: at low intensities both early and late (AS- and C-fiber) components were present with a bilateral negativity in the temporal leads (the "N1" wave) concurrent with a midline frontal positivity; the AS component was graded with stimulus intensity, and at higher intensities it occluded the C-fiber component; and the surface topography was consistent with a deep source in the temporal region (fig. 11B and C). After my colleagues returned home, they applied a computational model of a pediatric human head and calculated three-dimensional source reconstructions. The results provided remarkably clear evidence that is consistent with the presence of a nociceptive-specific current source in the dorsal posterior insular cortex, as illustrated in figure 11C.

In addition, by applying the laser in turn to the foot, hand, and ear, we were able to demonstrate an anteroposterior (face to foot) topography (shown in figure 11C), which is consistent with the topographic gradient of VMpo projections to the Idfp and clearly distinct from the mediolateral topographic gradient of the somatosensory cortices. Such clear evidence of a somatic topography has not been obtained yet with LEP recordings in humans. It seems likely that we succeeded because the brain of the macaque monkey is much smaller, making the main source in the dorsal posterior insular cortex a much larger proportion of the entire brain than it is in the human brain and so putting our recording electrodes that much closer to the source. My collaborators recently reported comparable topographic evidence in the human brain for noxious heat stimuli produced with (p.152)

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Figure 11 The anteroposterior somatotopic organization of laser-evoked potentials (LEPs) in the macaque monkey.

The photo (A) shows the positions of the central five electrodes (of eight) on the scalp of a macaque monkey (nose at top; the locations of the two lateral electrodes on each side are not (p.153) a laser, not by reconstructing LEP sources, but rather, by using functional magnetic resonance imaging (fMRI) in a high-resolution scanner (see below). [Baumgärtner et al., 2006, 2010]

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Plate 7 The anteroposterior somatotopic organization of laser-evoked potentials (LEPs) in the macaque monkey.

The photo (A) shows the positions of the central five electrodes (of eight) on the scalp of a macaque monkey (nose at top; the locations of the two lateral electrodes on each side are not a laser, not by reconstructing LEP sources, but rather, by using functional magnetic resonance imaging (fMRI) in a high-resolution scanner (see below). [Baumgärtner et al., 2006, 2010]

More recently, a postdoctoral scientist, Jin Huang from Beijing, China, joined me in using an intracortical multielectrode array (2 x 8, 1 mm spacing) to identify the main focus of anteroposteriorly organized LEPs as the fundus of the SLS. After fine-tuning the microelectrodes to achieve just the right combination of sensitivity and selectivity, we identified in several deeply anesthetized monkeys three different LEP sources: a large source in the fundus of the SLS that was clearly anteroposteriorly topographic, a small source more posterior in the lateral sulcus, and another small source more dorsal in the fundus of the central sulcus (area 3a).5 Our yet unpublished data include one particularly remarkable experiment, in which we made a small retrograde tracer injection at the focus of the large LEP elicited by laser stimulation of the medial hand, which we believed was in the Idfp. After waiting several days to allow transport of the tracer, we reanesthetized the monkey, and before the perfusion with fixative, we performed a second recording experiment, in which we drove a single microelectrode into the thalamus. There, we identified an LEP activated from the same location on the medial hand, and we moved the electrode until we found its focus, which was located at the site of VMpo neurons that responded to pinch and noxious heat stimulation of the medial hand. This LEP began approximately 10 milliseconds earlier than the large cortical LEP, which is an appropriate delay ("latency") for a direct thalamocortical projection. Finally, after we finished the histological processing, we saw that the tracer injection made at the source of the large cortical LEP was indeed located in the fundus of the SLS. In the thalamus, we saw a cluster of retrogradely labeled VMpo neurons at the location of the LEP focus we had identified. That evidence shows very clearly that the cortical projection of the VMpo to the Idfp in the fundus of the posterior SLS is the origin of the large LEP, which corresponds with the anteroposteriorly topographic LEP that we had recorded in the prior experiment from the monkey's scalp. These findings imply that the same is true for the major LEP recorded from the human scalp. [Huang and Craig, 2007, 2008]

(p.154) As mentioned above, the unpublished cases that I am still analyzing in the laboratory include anterograde double- and triple-labeling experiments in which recordings and microinjections were made in two or more of the major physiological cell groups in the VMpo (COOL, NS, and HPC). These require minutely detailed analysis, which is not yet complete. The evidence compiled to date supports the straightforward interpretation that is illustrated in figure 10C. The three categories of modality-specific neurons—COOL, NS, and HPC cells—are each represented in a separate somatotopic map in the VMpo that is organized in the anteroposterior dimension, and each type projects topographically to a separate portion of the Idfp. The NS and HPC neuron groups project to the medial and lateral walls of the SLS, respectively, extending anteriorly from its posterior end nearly to its midpoint. The border between them lies at the center of the fundus, where a cytoarchitectonic boundary in the Idfp can often be seen in a Nissl-stained section (see figure 3 in Evrard et al., 2014). (This morphological boundary is probably also the point of glial attachment that serves as the tensile anchor which produces the fold of the SLS during the development of the brain, as discussed near the end of this chapter.) The HPC map extends all the way to the anterior end of the Idfp, where it adjoins the Idfa at the anteroposterior middle of the SLS, but the NS map does not extend that far. At the anterior level of the broadening of the central sulcus, the NS map is replaced by the somatotopic COOL map, which is much shorter in the anteroposterior direction. This interpretation is supported by several double- and triple-labeling cases with microinjections at physiologically well-identified locations in different cell groups in the VMpo, and by cases with microinjections at different somatotopic levels in each one of the groups, including the COOL neurons. In one very fortunate experiment, three microinjections were made at different somatotopic locations in the VMpo COOL group, which produced three fluorescent-labeled terminal projections that were arranged in somatotopic sequence in the small, anterior portion of the medial wall of the Idfp that is assigned to COOL cells in figure 10C.

The larger injections described above that demonstrated the overall topography in the VMpo produced labeling across the entire width of the Idfp, and the labeling in the medial wall was slightly more posterior than the labeling in the lateral wall. That observation is consistent with this modality arrangement in the Idfp, because the HPC map is longer than the NS map. In a few of the cases with a larger injection in the VMpo, a disjunction in anterograde labeling was apparent at the midpoint of the fundus of the SLS, which is consistent with the differential involvement of only one of the separate nociceptive maps (e.g., cases ml37R and m58R in figure 10 of Craig, 2014).

Thus, the evidence supports the conclusion that there is a separate and distinct somatotopic map for each of the three major types of modality-specific lamina I spinothalamic neurons in the VMpo and in the Idfp. These data provide over-whelming support for the idea that these are indeed labeled lines, which in humans (p.155) correspond with qualitatively distinct feelings. The putative fourth lamina I modality in the VMpo indicated by the convergent double-labeling in its extreme anterior and dorsolateral corner (mentioned above) must be represented at the anterior end of the HPC map in the Idfp, which is consistent with the overall topography of the entire projection. Different submodalities of NS and HPC cells, such as the NS cells responsive to sharpness or the HPC cells that are muscle responsive, are likely embedded within the present schematic as discrete longitudinal strips, instead of occupying a segment at one end, analogous to the COOL cells.

There are ancillary projections from the VMpo to the dysgranular areas in the middle of insular cortex, as well as to the fundus of the central sulcus (area 3a) and the fundus of the cingulate sulcus (area 24c; this is also the cortical terminus of the ascending lamina I projections to the ventral caudal part of the medial dorsal nucleus of thalamus, MDvc, described in box 5). The projection to area 3a conveys all three major response types, and it is topographically organized me-diolaterally, in parallel (but not in register) with the somatotopy of S1 in area 3b. Investigators in another laboratory reported augmenting responses to quickly repeated noxious heat stimuli in area 3a, which is indicative of HPC input, and they suggested that area 3a might serve a role in the localization of pain sensations in humans. That seems possible, and it could explain particular clinical observations; however, area 3a is the selective cortical terminus of proprioceptive, lemniscal Group I joint and muscle spindle inputs relayed by VPM + VPL, and it also receives inputs from vestibular and vagal-activated sources, in addition to the input from all three major types of lamina I and VMpo neurons. Finally, it is a major source of corticospinal projections. To my mind, these facts strongly suggest that area 3a has a major role in cortical modulation of spinal motor activity. The localization of pain sensations can be supported by the Idfp, according to present data and other clinical findings (see below). Nevertheless, the presently available evidence suggests that area 3a is closely interconnected with the Idfp and area 24c, and thus it likely has an integral role in thalamocortical processing of all three major interoceptive modalities, if not all (see also the final section of this chapter). [Tommerdahl et al., 1996; Ito and Craig, 2003; Vierck et al., 2013]

The interoceptive pathway.

As outlined in chapter 2, these findings in the macaque monkey suggest that viewing the ascending lamina I pathway through the VMpo as a "pain and temperature" pathway is too simplistic. Rather, it makes more sense to regard it as part of a thalamocortical representation of homeostatic sensory activity, together with the VMb. In the thalamus of anthropoid primates, the VMpo is contiguous at its anterior end with the VMb, as seen in all three planes. These two anatomically conjoined structures both relay homeostatic sensory

activity, that is, interoceptive activity. The combined structure, VMpo + VMb, forms a topologically coherent structure that receives direct, high-resolution homeostatic sensory input from the body and projects to the entire granular cortical strip at the fundus of the SLS, the VMpo to its posterior half (Idfp) and the VMb to its anterior half (Idfa). The entire strip identified by color in figure 10C, that is Idfp + Idfa, is interoceptive cortex.

In all mammals, the VMb receives (mainly) ipsilateral, viscerotopically organized input from the parabrachial nucleus (PB) in the mid/upper brainstem, and it projects to the gustatory and insular cortices, which send descending controls to the PB, solitary nucleus (NTS), and lamina I. The viscerotopic map in the PB originates in part from the ipsilateral NTS in the lower medulla, which receives sensory input from the cranial autonomic (parasympathetic) nerves (i.e., facial, glossopharyngeal, and vagus). There is also input to the PB from lamina I neurons in the spinal dorsal horn (mainly contralateral) and the trigeminal dorsal horn (mainly ipsilateral), which is weakly topographic. As described in chapter 4, the PB is the major brainstem homeostatic sensory integration site in all mammals; it sends viscerosensory and gustatory activity from the NTS to the VMb, and thermosensory and nociceptive activity from lamina I mainly to the hypothalamus and the extended amygdala (see box 4), but also to the VMb. The PB is an obligatory gustatory relay in all mammals, except in anthropoid primates, where the VMb instead receives direct gustatory input from the NTS, and little or no gustatory input is processed in the PB. [Saper, 2002; Beckstead et al., 1980]

It is particularly striking that both the direct NTS input to the VMb and the direct lamina I input to the VMpo are phylogenetically unique to primates; the entire combined high-resolution homeostatic sensory pathway is a primate innovation. Morphologically, the operculated insula formed by the SLS is present only in anthropoid primates (monkeys, apes, and humans). That makes sense if the cortical terminus of this pathway, interoceptive cortex, serves as the tensile anchor that produces the cortical fold which becomes the SLS, because this pathway (p.158) is present only in anthropoid primates and it is not present in carnivores or rodents. In the macaque monkey, the VMpo receives via lamina I neurons the homeostatic sensory input from all spinal autonomic, sympathetically innervated tissue sources in the body, and the VMb receives via NTS and PB neurons the complementary homeostatic sensory input from all cranial autonomic, parasympathetically innervated tissue sources. The posterior-to-anterior topographic order in the VMpo continues through the VMb, where vagal and gustatory inputs are also differentially organized from posterior to anterior (as well as dorsoventrally). The projections from the VMpo to the Idfp and from the VMb to the Idfa in the fundus of the SLS are similarly organized with an overall posterior-to-anterior gradient. [Pritchard et al., 1986; Carmichael and Price, 1995]

Thus, the combination VMpo + VMb represents virtually all homeostatic sensory inflow and forms a posterior-to-anterior column in the thalamus that is orthogonal to the medial-to-lateral somatotopic gradient of the mechanoreceptive and proprioceptive representations in the VPM + VPL, to which it is connected at the representation of the mouth (which is the embryological junction of the inside and outside of the body). These two topographic gradients must develop from two distinct neurotrophic chemical gradients that represent a fundamental neurobiological and genetic distinction. Just as the interoceptive and exteroceptive sensory inflow is processed by two distinct dorsal horn regions in the spinal cord and then sent to the forebrain in two distinct ascending pathways, in primates there are also two distinct thalamocortical substrates that process interoceptive and exteroceptive sensory inputs at the highest resolution possible. The interoceptive role of Idfp + Idfa is directly supported by evidence in the macaque monkey that it is delimited by labeling for the receptors of corticotropin-releasing factor, which many investigators regard as a definitive marker for the neural substrates of homeostasis. [Craig, 2014; Evrard et al., 2014; Sanchez et al., 1999]

In the macaque monkey, however, this pathway may not convey all interoceptive activity to the insular cortex at high resolution. Approximately 85% of the lamina I cells that project to the VMpo are COOL, NS, and HPC cells, which seems to support the simple assertion that the VMpo is a specific substrate for discriminative pain and temperature sensations. In contrast to the evidence in humans described below, there seems at best to be only a small representation of other modalities, such as itch, vasculature distension, and C-fiber touch. The possibility that the interoceptive pathway in monkeys is mainly used for discriminative sensation is also supported by evidence we obtained in nocturnal New World owl monkeys. We found that they have an enlarged outgrowth of trigeminothalamic COOL lamina I neurons, which forms a specialized pathway to the VMpo that conveys high-acuity perinasal thermosensory activity. We recognized that this pathway is probably important specifically for following scent-marked arboreal trails, thereby enhancing survival in the pitch-dark rain forest. That interpretation suggests that this specialized discriminative thermosensory pathway primarily (p.159) serves a behavioral motoric role, rather than an autonomic role. Thus, the high-resolution cortical representations of COOL, NS, and HPC activity in the Old World macaque monkey may also be most important as discriminative sensations that support motor behavior (and thus might be regarded as serving exteroceptive functions; see box 3). [Craig et al., 1999]

Similarly, the investigators who identified the direct NTS pathway to the VMb in the macaque originally regarded it as a specific substrate for discriminative taste (gustation). That conclusion was suggested by the observation that a retrograde tracer injection in the VMb labeled only cells in the anterior third of the ipsilateral NTS, which contains gustatory neurons, but very few in the intermediate NTS and none in the posterior third of NTS, which contain viscerosensitive neurons. These considerations together seem to support the notion that the ascending homeostatic sensory pathway in the monkey serves primarily the discriminative sensations of taste, pain, and temperature by way of direct, high-resolution inputs to the VMpo + VMb from lamina I and NTS. [Beckstead et al., 1980]

Evidence that this perspective is in fact too simple is presented in the next paragraph, but it's important to note that this perspective does raise an interesting and significant hypothesis for consideration. Human feelings from the gut are normally blunt, dull, and difficult to localize, which contrasts starkly with the bright, well-differentiated nature of our sensations of taste, pain, and temperature. If there is such a paucity of direct input to the VMb from the posterior two-thirds of the NTS in humans, then the visceral feelings from the gut would be supported by the evolutionarily older, indirect ascending pathway to the VMb from the PB, while the discriminative sensations of taste, pain, and temperature would be supported by the evolutionarily newer, direct, high-resolution pathway. That arrangement might explain the contrast between visceral feelings and discriminative taste, temperature, and pain. Perhaps it could also explain why feelings of visceral pain are bright, since they depend on ascending activity from lamina I spinothalamic neurons, and why the feelings of central pain are regarded as dark and unspecific, since they must depend on indirect ascending input from the PB (see box 4). This potential explanation certainly deserves further exploration; for example, it might also explain why a toothache is so overwhelming, since both the indirect projections from the trigeminal lamina I to the ipsilateral VMb by way of the ipsilateral PB and the direct projections from lamina I trigeminothalamic neurons to the contralateral VMb would be strongly activated. These considerations also emphasize the need for future studies to address the functional consequences of mixed laterality in this dual-layered pathway, since the direct NTS and the PB projections to the VMb are almost entirely ipsi-lateral, whereas the direct lamina I projections to the VMpo and the indirect lamina I projections to the VMb via the PB are almost entirely contralateral. [Beckstead et al., 1980; Pritchard et al., 1999, 2000; Ito and Craig, 2007]

(p.160) Yet ascending interoceptive activity is somewhat more complicated in the monkey and considerably more complicated in the human. When we examined retrograde labeling from the VMb in the macaque monkey with a more sensitive tracer than the one used twenty years earlier, we confirmed the dense input from the anterior third of the NTS, that is, the gustatory portion, and the sparse input from the posterior NTS, but we found weak to moderate labeling in the intermediate NTS, which contains overlapping inputs from most, if not all, visceral sources. In addition to the ipsilateral inputs from the NTS and PB, we also observed weaker input to the VMb from these nuclei on the contralateral side. Consistent with that anatomical evidence of mixed laterality, we also recorded vagal-evoked potentials of comparable size in the neighborhood of the VMb on both right and left sides, from both the right and left vagus nerves. [Ito and Craig, 2005, 2007]

The ecological and evolutionary value of the high-resolution sensory representations of taste, temperature, and pain in macaque monkeys must certainly stem from their significance for discriminative motoric behavior. Nevertheless, these interoceptive modalities are conveyed in a distinct ascending pathway, and this pathway has an intrinsic morphological (i.e., genetic) identity as a homeostatic sensory substrate and conveys additional types of interoceptive activity even in the monkey. To my mind, the inherent capacity of this pathway to provide a complete representation of all aspects of the physiological condition of the body in interoceptive cortex is of the greatest significance, because that capacity led to the integration of homeostatic salience in the insular cortex that underlies the phenomena of homeostatic sentience and subjective awareness in hominids, as described in the next section and in chapters 6 and 7.

The human VMpo.

The earliest reports that pain could be elicited in awake human patients upon microstimulation in the thalamus came in 1959–60 from Rolf Hassler, a neurosurgeon in Germany. He made these observations while implanting a stimulation electrode in the thalamus for the treatment of movement disorders. The pain reports were obtained only when the electrode was positioned at the lower, rear (posterior and inferior) margin of the main somatosensory nucleus (which is called Ventrocaudalis or Ventralis caudalis, abbreviated V.c; it is equivalent to the VPM + VPL in the monkey). He clearly noted the contrast between these reports of well-localized pricking or burning pain sensations and the reports of tingling or buzzing ("pareasthesia") that the patients gave upon stimulation within the main somatosensory thalamus. He initially associated the pain-related region with dense spinothalamic terminations that he had identified in the posterolateral thalamus of a baboon using silver-stained fiber degeneration after cutting the spinothalamic pathway; he named that region the Ventrocaudalis portae (V.c.po.), which literally means the "gateway" (or, portal) to the main so-matosensory nucleus. That region is certainly the VMpo; not only is that the appropriate (p.161) location, but also the description and the name match the histological appearance of the VMpo in the thalamus of humans and great apes, because it visibly enwraps the lemniscal and spinothalamic fibers entering posterolateral thalamus, as we illustrated in our cytoarchitectonic description of the human VMpo and as shown in figure 9A (see also figures 6A and 6B in Blomqvist et al., 2000). The same region was later described independently by two neuroanato-mists, each of whom reported a dense field of silver-stained, thin-fiber terminations in a cytoarchitectonically distinct structure in a human patient with spino-thalamic damage. Later, however, Hassler associated pain with a region he named the Ventrocaudalis parvicellularis (V.c.pc.), or the small-celled region of the main somatosensory nucleus, which is equivalent to the monkey's VPI (see box 6). [Hassler and Riechert, 1959; Hassler, 1960, 1970; Mehler 1966; Mesulam, 1979; Blomqvist et al., 2000]

Subsequent clinical reports published by several neurosurgical teams corroborated Hassler's reports of pain evoked by stimulation in the region just behind and under (posterior and inferior to) the main somatosensory nucleus of the human thalamus. One such team worked in Toronto, under the direction of the neurosurgeon R. Tasker, and Jonathan Dostrovsky (University of Toronto) was the lead neurophysiologist for their stimulation and recording studies. Jonathan and I had been exchanging visits to each other's laboratory for several years, spending long days and evenings together to study the projections of thermoreceptive-specific COOL lamina I neurons, using microelectrode recordings and antidromic activation, one at a time. During one of my visits, they invited me to be present in the operating room for a human procedure. I jumped at the opportunity, of course. [Dostrovsky and Craig, 1996; Craig and Dostrovsky, 2001; see Craig, 2014]

The microelectrodes that Jonathan made for use in humans produced excellent recordings, and the sounds from the audio monitor as they touched the patient's arm to map the mechanoreceptive RFs were very reminiscent of the recordings we were making from the cat thalamus in Jonathan's laboratory, as well as of my earliest experience mapping the somatosensory thalamus in the raccoon. As they slowly advanced the electrode, the RFs progressed as usual down the arm and onto the hand and the fingers. When we realized that the recording microelectrode had reached the posterior and inferior margin of the somatosensory thalamus (V.c), Jonathan and I both looked at each other expectantly; and when Jonathan activated the brief (50 μ‎sec), tiny (5 microampere, or μ‎A) electrical stimulus, the patient immediately grabbed her upper arm and said, "Ow, that hurts!"

"Oh, wow!" I said, silently.

The Toronto team elicited both sharp and burning pain, and cool sensations as well, with microstimulation in that region in numerous patients. The reports of cool sensations contrasted starkly with the warm sensations that they and others had uniformly elicited upon stimulation of ascending spinothalamic fibers (which

must have resulted from a stimulation-induced interference or inhibitory effect on COOL fibers). They also recorded thermoreceptive-specific (COOL) units at the same sites from which they elicited reports of cool sensations in awake patients, and they rightfully attributed those to VMpo neurons. Single-unit recordings from nociceptive-specific neurons were reported in the same region by the Toronto team and by others. Nevertheless, no one has yet obtained definitive histological reconstructions of such recording sites that are comparable to the precise neuroanatomical correlations that we made in experimental animals. [Tasker, 1982; Dostrovsky et al., 1992; Lenz et al., 1993; Davis et al., 1999; for further references, see Craig, 2014]

Attempts to make such a structure/function correlation and to identify the VMpo anatomically in human thalamus were made by several groups in individual stroke patients, in whom a small hole in thalamus ("lacunar infarct," due to a vascular failure) visible on magnetic resonance images had produced thermanes-thesia (and central pain; see box 5). Although those attempts had insufficient resolution and didn't associate the VMpo with specific cool sensations, a recently reported study used a rigorous mathematical alignment method to merge the images from a group of ten such patients, and the authors reported that the critical structure is the region named Ventrocaudalis portae in a reference atlas. They identified the epicenter of such lesions at standardized MRI (x, y, z) millimeter coordinates (–14, –23, 0 or 1). That is essentially identical to the point that we reported as the center of the 4 × 2 × 4 millimeter area that we identified histo-logically as the human VMpo, which in their notation is (–14, –22, 0). [Sprenger et al., 2012; for further references, see Craig, 2014] (p.164) Our anatomical examination of the VMpo in the human thalamus began when Anders Blomqvist (Linköping University, Sweden) asked me during a telephone call, "Would you like to see VMpo in the human thalamus, Bud? I know where I can get some fresh, properly fixed brains." We had been collaborating already for several years, mainly by sending each other boxes of stained slides, occasionally accompanied by a technician, a student, and/ or a postdoctoral scientist. But this time we worked together for many days, taking turns looking through the microscope, trading criteria for identifying each group of cells, spending hours in the darkroom making high-contrast prints from 4- x 5-inch black-and-white negatives and probing the limits of each other's wit, knowledge, patience, and integrity.

Anders came to my lab for two weeks after he and his staff had collected, processed, and shipped a total of ten brains to my lab. There were at least two well-stained brains cut in each of the three major planes of section, and some of the stained sections were very impressive. The best sections through the VMpo that were stained for calbindin could be held up to a well-lit window at arm's length, and the large, fuzzy, brown-colored blotch that could be seen very clearly in the middle of a series of sections was the dense calbindin-positive fiber plexus in the VMpo. One of the adjacent pairs of Nissl-stained and calbindin-stained sections that we showed in black and white in our paper is shown in color in figure 9A. As you can see, the VMpo does not jump out in the Nissl-stained section, but the large region of dense terminal labeling in the calbindin-stained section is very conspicuous (shown at higher magnification in figure 9B).

We used the calbindin-terminal-fiber labeling as a guide, based on the staining in the macaque brain, where the PHA-L–labeled lamina I terminations had been the definitive guide. Using photographs and both low-power and high-power microscopes, we compared different cell groups in different brains cut in different planes. At times, we both could easily differentiate the VMpo in the Nissl-stained sections, but sometimes there was plenty of discussion, over several days even, before we agreed on the outlines of the thalamic nuclei that we drew for each brain.

As in the macaque, the human VMpo consists of multiple small groups of neurons traversed by multiple small fiber bundles, but in contrast to neighboring nuclei, the VMpo neurons in each group have a homogeneous shape, size, and staining intensity, and the cells in each group tend to roll together across sections, as in the monkey, though not quite as rapidly or pronounced. The many fiber bundles give it both a disheveled and a lobulated appearance, but they must be ignored in order to see the patterns of blue, Nissl-stained neurons. The many fiber bundles essentially camouflage the VMpo, as does the background of tiny blue dots. Those dots are the glial cells that reacted to tissue hypoxia before the brain was immersed in the formaldehyde fixative ("reactive gliosis," as mentioned earlier); if the postmortem interval had been longer than twenty-four (p.165) hours, their number was greatly increased, which made visual analysis of Nissl-stained sections nearly impossible.

Again as in the macaque, the dense, patchy, calbindin-terminal-fiber labeling proved to be a reliable marker for VMpo cell groups, even though it did not stain all of them. On particular frontal or transverse sections through the posterior thalamus, at low-power the dense calbindin-terminal-fiber staining looks like a baseball catcher's mitt that has a saclike opening around the spinal fibers as they enter the thalamus. At high-power, numerous calbindin-positive fibers can be seen streaming into the densely labeled patches from the ascending bundles of spinothalamic fibers. The medial lemniscus fibers lie just ventral to the VMpo and course laterally to join the thick fiber layer that lines the bottom of the thalamus (the "external medullary lamina"). In these sections, the VMpo looks just like the region that Hassler described as the Ventrocaudalis portae.

In sagittal or horizontal sections, the VMpo is more easily differentiated because it is a coherent region of homogeneous neurons that is surrounded by a thin, fibrous lamina. In sagittal sections, most VMpo cells are cut crosswise and appear thin and lightly stained. It occupies exactly the location that the neuro-anatomist W. R. Mehler described as the "ventrocaudal portion of the Ventrocaudal nucleus," where he saw dense, thin-fiber spinothalamic degeneration in the sagittal sections from one human spinothalamic tractotomy patient, mentioned above. This equivalence supports the identification of the VMpo as the target of lamina I spinothalamic terminations in the human. [Mehler, 1966]

The sections that had been stained for the peptides SP (substance P) and CGRP (calcitonin gene-related peptide) in Anders's laboratory were most illuminating. Successful staining for these two markers had not been available in monkey thalamus, but in human sections, we saw several dense bursts of SP-stained terminal fibers in the VMpo; some of these were also stained for calbindin (in the adjacent section) and a few were not. Since lamina I spinothalamic neurons can stain for SP in the cat and rat, we interpreted these bursts as terminals of different modalities, and we inferred that the negative SP labeling we had obtained in the monkey thalamus had probably resulted from inadequate chemistry (i.e., antibody specificity or hidden epitopes). There were also dense bursts of terminal labeling in the sections stained for CGRP, but in contrast to SP, the CGRP-positive terminal bursts were never stained for calbindin. A few were located in the VMpo, always in gaps between calbindin-positive bursts, but most were just outside the borders we drew for the VMpo, within a narrow region between the VMpo and the somatosensory nuclei VPM + VPL. In addition, there was moderate CGRP terminal labeling in part of the VMb close to the VMpo, while most of the rostral VMb was not stained for any of these markers.

These results provided strong evidence indicating that the VMpo nuclear region in humans contains a complex mixture of several distinct types of cell groups that receive dense terminations having different staining characteristics, and they (p.166) therefore most likely have different functional specificity. As in the macaque, the human VMpo is well delimited by the calbindin-positive fiber terminations, which can definitely be associated with lamina I spinothalamic terminals. But it is clearly more complex than the macaque VMpo, and it likely contains many functionally distinct modalities of interoceptive activity. Based on the knowledge of rat PB that Anders had gained working with his students, we related the overall staining pattern to the interoceptive map in the PB, albeit flipped 180 degrees anteroposteriorly and modified for primates. In this scheme, the CGRP staining in the gap that surrounds the VMpo laterally, medially, and anteriorly signifies general visceral sensory inputs from the PB, which includes both cranial and sacral autonomic (parasympathetic, i.e., vagal and pudendal) activity and incorporates the caudal VMb. The CGRP-negative region in the rostral VMb corresponds with the gustatory region, which in primates receives direct input from the rostral third of the NTS but not from the PB. The VMpo fits in this rotated interoceptive PB map if it is topographic in the anteroposterior direction (face to foot), just as it is in the macaque. That must be true, because the evidence described below indicates that the human dorsal posterior insula is organized anteroposteriorly (face to foot), as it is in the macaque. [H. Herbert et al., 1990]

The human dorsal posterior insula.

As I described in chapter 2, our positron emission tomography (PET) study of graded innocuous cooling identified human discriminative thermosensory cortex, and it provided the first functional imaging evidence indicating that the VMpo cortical projection target in human cortex is the fundus of the posterior SLS, the dorsal posterior insular cortex. That was the only site in the contralateral cortex with activation linearly correlated with the innocuous cool stimulus temperatures. The only ascending pathway that conveys a linear representation of innocuous cool stimulus temperatures is the lamina I spinothalamic input to the anteromedial VMpo. The location of thermosensory cortex identified by our PET data is not in the somatosensory cortical areas; rather, it matches the anterograde tracing data we obtained in the macaque monkey remarkably well. Thus, our finding confirms the fundamental distinction of interoception and exteroception in the human brain. [Craig et al., 2000, 1994, 2001; Dostrovsky and Craig, 1996; Craig, 2004a, 2014]

Our subsequent fMRI study and a surface recording study by others corroborated our finding for cool stimulation. Two studies reported a comparable localization for activation by warm stimuli using electroencephalographic (EEG) analyses. In addition, we demonstrated an anteroposterior (face to foot) somatotopic gradient for innocuous cooling activation in the dorsal posterior insula with fMRI. Figure 12A (plate 8) shows our results, where white (green) indicates activation from the neck and dark gray (red) from the hand. The anteroposterior topography is clear in both the axial plane (left) and the sagittal plane (far right). [Maihöfner et al., 2002; Hua et al., 2005; Greenspan et al., 2008; Iannetti et al. 2003, Stancak et al. 2006] (p.167)

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Figure 12 The anteroposterior somatotopic organization of the human interoceptive cortex.

The magnified images (left) from the boxed regions show the anteroposterior organization in the horizontal plane for the activation during sensations of innocuous cool (A; neck vs. hand), muscle pain (B; arm vs. leg), skin heat pain (C; face vs. hand vs. foot), and affective touch (D; arm vs. leg; in individuals rather than ensembles). See plate 8 for unambiguous color comparisons.

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Plate 8 The anteroposterior somatotopic organization of the human interoceptive cortex.

The magnified images (left) from the boxed regions show the anteroposterior organization in the horizontal plane for the activation during sensations of innocuous cool (A; neck vs. hand), muscle pain (B; arm vs. leg), skin heat pain (C; face vs. hand vs. foot), and affective touch (D; arm vs. leg; in individuals rather than ensembles). See plate 8 for unambiguous color comparisons.

(p.168) Those results imply that nociceptive-specific activation also occurs in the fundus of the posterior SLS of the human cortex, as in the macaque monkey. The earliest functional imaging (PET and fMRI) and LEP studies of heat-evoked pain in humans reported activation in insular cortex, but they could not distinguish activation in the posterior insula from activation in the anterior insula or from activation in the somatosensory S2 field in the nearby parietal operculum, and the term operculo-insular cortex arose. Now, strong evidence is provided by studies that showed specific activation in the dorsal posterior insular cortex by graded noxious heat, C-fiber laser-evoked heat, noxious esophageal or colonic distension, intramuscular hypertonic saline injection (muscle pain), and pinprick. Three-dimensional reconstructions of EEG recordings showed that a source in the dorsal posterior insula is responsible for the earliest cortical LEP and also showed that it grades with stimulus strength and correlates with perceived pain intensity. The assertion that this region contains primary nociceptive cortex is supported as well by the attention-related enhancement of activity when subjects perform a nociceptive discrimination task. [Peyron et al., 2002; Frot and Mauguiere, 2003; Apkarian et al., 2005; Brooks et al., 2002; Keltner et al., 2006; Kakigi et al., 2003; Strigo et al., 2003; Kupers et al., 2004; Baumgärtner et al., 2010; Iannetti et al., 2005; Schlereth et al., 2003; see Craig, 2014]

In addition, several studies reported that specific nociceptive activation in the dorsal posterior insular cortex of humans is somatotopically organized along an anteroposterior gradient. This has been demonstrated for both intracutaneous and intramuscular hypertonic saline injection (figure 12B), contact noxious heat (figure 12C), and pinprick. One study also reported that the LEP from the face was located anterior to the LEP from the hand in one patient. Such evidence clearly differentiates this specific nociceptive representation from the neighboring S2 somatosensory cortical areas, which are organized along an orthogonal lateral-to-medial (face to foot) somatotopic gradient in the parietal operculum. These studies clearly support the conclusion that the specific representation of nociceptive activity in human dorsal posterior insular cortex corresponds directly to the specific representation in the fundus of the SLS of the macaque monkey that we demonstrated with anatomical tracing and physiological recording evidence. [Henderson et al., 2007, 2011; Brooks et al., 2005; Baumgärtner et al., 2010; Vogel et al., 2003]

Clinical support for this conclusion is provided by several reports that a lesion of this same region of cortex caused localized contralateral loss or dysfunction of pain and temperature sensibilities. In contrast, large lesions of S1 and S2 do not reduce pain sensation, unless they involve area 3a and/or the dorsal posterior insula. Clinical lesion effects can vary, however, due to various factors that cannot be controlled in inhomogeneous samples (e.g., compensatory plasticity, undetected damage, ipsilateral pathways), which might explain two contradictory reports. [Schmahmann and Liefer, 1992; Greenspan et al., 1999; Birklein et al., (p.169) 2005; Perl, 1984; Ploner et al., 1999; Craig and Blomqvist, 2002; Craig, 2003a; Olausson et al., 2002; Voets et al., 2006; Starr et al., 2009; Veldhuijzen et al., 2010]

Human patients report pain sensations when the VMpo is electrically stimulated, which suggests that pain might also be elicited by stimulation of the dorsal posterior insular cortex. In his pioneering studies, the neurosurgeon Wilder Pen-field never found a region of cortex from which pain could reliably be elicited, even though he stimulated the cortical surface of hundreds of patients. However, he could not stimulate the dorsal posterior insula because access was prevented by the many large blood vessels in the deep folds of cortex that surround the insula. But evidence has been provided by a team of clinicians who implant one or more electrodes deep in the brain for electrical stimulation in human epilepsy patients in order to differentiate epileptogenic and functional cortex. The recently published summary of their experience—4160 cortical stimulation sites in 164 patients across twelve years—states definitively that "the medial parietal operculum and the posterior insula are the only areas where electrical stimulation is able to trigger activation of the pain cortical network and thus the experience of somatic pain" in awake humans. [Mazzola et al., 2012]

They never elicited reports of localized contralateral pain sensations upon stimulation of the primary somatosensory cortex, a result consistent with Pen-field's reports. Yet they were also unable to evoke pain with stimulation of the cingulate cortex, the limbic motor area on the midline that is strongly associated with pain unpleasantness, on the basis of functional imaging and lesion results. Since the primary somatosensory cortex is also activated in many imaging studies of pain, the authors concluded that activation of a distributed network must generate the experience of pain, rather than a specific sensory cortical representation, and that their stimulation method was not an efficient means of activating that network. Thus, they emphasized that pain was evoked in only 10% of stimulation trials in the insula. That conclusion is consistent with the consensus view that the feeling of pain is supported by activation of a network of areas (referred to as a "pain matrix," a term that to me is nonsense).

However, figure 3 of their report shows a remarkable data set that is considerably more specific. That figure presents a graph that shows the location of every stimulation site in a group of 25 patients in whom more than one electrode was implanted in insular cortex; the graph shows the distance from the dorsal posterior pole of the insula for each stimulation site, and each location is colored to show whether stimulation at that site produced a pain sensation or not. The graph indicates very clearly that stimulation at sites within 2 centimeters of the dorsal posterior pole of the insula produced pain on 100% of trials (15 out of 15), whereas stimulation at sites between 2 and 3 centimeters away from the dorsal posterior pole caused pain on 53% of trials (10 out of 19), and stimulation at sites between 3 and 4 centimeters away caused pain on 22% of trials (6 out of 27); but (p.170) if an electrode was further than 4 centimeters distant, then stimulation in the insular cortex never caused pain (0 of 15). To my mind, those findings strongly support the conclusion that the primary nociceptive sensory cortex is located in the dorsal posterior insular cortex of humans, which is consistent with all of the evidence described above. In a separate article, they even reported an anterior-to-posterior (face to foot) topographic organization of stimulation-produced pain in insular cortex. [Mazzola et al., 2009, 2012]

The human interoceptive cortex.

The dorsal posterior insula in humans is certainly activated by "pain and temperature" stimuli; the findings described in this section show that it is also activated by a wide variety of stimuli associated with lamina I neuron activity. Although the dorsal posterior insula represents primarily "pain and temperature" in the macaque monkey, the evidence indicates that in human it represents most, and quite likely all, aspects of the physiological condition of the body that are signaled by small-diameter sensory inputs. The following findings are consistent with the idea that this pathway subserves interoception and that the dorsal posterior insula in human contains primary interoceptive cortex.

Graded activation in the dorsal posterior insula was demonstrated by an fMRI study of itch; they used intracutaneous histamine, the same chemical we used to demonstrate graded responses in itch-specific lamina I spinothalamic neurons, because it produces a relatively pure itch sensation in skin. Graded activation in the dorsal posterior insula has also been demonstrated using nonnoxious esophageal or rectal distension, including subliminal rectal distension stimuli that do not reach awareness. Both of these stimuli can involve vagal sensory activity via the VMb to the interoceptive cortex (see below) as well as lamina I activity via the VMpo. Graded activation of the dorsal posterior insular and mid-insular cortex was demonstrated in several studies using isometric exercise (handgrip), dynamic exercise (bicycling), central motor command, or free movement; and one study showed evidence that activation was correlated with intensity of exercise and was directly related to muscle blood flow, with heart rate and blood pressure effects removed. These nonnoxious modalities activate lamina I neurons responsive to small-diameter sensory fibers from muscle, which variously report waste metabolites, contraction and vascular distension, as described in chapter 3; they have fundamental significance for energy utilization and for the concept of homeostatic sentience that is proposed in chapter 6. Several other demonstrations of cardiovascular responses that probably involve VMb input are described next. [Drzezga et al., 2001; Strigo et al., 2003; Larsson et al., 2012; Williamson et al., 1997, 1999; King et al., 1999; Cechetto and Shoemaker, 2009]

If the dorsal posterior insula in human does contain a coherent interoceptive cortex that represents all homeostatic input, then its anterior half must represent (p.171) the cranial autonomic sensory inputs and lie just in front of the posterior half, as in the macaque monkey. Determining whether human insular cortex contains a coherent interoceptive representation was complicated by early reports that human taste (or, gustatory) cortex lies in the anterior insula. That is far removed from the region where we found activation using thermosensory stimuli in the human posterior insula. Those early reports originated mainly from a group that studied taste in the macaque monkey, and they had simply presumed that the taste cortex in humans should be located in the anterior insula, because that's where it is located in the monkey. However, the taste cortex in the monkey lies in the anterior insula because interoceptive cortex occupies the entire length of the dorsal insula. The human anterior insula is involved in the discrimination of different taste stimuli, as they had observed, but it is involved in all discrimination tasks, and so much more, as a growing number of studies reported, similar to our PET study of thermosensation. Fortunately, within a few years, other investigators studying taste recognized that the anterior insula is involved in subjective judgments of any sensation, and they showed with critical experiments that the primary taste cortex lies in the dorsal mid-/posterior insula. It lies anterior to the region activated by spinal small-diameter inputs such as pain and temperature, which directly supports the presence of a coherent, unified interoceptive cortex. [Small, 2010; Veldhuizen et al., 2007, 2011]

The dorsal mid-/posterior insula is activated bilaterally (stronger on the left side with attention, see chapter 6) also by a range of modalities associated with cardiovascular or visceral sensations that drive vagal sensory fibers, such as non-noxious stomach distension; the Valsalva maneuver (blocked expiratory pressure); manipulations of blood pressure and heart rate by maximal inspiration, by post-exercise ischemia, active cycling, free cycling, or intravenous injection of lactate (tachycardia and panic) or cholecystokinin (tachycardia and anxiety); verbally triggered bronchiole constriction in asthma patients; intravenous hypertonic saline injection (which generates thirst); the sight of food in fasted obese patients (hunger); and "air hunger" (inspiratory restriction with constant carbon dioxide). Also, a recent study reported that paying attention to a breathing task produces selective enhancement of activation in the dorsal posterior insula (on the right side), and it revealed significant functional connectivity with a region they labeled as "posterior ventromedial thalamus," which could have meant either the VMb or VMpo or both. The topographic order has not yet been resolved, but the primary gustatory cortex should lie anterior to these general visceral activation sites, as it does in the VMb. [G. Wang et al., 2008; Henderson et al., 2003; King et al., 1999; Williamson, et al., 1999; Benkelfat et al., 1995; Reiman et al., 1997; Denton et al., 1999; Tataranni, et al., 1999; Delparigi et al., 2005; Banzett et al., 2000; Farb et al., 2013a]

Thus, the evidence supports the assertion that a complete interoceptive cortex in humans is present in the posterior one-third of the dorsal insula, which includes (p.172) both a posterior half that receives input from the VMpo and an anterior half that receives input from the VMb. In humans, this region serves as primary sensory cortex for each of the distinct affective bodily feelings that we can perceive as discriminative sensations, such as cool, warm, itch, first pain (pricking), second pain (burning), sharpness, and taste. It also underlies the less well discriminated interoceptive feelings from the body, such as muscle burn, joint ache, and deep cramp, as well as the general visceral feelings of fullness, vasomotor flush, generalized illness, nausea, hunger, "air hunger," and thirst. It is also activated by subliminal visceral stimuli that we do not perceive, and by vascular distension, which can be reported as "pressure," but which may normally be involved with the ongoing feeling of being alive (see chapter 6).

The presence of a distinct somatotopic map for each discriminative interoceptive modality in both the VMpo and the dorsal posterior insula of the macaque monkey strongly suggests that distinct somatotopic maps are present for each of the distinct discriminative feelings from the body that are represented in the interoceptive cortex of humans, which is consistent with the fact that we can localize such feelings in the body. Evidence for somatotopy is so far only available for cool, heat pain, pinprick pain, and muscle pain, as noted above, and so far there is no evidence regarding the relative distinctness and location of these maps in humans. It remains to be determined whether separate areas exist for the interoceptive activity that underlies the less well discriminated feelings, the general visceral feelings, and the subliminal responses not associated with subjective feelings; these may instead emerge from a general interoceptive area that is not specialized. The possibility that general visceral feelings are qualitatively different ("blunt") because they emerge from a poorly differentiated, polysynaptic representation of homeostatic input to the VMb from the PB (as suggested above) deserves serious consideration.

Recent anatomical studies of the cellular architecture of the posterior operculo-insular region of the human cortex differentiated two areas in the parietal operculum that were identified with fMRI activation as the twin S2 areas (OP1 and OP4) and two more areas medial to S2 that include the fundus of the SLS at the posterior pole of insular cortex (OP2 and OP3; in the green region of figure 10D but not identified). The more posterior area (OP2) extends well into the lateral sulcus behind the insula, and it was identified as primary vestibular cortex by activation in fMRI experiments. In macaques, that area is called the retroinsular area (Ri; see box 6), because it lies in the fundus of the lateral sulcus just posterior to the insula and Idfp, the posterior half of the interoceptive cortex. Thus, the human area OP3 in the fundus of the posterior SLS just anterior to OP2 could be the human equivalent of Idfp. If that is the case, then there should be another area in the fundus of the SLS anterior to OP3 that will correspond with the Idfa and contain primary gustatory cortex. An alternate possibility is that the OP3 in human is homologous with the combined Idfp and Idfa in monkey, that is, the (p.173) entire interoceptive cortex. [Eickhoff et al., 2006a, 2006b; Kurth et al., 2010b; Chen et al., 2010]

Interoceptive touch.

A particularly interesting sensory modality is affective touch, also called sensual touch. As a cutaneous tactile modality, one might normally expect that it is served by large-diameter, fast-conducting myelinated fibers that send a collateral up the dorsal columns and the medial lemniscus to the somato-sensory thalamus and from there to the somatosensory cortices. That's the path of all cutaneous low-threshold mechanoreceptors, which subserve exteroception and fine adjustments in skeletal motor behavior. But none of that is true for affective touch. Affective touch is an interoceptive modality, and it subserves homeostasis not only at the level of the individual but also at the level of the social community; it supports the health and well-being of the individual and the species.

Affective touch is conveyed by small-diameter, unmyelinated C-tactile ("slow brush") sensory fibers with peripheral endings in the epidermis and central terminals in the superficial dorsal horn, as mentioned in chapter 3. Microneuro-graphic recordings in human participants (as described in box 2) indicate that they conduct spikes very slowly (~1 m/ s), and they fire at very low rates (which saves energy). They innervate hairy skin all over the body; they are not found in glabrous skin (e.g., the palms), where high acuity mechanoreception is needed. They are reported to be at least as numerous as C-nociceptors or myelinated mechanoreceptors. The C-tactile sensory fibers are activated only by light brushing within a limited range of slow velocities that are not fast enough to activate large-diameter cutaneous mechanoreceptors. More recent evidence indicates that C-tactile sensory fibers in mice are genetically characterized by a particular receptor molecule that differentiates them from both C-nociceptors and hair follicle C-mechanoreceptors; that article reported also that they are activated by "massage-like stroking of hairy skin," are anxiolytic, and generate a positive affective behavioral motivation in mice. [Vallbo et al., 1995; Olausson et al., 2010; Vrontou et al., 2013]

Only one lamina I spinothalamic neuron specifically responsive to C-tactile sensory fibers has been identified; but responsive cells in the superficial dorsal horn have been documented, and (nonselective) lamina I cells that respond to C-tactile fibers and project to PB have been recorded in rats. Affective touch is lost along with pain and temperature and itch after the lateral spinothalamic pathway is cut in human patients. Selective stimulation of C-tactile sensory fibers by slow brushing activates the contralateral dorsal posterior insular cortex, and that activation displays the characteristic velocity dependence of the C-tactile sensory fibers. The activation in the dorsal posterior insula is somatotopically organized, with arm activation anterior to leg activation, that is, with the same (p.174) somatotopic gradient observed in the interoceptive cortex for the discriminative interoceptive modalities of cool, heat pain and pinprick pain, and the deep modality of muscle pain. Figure 12D shows the anteroposterior somatotopy for affective touch in six normal healthy participants at the right, compared with one of the patients described in the next paragraph on the left. [Craig and Kniffki, 1985; Andrew, 2010; White and Sweet, 1969; Olausson et al., 2002; Löken et al., 2009; Björnsdotter et al., 2009, 2010]

In several studies, investigators examined two particular polyneuropathy patients who lack large-diameter cutaneous fibers but still have peripheral C-fibers. Both patients report no feelings of touch, yet they can identify significantly often which limb was stimulated by slow brushing, and they vaguely perceive a pleasant feeling in response. (Conversely, neuropathy patients who have reduced C-fiber skin innervation report significantly lower pleasantness ratings to slow brushing.) "Slow-brush" stimulation produces reflex sympathetic activity in their cutaneous nerves (with unknown function). The C-tactile pathway in these two patients was found to activate the contralateral dorsal posterior insular cortex, with no response or deactivation in the S1 or S2 cortices. The anteroposterior somatotopy in patient GL is shown in figure 12D (left). Activation in the mid-insula (see figure 14B) and in the left anterior insular cortex was also observed in these patients. [Olausson, 2010; Björnsdotter et al., 2010]

Affective touch results from external contact, but only of the kind produced by gentle conspecific contact, like during a caress or a hug or maternal cuddling or when monkeys groom each other. In contrast to the rapid, high-acuity discriminative touch substrates, affective touch is conveyed by slow, low-resolution peripheral and central interoceptive components. Instead of guiding skeletal motor adjustments, affective touch communicates social emotion and modulates the homeostatic well-being of the individual and the group. Affective touch epitomizes the fundamental distinction between interoceptive and exteroceptive feelings.

To my mind, the C-tactile receptors are safety detectors that activate the "calm and connection" system, the energy nourishment system, in opposition to the danger signals that activate the arousal and stress system, the system that expends energy, as discussed in chapter 8. Conceptually, there is a remarkable parallel with the social aggregation and feeding behavior of the worm C. elegans, which is facilitated by conspecific touch, pheromones, and oxygen but opposed by noxious heat or cold, "harsh touch," and abnormal levels of carbon dioxide; their behavior is also modulated in a state-dependent fashion by octopamine (the invertebrate equivalent of adrenalin), serotonin, and neuropeptides, again similar to mammals. Even more amazing, just as Harry Harlow demonstrated for monkeys, a lack of gentle mechanical touch stimulation during larval development in C. elegans produces dysfunctional social behavior, retards reproduction, and alters neuronal connectivity. [Uvnas-Moberg et al., 2005; Rose et al., 2005; (p.175) Barrios et al., 2008; Bretscher et al., 2008; Chatzigeorgiou et al., 2010; Flavell et al., 2013]

Summary, and an interoceptive perspective on cortical gyrification.

The ascending axons of lamina I neurons course in the lateral spinothalamic pathway, precisely where spinal lesions selectively interrupt feelings of pain, temperature, itch, and affective touch (or sensual touch) from the contralateral body. The thalamic projections of lamina I neurons in primates terminate densely and somatotopically in a specific thalamocortical relay nucleus in the posterolateral thalamus, the VMpo, which projects to the dorsal posterior insular cortex. The VMpo is organized somatotopically in the anterior-to-posterior direction (face to foot), which is orthogonal to the mediolateral somatotopic gradient (face to foot) in the main somatosensory nuclei, VPM + VPL (or Ventrocaudalis in human thalamus, V.c), which relay mechanoreceptive and proprioceptive activity to the main somato-sensory cortices, S1 and S2. This dual thalamocortical representation is the completion of the fundamental morphological (i.e., genetic) distinction between interoceptive and exteroceptive processing described in chapter 3, which includes the differentiation of A-cells and B-cells in the dorsal root ganglia, of large-diameter and small-diameter sensory fibers in the medial and lateral dorsal root entry zones, of the superficial dorsal horn and the deep dorsal horn, and of the two ascending sensory pathways—the dorsal column–medial lemniscus and the spinothalamic pathway.

In the macaque monkey, the VMpo contains three distinct modality-specific components, COOL, NS, and HPC neurons, which correspond to discriminative sensations of innocuous temperature, first pricking pain, and second burning pain; it also contains at least one more small component of unidentified modality that represents approximately 10% of lamina I spinothalamic neurons. The COOL, NS, and HPC neurons all display responses like lamina I neuron responses, and they are organized somatotopically from anterior to posterior (face to foot), consistent with the interpretation that these are labeled lines. The COOL neurons are located most medially and concentrated in the anterior one-third of the VMpo; the NS and HPC neurons are located in two successively more lateral clusters in the trigeminal portion of the VMpo but appear to be intertwined in the more posterior two-thirds of the VMpo.

The size of the VMpo varies considerably in macaque monkeys: it is smaller in New World monkeys, and much larger and more complex in hominids (apes and humans). In strepsirhine primates, a very small cluster of round neurons may be homologous, but no tracing evidence is available.

The VMb that adjoins the VMpo anteriorly in anthropoid primates receives direct high-resolution taste ("gustatory") sensory input from the NTS and general homeostatic ("visceral") sensory input from the PB (and perhaps somewhat from (p.176) NTS). Thus, the combined structure VMpo + VMb represents homeostatic sensory activity from both spinal autonomic sources (in the VMpo) and cranial autonomic sources (in the VMb and PB). (The sacral lamina I projections have not been mapped.) Both the lamina I spinothalamic pathway to the VMpo and the NTS taste pathway to the VMb are clearly present only in anthropoid primates.

The entire homeostatic sensory representation is relayed with the same posterior-to-anterior topography to the continuous strip of granular cortex in the fundus of the SLS at the dorsal margin of insular cortex, which we refer to as interoceptive cortex. Projections from the VMpo terminate in the posterior half, or Idfp, and projections from VMb terminate in the anterior half, or Idfa. Evidence currently being analyzed suggests that COOL, NS, and HPC VMpo neurons terminate somatotopically in individual portions of the Idfp, the NS on the medial wall of the SLS, COOL on the medial wall most anteriorly, and HPC on the lateral wall.

This pathway potentially conveys all interoceptive inputs representing the physiological condition of all tissues of the body. In monkeys, however, the VMpo appears to have evolved to convey primarily the high-acuity sensory channels that support ecologically significant discriminative behaviors. This conclusion is exemplified by the specialized perinasal COOL pathway in owl monkeys that likely aids their nocturnal foraging. Thus, the interoceptive cortex in monkeys may receive only a limited portion of the available interoceptive inputs. Nevertheless, the "anomalous" characteristics of the COOL, NS, and HPC cells (described in chapter 3) suggest an essential role as feedback to the homeostatic control systems, which are integrated in the insular cortex for forebrain control of homeostatic emotional behavior. The cytoarchitectonic organization of macaque insular cortex in longitudinal strips parallel to the interoceptive cortex suggests, together with connectional and physiological evidence, that the interoceptive cortex provides the longitudinal index for insular integration of sensory activity related to the emotional significance of hand and mouth movements, interspecies emotional communication, and feeding behavior. It does not support the broad cognitive integration that characterizes the human anterior insula, as discussed in the following chapters. Homeostatic integration in the insular cortex can refine and optimize energy utilization for all of these functions; the somato-topic inputs from tissue-specific subtypes of thermoreceptive cells and of cells responsive to interoceptive deep input, including metaboreceptors, ergorecep-tors, and vasoreceptors, are essential for such integration. Thus, the monkey interoceptive thalamocortical pathway may be specialized to represent high-acuity, discriminative COOL, NS, and HPC sensory channels, but it is more than a simple "pain and temperature" pathway.

By contrast, the VMpo in the thalamus of hominid primates is greatly enlarged, clearly distinct, lobulated, and anatomically and chemically more complex. The VMpo + VMb and the interoceptive cortex in humans are both quantitatively (p.177) (proportionately) larger and qualitatively more developed than they are in the macaque monkey. The functional imaging evidence shows that the dorsal posterior insular cortex in humans is activated by many interoceptive modalities in addition to pain, temperature, and taste, and that it is organized along the same anteroposterior gradient (face to foot) as in the macaque monkey. Clinical stimulation and lesion findings in humans support the presence of primary cortical representations of contralateral discriminative pain, temperature, and taste sensibilities in the dorsal posterior insula, and much more.

In humans, this interoceptive pathway represents all aspects of the physiological condition of the body. The labeled lines for distinct, well-discriminated feelings, such as cool, sharp pain, itch, and salty, are almost certainly represented in separate topographic maps that are available to attention-related cognitive processes, yet they retain their homeostatic character because they are linked with autonomic reflexes, homeostatic emotions, and energy utilization. Interoceptive cortex also contains representations of less well discriminated and less well localized but affectively distinct feelings from the body, such as itch, isometric and dynamic exercise, muscle ache, deep cramp, nonnoxious esophageal, gastric, or rectal distension, vascular distension, and affective touch; these interoceptive modalities might occupy distinct maps, or they might involve subcategories of cells that can be roughly located because they are contained within a general map. They are closely linked with homeostatic control, and they are characterized by strong affective feelings that help guide emotional behavior, such as vigor, fatigue, fullness, and pleasant social contact. This pathway also supports representations of general visceral feelings, such as vasomotor flush, thirst, hunger, and "air hunger," which are more vague (less "bright") and might involve more ancient homeostatic integration because they are still based largely on ascending activity from the brainstem PB via the VMb. And last, the interoceptive cortex supports subliminal interoceptive activity that does not generate a feeling that enters human awareness yet modulates homeostasis and emotional behavior (as discussed in the following paragraphs).

These findings support the assertion that the interoceptive cortical image of the physiological condition of the body in humans emerged evolutionarily as an extension of the hierarchical homeostatic system. In other words, feelings from the body in humans reflect its homeostatic condition. These findings also support the idea that interoceptive cortex in humans represents homeostatic sensory feedback, that is, it provides a high-resolution sensory image of the functional actions of the autonomic nervous system, which is the foundation for the concept of homeostatic sentience that is discussed in chapter 6. The interoceptive cortex may well represent all homeostatic sensory signals, whether they produce subjective feelings or not; such signals include specific osmolarity changes, plasma concentrations of salt, steroids, or immune factors, and any condition that changes the (p.178) activity of small-diameter sensory fibers. Importantly, these fibers report the physiological conditions in the body on an ongoing basis, not just during an emergency. They convey slow activity at slow conduction velocities, which is consistent with the primal need to conserve energy, because they are continuously active. The physical proximity of the main interoceptive cortical targets produced by novel embryological gyrification in the anthropoid primate brain, shown in figure 13 and described below, provides strong evidence for this conceptual perspective.

Notably, the potential representation in the interoceptive cortex of physiological conditions that may not produce subjective feelings implies the possibility of subconscious interoceptive signals that affect emotional behavior in human beings. Such signals could have significant clinical impact, particularly if behaviorally modifiable. For instance, significant differences were observed in the activation of the dorsal mid- and posterior insula due to inspiratory load in exercising elite athletes or, surprisingly, due to visual threat in highly trained military warriors, in comparison to healthy control subjects; these were interpreted as interoceptive signals that produced a "body prediction error" thought to be crucial for the attainment of optimal behavioral performance. That interpretation is consistent with other evidence indicating that interoceptive awareness and insular cortical processing mediate self-regulation of physical exertion and energy expenditure without conscious awareness (as mentioned in chapter 1). Nevertheless, the physiological origin of the interoceptive signals that generated the observed activation in dorsal mid- and posterior insula of those athletes and warriors is unknown, and identification of the source of those signals could have immediate practical value. If the source is an interoceptive signal from the body, then it ascends via the VMpo + VMb. Yet the incorporation of all interoceptive activity within the dorsal posterior insular cortex remains to be demonstrated, as does the organization of distinct interoceptive feelings in separate maps, including, for example, the putative distinct representations of NS (first, or pricking, pain) and HPC (second, or burning, pain) activity. [Paulus et al., 2010, 2012; Herbert et al., 2007;Hilty et al., 2011a]

In primates, the high-resolution homeostatic sensory activity representing the physiological condition of the body ascends directly from lamina I to the VMpo, to the VPI (see box 6), and to a site in medial thalamus, the MDvc (see box 5). These regions, in turn, activate the three sites in the cortex that are highlighted in figure 13A—the fundus of the SLS (by way of the VMpo), the fundus of the central sulcus (also by way of the VMpo), and the fundus of the cingulate sulcus (by way of the MDvc). The image in figure 13A is from an fMRI experiment; it shows functional activation at these three cortical sites induced by noxious cold stimulation of the contralateral foot of an anesthetized monkey (applied for 45 sec every 2.5 min). (p.179)

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Figure 13 The cortical targets of the ascending homeostatic sensory pathway.

The fMRI activation produced in a monkey by repeated 45-second applications of ice to one foot is displayed on the structural MR image (A; 1 × 1 mm pixels) and on a standard coronal atlas drawing (B). This activation pattern highlights the fact that the lamina I spino-thalamo-cortical pathway terminates in the fundus of each of these three major sulci (CgS, cingulate sulcus; CS, central sulcus; LS, lateral sulcus). These sulci are present only in the brains of anthropoid primates, which are the only animals that have this phylogenetically novel pathway. This image suggests that these termination sites are the anchors that produce these cortical sulci, and that the cortex is folded so that these three homeostatic sensorimotor processing sites are as close together as possible. These sites are continuously active and interactive, with the goal of optimizing energy utilization in the body and the brain; thus, these folds advance the goal of this network by producing a continuous savings in the energy budget of the brain.

The Interoceptive Pathway to the Insular CortexLamina I Spinothalamic Input to the Thalamus and Cortex in Primates

Plate 9 The cortical targets of the ascending homeostatic sensory pathway.

The fMRI activation produced in a monkey by repeated 45-second applications of ice to one foot is displayed on the structural MR image (A; 1 × 1 mm pixels) and on a standard coronal atlas drawing (B). This activation pattern highlights the fact that the lamina I spino-thalamo-cortical pathway terminates in the fundus of each of these three major sulci (CgS, cingulate sulcus; CS, central sulcus; LS, lateral sulcus). These sulci are present only in the brains of anthropoid primates, which are the only animals that have this phylogenetically novel pathway. This image suggests that these termination sites are the anchors that produce these cortical sulci, and that the cortex is folded so that these three homeostatic sensorimotor processing sites are as close together as possible. These sites are continuously active and interactive, with the goal of optimizing energy utilization in the body and the brain; thus, these folds advance the goal of this network by producing a continuous savings in the energy budget of the brain.

(p.180) The overall pattern of this homeostatic integration mimics the sensorimotor organization of the spinal cord and the entire homeostatic hierarchy. Thus, there is interoceptive input to the separate sensory and motor components of the central homeostatic system at the thalamocortical level in primates. The lamina I spinothalamic projections to the VMpo and MDvc are both phylogenetically unique to primates. The insular and cingulate cortices that receive these inputs can be regarded as limbic sensory and limbic motor cortices, respectively, because their major descending projections are to the PB and PAG, respectively. As suggested in chapter 2, the dual activation of limbic sensory and motor cortices generated by this pathway constitutes the feeling and the motivation of homeostatic emotions that clearly subserve the maintenance and well-being of the body. The fundus of the central sulcus (area 3 a) can be regarded as a viscerosomatic motor site, as discussed earlier. It projects strongly to spinal motor-control regions, as does the fundus of the cingulate sulcus, which contains the cingulate motor areas that drive emotional behavior.

The ascending lamina I homeostatic sensory pathway projections to the three activation sites shown in figure 13 are present only in anthropoid primates (Old World monkeys, apes, and humans). Furthermore, the three sulci that contain those activation sites—namely, the lateral sulcus, the central sulcus, and the cingulate sulcus—are also present only in anthropoid primates. I find that striking. The fact that homeostatic sensory activity ascends to the fundus of each of these three sulci suggests strongly, on the basis of the tension model of cortical morphogenesis, that each of these projections serves as a tensile anchor for cortical gyrification. That is, each sulcus is formed during the embryological development of the brain by being anchored to a center point while the cortex expands. That means that the interoceptive projections to the fundus of all three of these sulci are morphological anchors of the human brain. [Craig, 2011b]

To my mind, this image says that embryological gyrification in anthropoid primate brains brings these three interoceptive processing regions as close together as possible. This could only occur for one reason—because it optimizes the energy efficiency of homeostatic processing. These three sites are the three main targets of the new pathway, a primate innovation that brought high-resolution homeostatic sensory information to the cortex; these three sites are slowly but continuously active, and they are continuously communicating with each other about maintaining the well-being of the body and optimizing homeostatic senso-rimotor integration and homeostatic emotional behaviors. In order to optimize homeostasis, it is crucial to achieve the most efficient utilization of energy. To my mind, the image in figure 13 says that a significant energy savings was achieved by folding the cortex so that these three sites are as close together as possible. The morphological folds of the anthropoid primate cortex provide a visible record—fossilized evidence—of the evolutionary significance of optimal homeostatic energy utilization.

(p.181) The lateral sulcus is the first fold to appear in the developing fetal human brain, even though it's one of the last to have appeared during evolution. It doesn't appear first because it is repeating an evolutionary progression. The lateral sulcus, the Sylvian fissure that forms the operculated insula, is the first fold that appears in the developing human brain because that saves the most energy!


(1) Regarding the name VMb: The neuroanatomist who gave it the name VMb in an article published in 1942, J. Rose, worked on the somatosensory map in the VPM+VPL of cat and rabbit with V. B. Mountcastle and C. Woolsey, and he predicted (correctly) that the VMb would be associated with visceral sensation, based on its neuroanatomical characteristics. Later investigators named it the parvicellular part of the ventral posterior medial nucleus, or VPMpc, based on their neuroanatomical opinions. In my view, that name incorrectly associates it with the exteroceptive somatosensory nuclei VPM + VPL, which project mainly to area 3b, or S1. In contrast, the main cortical projection of VMb is to the insular cortex, in parallel with the VMpo, and like the VMpo, it also has an ancillary projection to area 3a in the fundus of the precentral sulcus.

(2) Tracing studies have not yet tested whether a direct projection to thalamus from lamina I and NTS is present in strepsirhine primates.

(3) We named these areas according to the naming convention that had been previously used by a prominent group studying the orbitofrontal cortex, because the anterior areas we identified match almost exactly with the areas they had previously recognized adjacent to the orbitofrontal areas.

(4) It had been previously reported that anesthesia prevents LEP recordings in monkey and human; we succeeded by using a less well-known anesthetic (Saffan, or alfaxalone/alphadolone) that I had learned about from physiologists who study viscerosensory and autonomic activity, because it leaves many central responses to small-diameter sensory inputs intact.

(5) The small posterior source was also anteroposteriorly topographic; it was localized to the dorsal wall of the lateral sulcus, where VPI projections are located in the anterograde tracing data referred to above (box 6).

visible in the photograph). The graded LEPs (B) recorded on all eight electrodes showed consistent topographies (that is, relative size distributions) after brief laser stimuli (5 msec) applied to the contralateral hand and foot at the indicated intensities (dot marks stimulus; W means watts). The image in (C) shows the reconstructed LEP source locations in the contralateral dorsal posterior insula (blue in plate 7) on a cutaway MRI three-dimensional reconstruction of one monkey's head (triangle, ear; circle, hand; box, foot).