Pain matrices and neuropathic pain matrices: A review

TL;DR: The pain matrix is conceptualised here as a fluid system composed of several interacting networks, including posterior parietal, prefrontal and anterior insular areas, which ensures the bodily specificity of pain and is the only one whose destruction entails selective pain deficits.

Abstract: The pain matrix is conceptualised here as a fluid system composed of several interacting networks. A nociceptive matrix receiving spinothalamic projections (mainly posterior operculoinsular areas) ensures the bodily specificity of pain and is the only one whose destruction entails selective pain deficits. Transition from cortical nociception to conscious pain relies on a second-order network, including posterior parietal, prefrontal and anterior insular areas. Second-order regions are not nociceptive-specific; focal stimulation does not evoke pain, and focal destruction does not produce analgesia, but their joint activation is necessary for conscious perception, attentional modulation and control of vegetative reactions. The ensuing pain experience can still be modified as a function of beliefs, emotions and expectations through activity of third-order areas, including the orbitofrontal and perigenual/limbic networks. The pain we remember results from continuous interaction of these subsystems, and substantial changes in the pain experience can be achieved by acting on each of them. Neuropathic pain (NP) is associated with changes in each of these levels of integration. The most robust abnormality in NP is a functional depression of thalamic activity, reversible with therapeutic manoeuvres and associated with rhythmic neural bursting. Neuropathic allodynia has been associated with enhancement of ipsilateral over contralateral insular activation and lack of reactivity in orbitofrontal/perigenual areas. Although lack of response of perigenual cortices may be an epiphenomenon of chronic pain, the enhancement of ipsilateral activity may reflect disinhibition of ipsilateral spinothalamic pathways due to depression of their contralateral counterpart. This in turn may bias perceptual networks and contribute to the subjective painful experience.

Summary

From nociception to pain: a second-order perceptual matrix

• The 'classical' PM encompasses activity in many areas distinct from the nociceptive network described above, the most consistent being the mid-and anterior insulae, the anterior cingulate, prefrontal and posterior parietal areas, and with less consistency the striatum, supplementary motor area (SMA), hippocampus, cerebellum, and temporo-parietal junction.
• The mid-and anterior insulae participate almost constantly to the PM.
• Their activation may reflect a posterior-to-anterior information flux within the insula (.

From immediate perception to pain memories: third-order networks

• Impressive changes of the pain experience can occur without changes in the matrices described above.
• The enhancement of subjective pain during the observation of other Of importance, areas involved in the 'reappraisal' matrix such as the perigenual cingulate and orbitofrontal cortices are themselves strongly interconnected with subcortical regions crucial for descending pain control (notably the periaqueductal grey matter).
• Tonic activation of these areas may therefore not only support modifications in the subjective value of the nociceptive stimulus, but also contribute to a loop which changes the activity of ascending nociceptive systems, and by this bias influence the ascending input to cortical targets (see e.g. Leknes et al 2013) .
• This point will be of importance when discussing PM changes during neuropathic pain below.

The pain experience: an intersection of matrices

• The great merit of the PM concept was to underscore that pain experiences result from coordinated activity in a number of brain regions -i.e. the absence of any single "pain centre".
• While some investigators consider the PM as a genuine biomarker of the pain experience -a direct measure of the actual pain-others claim that the PM simply reflects a non-specific system of salience detection.
• The viewpoint proposed here, which elaborates and expands existing notions (e.g. Loeser.

2000, Tracey & Mantyh 2007

• ) is that the final experience of pain (i.e. the pain the authors shall remember) results from the convolution of three orders of brain processing with progressive complexity, in networks that they may tentatively label as 'nociceptive', 'perceptive-attentional' and 'reappraisalemotional' matrices .
• Regions receiving spinothalamic input ensure the somatic-specific ('corporal') quality of the sensation; they trigger activity in parietal, frontal and anterior insular circuits supporting conscious perception, vegetative reactions and their modulation by attention and vigilance.
• The immediate perception issued from these activities can itself be modulated by higherorder networks driven by emotional contexts and internal states.
• This reappraisal determines a private generated assessment of instant percepts, tuning them up or down as a function of affective states and previous memories, and building what will represent the "subjective experience" available to long-term memory buffers.
• While dissociation of such different processing levels is useful for conceptualisation purposes, in real life their activity is interdependent and extremely fluid (see e.g. Craig 2009 ); hence the perception of pain appears as an active process, continuously re-constructing itself by integration of sensory inputs with ongoing memories and internal representations (Gregory 1997, Loeser 2000, Nakamura and Chapman 2002).

2. IS THERE A "NEUROPATHIC PAIN MATRIX" ? "Constructing a model is making a bet" Anonymous

• Previous literature has abundantly discussed brain activation differences between experimental and neuropathic pain (NP), but systematic investigations in large samples of NP patients remain an exception.
• Functional imaging in NP cannot be examined with the same confidence as the experimental studies reviewed above.
• Some features observed in NP are reproducible across studies, while others remain controversial; some particularities have been described in single case reports, but not reproduced in larger series.
• In some cases, similar results have been interpreted in different ways.
• While the set of brain structures activated during neuropathic hyperalgesia and allodynia grossly correspond to those of the "pain matrix", a number of features have been described that are highly characteristic of the neuropathic state.

Thalamic hypoactivity in ongoing NP

• In one case with bilateral thalamic deafferentation, the thalamus contralateral to pain, albeit the less deafferented, showed the more severe hypoperfusion (Garcia-Larrea et al 2006).
• The contribution of deafferentation should not be neglected, however, as some thalamic asymmetry may persist even when differences in pain intensity are removed (Kupers et In summary, the functional thalamic depression in NP appears to reflect mechanisms that, once triggered by anatomical deafferentation, favour the transition to neuropathic pain.
• Resting thalamic hypoactivity may represent the metabolic counterpart of abnormal thalamic bursting observed in these patients.
• Before considering this phenomenon as a putative marker of neuropathic pain, direct comparison of thalamic activity in series of patients with similar lesions, but presenting or not with NP, appears mandatory.
• Longitudinal studies are also needed to determine whether or not patients with thalamic functional findings at NP onset are more likely to develop uncontrolled NP.

Provoked pain: allodynia and hyperalgesia (A/H)

• Provoked pain permits easier access to haemodynamic imaging than ongoing pain, and therefore studies of stimulus-evoked NP largely outnumber those assessing its continuous component.
• Provoked neuropathic pain can be contrasted with surrogate models of allodynia or hyperalgesia which generate experimentally controlled abnormal pain sensations.
• These models, mostly based on capsaicin injections, share with NP the induction of anomalous peripheral and central sensitisation, but lack the somatosensory deafferentation which is a key feature of NP.

Experimental hyperalgesia and allodynia. (Table 1)

• When compared with either a resting state or non-painful stimuli, experimental allodynia is consistently associated with activation in the posterior operculo-insular region, the anterior insulae, the mid-and anterior cingulate, and the posterior parietal and prefrontal cortices (Table 1 ).
• Responses in "3 rd -order" regions linked to emotional appraisal are more variable: activation of perigenual and orbitofrontal cortices was reported in a majority of experimental studies (i.e. Iadarola.

Neuropathic hyperalgesia / allodynia (Table 2)

• Grouped analysis of neuropathic A/H is hindered with drawbacks that have been well summarised by Kupers and Kehlet (2006) .
• Since the crucial variable was consistency, the authors may have underestimated important changes that only a few studies were able to tag.
• In other cases both posterior insulae were activated, but only the ipsilateral side remained significant after contrasting allodynic versus non-painful control stimuli (e.g. Schweinhardt et al 2006) .
• While activation enhancement in sensory regions has been similarly reported in experimental or neuropathic allodynia , a conspicuous feature of neuropathic allodynia is the lack of activation of ventromedial (perigenual and orbitofrontal) PFC.
• As a maladaptive consequence of persistent pain, lack of ventromedial responsiveness would not only change the subjective appraisal of the pain experience, but also limit the system's capacities to react adaptively to ascending pain signals.

Conclusions

• Data reviewed in this paper allow drawing some tentative mechanistic conclusions on the relation between brain activity and normal and abnormal pain sensations.
• Yet, although largely incomplete, existing data also show that functional imaging is now able to go beyond the phenomenological description of a 'physiological photograph', and propose testable hypotheses that will, or will not, come true in the following years.
• Above all, data from dozens of laboratories in the world underscore that pain, normal or abnormal, is an emergent property of the brain, lending substance to the nociception-perception-suffering model (Loeser 2000) .
• Arrows indicate significant differences in reported frequency.
• Activation of the ipsilateral opercular and insular cortices, relative to their contralateral counterparts, was more frequently reported during neuropathic than experimental allodynia (lower right panel).

Full text

HAL Id: inserm-00877368
https://www.hal.inserm.fr/inserm-00877368
Submitted on 28 Oct 2013
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-
entic research documents, whether they are pub-
lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diusion de documents
scientiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Pain matrices and neuropathic pain matrices: A review.
Luis Garcia Larrea, Roland Peyron
To cite this version:
Luis Garcia Larrea, Roland Peyron. Pain matrices and neuropathic pain matrices: A review.. PAIN,
Elsevier, 2013, 154 (Suppl 1), pp.S29-43. �10.1016/j.pain.2013.09.001�. �inserm-00877368�

Pain matrices and neuropathic pain matrices: a review
Luis Garcia-Larrea & Roland Peyron
Abstract: The “pain matrix” is conceptualised here as a fluid system composed of several interacting
networks. A ‘nociceptive matrix’ receiving spinothalamic projections (mainly posterior operculo-insular areas)
ensures the bodily specificity of pain and is the only whose destruction entails selective pain deficits. Transition
from cortical nociception to conscious pain relies on a second-order network including posterior parietal,
prefrontal and anterior insular areas. 2
nd
-order regions are not nociceptive-specific: focal stimulation does not
evoke pain and focal destruction does not produce analgesia, but their joint activation is necessary for
conscious perception, attentional modulation and control of vegetative reactions. The ensuing pain experience
can still be modified as a function of beliefs, emotions and expectations through activity of 3
rd
-order areas,
including orbitofrontal and perigenual/limbic networks. The pain we remember results from continuous
interaction of these subsystems, and substantial changes in the pain experience can be achieved by acting on
each of them. Neuropathic pain (NP) is associated with changes in each of these levels of integration. The most
robust abnormality in NP is a functional depression of thalamic activity, reversible with therapeutic
manoeuvres and associated with rhythmic neural bursting. Neuropathic allodynia has been associated with
enhancement of ipsilateral over contralateral insular activation, and lack of reactivity in
orbitofrontal/perigenual areas. While lack of response of perigenual cortices may be an epiphenomenon of
chronic pain, the enhancement of ipsilateral activity may reflect disinhibition of ipsilateral spinothalamic
pathways due to depression of their contralateral counterpart. This in turn may bias perceptual networks and
contribute to the subjective painful experience.
Key words: pain matrix; functional imaging; neuropathic pain; insula; thalamus; cingulate
1. TRIALS AND TRIBULATIONS OF THE PAIN MATRIX
“Concepts are tools: they wear out by ensuring their function”
Claude Bernard
The concept of “pain matrix” defines a group of brain structures jointly activated by painful stimuli. It
owes much to the notion of “neuromatrix” developed by Ronald Melzack, who proposed that the
anatomical substratum of the physical self is a network of neurons extending throughout widespread
areas of the brain (a “neuromatrix”), and generating characteristic patterns of neural impulses
distinguishing each bodily sensation. In Melzack’s words, “the neuromatrix for the physical self (..)
generates the neurosignature pattern for pain” (Melzack 1990).
The Pain Matrix (PM) notion represented a conceptual advance over many prevailing
concepts, which viewed pain-related emotional and cognitive phenomena as “reactions to”, rather
than components of pain (e.g. Hardy et al 1952). As early as 1968, Melzack and Casey suggested that

the pain experience reflected interacting sensory, affective and cognitive dimensions which could
influence each other. This view, of which the PM was an expansion, implied that there was no such
thing as a “brain pain centre”: pain was considered multidimensional and produced by distributed
neural patterns, usually triggered by sensory inputs but potentially generated independently of
them.
These theoretical notions were rapidly endorsed by functional imaging studies. Using
positron-emission tomography (PET), two seminal papers in 1991 reported that noxious stimuli
activated a distributed pattern of brain structures consistent with the notion of a ‘pain matrix’ (Jones
et al 1991, Talbot et al 1991). A bulk of consistent data rapidly accumulated showing not only
distributed activity to noxious inputs, but also linear and non-linear correlations between energy of
the stimulus, subjective perception, and PM responses (e.g. Bornhovd et al 2002, Coghill et al 1999,
Derbyshire et al 1997). It became also clear that most of the activated areas were not specific for
pain: PM regions such as the anterior cingulate cortex (ACC), the anterior insula (AI), or the
prefrontal and posterior parietal areas (PFC, PPC) showed enhanced activity in a wide range of non-
pain experiments, especially in emotionally or cognitively-laden contexts, whereas the sensory
encoding of noxious intensity was reflected by very tiny brain activations (Peyron et al 1999,
Ploghaus et al 1999, Sawamoto et al 2000, Bantick et al 2002, reviews Peyron et al 2000a, Apkarian
et al 2005).
Among the contrasting conceptual positions that have emerged since, some authors posit
that functional imaging data may contain a genuine and objective “signature” of the painful
experience (e.g. Tracey & Mantyh 2007, Wager et al 2013), which for a number of investigators may
imply that functional imaging could be used to derive an ‘individual pain phenotype’ (see Borsook et
al 2010, Cecchi et al 2012, Lu et al 2013, and comment in Apkarian 2013). Pushing this logic further,
activation of PM subsets (essentially the anterior insulae and cingulate) has been sometimes equated
with physical pain, leading to questionable conclusions such as that “social rejection hurts physically”
(e.g. Eisenberger et al 2003, MacDonald & Leary 2005). In contradistinction, other investigators have
assailed the very concept of a specific pain-related network, claiming that most, if not all, the regions
present in the PM represent a non-specific salience-detection system for the body, activated by
relevant events “regardless of the sensory channel through which these events are conveyed”
(Iannetti et al 2008, Iannetti and Mouraux 2010; review Legrain et al 2010). In between, the idea
that the pain matrix cannot be unequivocally defined, the role of different regions being dependent
on the context where stimuli are delivered, has been put forward by a few investigators (e.g. Peyron
et al 2000, p.282; Tracey and Mantyh 2007, p. 379).

Because spinothalamic projections inform brain networks on the bodily nature of the input,
the healthy brain recognises at once whether a menacing signal arrives through a somatosensory
channel. Information on the somatic origin is likely to be transferred to PM regions secondarily
ignited, hence giving to perceptive networks a stable reference to the own body. In this review, the
neural substrate of the pain experience will be conceptualised at different levels of progressively
higher-order cortical networks, from cortical nociception to the conscious experience we call “pain” –
itself subject to reappraisal by internal states, feelings and beliefs prior to stabilisation into memory
stores.
First-order processing: a nociceptive cortical matrix
The primate spinothalamic system (STS) chiefly originates from neurons in spinal laminæ I, V,
and VII, whose axons terminate in multiple nuclei of the posterior thalamus, essentially the ventral
posterior, centrolateral, mediodorsal and posterior group nuclei (Apkarian and Hodge 1989a,b;
Mehler 1962, Rausell et al 1999). Using trans-synaptic viral transport, main spinothalamic cortical
targets in primates were defined in the posterior insula (~40%), medial parietal operculum (~30%)
and mid-cingulate cortex (~24%) (Dum et al 2009). These receiving regions are the source of the
earliest responses to noxious stimuli recorded in the human brain (Frot et al 1999,2012; Lenz et al
1998c,d; Ohara et al 2006) and contain a ‘nociceptive matrix’ specific for spinothalamic projections.
The posterior insula and inner operculum are the only regions in the human brain where stimulation
triggers acute pain (Mazzola et al 2006, 2012), where focal lesions entail selective pain deficits
(Biemond 1956, Greenspan et al 1999, Garcia-Larrea et al 2010), and where cortical injury gives rise
to neuropathic pain (Biemond 1956, review Garcia-Larrea 2012). Lenz et al (1995) reported “full pain
experiences” when stimulating thalamic regions projecting to the posterior insula and operculum,
and intense pain was specifically reported during surgical dissection of the operculo-insular area,
(Pereira et al 2005). Focal epileptic activity in the posterior insula can trigger painful seizures by
igniting other PM areas in less than 100 ms; in a recent report, painful seizures were stopped by
millimetric thermo-coagulation of the posterior insular focus, and the patient has remained free from
pain seizures for more than 2 years (Isnard et al 2011).
However, while this ‘nociceptive matrix’ appears as a necessary entry to generate
physiological pain experiences, it cannot provide the countless nuances that characterise human
pain, and is unable to sustain consciousness: indeed, activation of the nociceptive matrix persists
during sleep, coma or vegetative state (Bastuji et al 2012, Boly et al 2008, Kassubek et al 2003). The
transition from cortical nociception to conscious pain and its multiple attentional-cognitive
modulations needs the recruitment of a second set of cortical networks.

From nociception to pain: a second-order perceptual matrix
The ‘classical’ PM encompasses activity in many areas distinct from the nociceptive network
described above, the most consistent being the mid- and anterior insulae, the anterior cingulate,
prefrontal and posterior parietal areas, and with less consistency the striatum, supplementary motor
area (SMA), hippocampus, cerebellum, and temporo-parietal junction. These second-order areas
share a number of features: (i) none of them is a direct target of the spinothalamic system; (ii) direct
stimulation does not evoke pain; (iii) selective destruction does not induce analgesia; (iv) they are
also activated in contexts not involving pain, and (v) their contribution to the PM, from nil to
predominant, depends on the context where noxious stimuli are applied.
The mid- and anterior insulae participate almost constantly to the PM. Their activation may
reflect a posterior-to-anterior information flux within the insula (Frot et al 2013, Pomares et al 2012),
supporting the transformation of sensory events into vegetative reactions and associated internal
feelings (Caruana et al 2011, Craig 2009, Wicker et al 2003). The “cognitive section” of the anterior
cingulate (BA 24–32) is also consistently activated by painful stimuli (review Vogt 2005), and together
with prefrontal and posterior parietal areas is thought to sustain attentional and evaluative
processes of anticipation, learning and cognitive control. The contribution of these areas to the PM
varies enormously with contextual factors (e.g. Buchel et al 2002, Peyron et al 1999, Seminowicz et
al 2004; review Apkarian et al 2005), and their activation can be dissociated from actual stimulus
intensity (e.g. Bornhovd et al 2002, Peyron et al 2007b). Activity in this 2
nd
-order contextual matrix
can influence the nociceptive areas via top-down projections: attending actively to noxious stimuli
enhances activity in sensory and orienting areas receiving STS afferents, such as the posterior insula
and mid-cingulate, whereas distraction tends to suppress such activities (Bantick et al 2002, Garcia-
Larrea et al 1997, Lorenz et al 2005b, Ohara et al 2004, 2006; Valet et al 2004, review Wiech et al
2008). Such top-down influences modify perception by changing the sensory gain ‘at the source’, i.e.
in cortical receiving areas (e.g. Garcia-Larrea et al 1991), thalamus (Vanhaudenhuyse et al 2009), and
even at brainstem (Tracey et al 2002) and spinal cord (Sprenger et al 2012, Willer et al 1979). Further,
vegetative peripheral reactions driven by anterior insular networks generate new ascending
information through splachnic and vagus nerves, STT and dorsal column systems, thus providing new
input to cortical and subcortical nociceptive targets (Tattersall et al 1986, Willis et al 1999). Of notice,
hypnotic suggestions of hypo/hyperalgesia appear to influence either nociceptive, second-order or
both matrices depending on the instructions given to the subject (eg Faymonville et al 2003,
Hofbauer et al 2001, Rainville et al 1997,) consistent with its action through top-down influences. A
dissociation between preserved activity in posterior insula but abated response in 2
nd
-order parietal
and temporal cortices has also been described under hypnosis (Abrahamsen et al 2010).

Frequently asked questions

Q1. What are the contributions in "Pain matrices and neuropathic pain matrices: a review" ?

This in turn may bias perceptual networks and contribute to the subjective painful experience. 

Q2. What is the effect of prolonged immobilization on ventromedial PFC?

Chronic stressors entail morphological changes in ventromedial PFC: prolongedimmobilization simplifies the branching and shorten the apical dendrites of rat ACC neurons –a damage reversible following a stress-free period (Radley and Morrison 2005). 

Q3. What is the importance of the reappraisal matrix?

Of importance, areas involved in the ‘reappraisal’ matrix such as the perigenual cingulate andorbitofrontal cortices are themselves strongly interconnected with subcortical regions crucial for descending pain control (notably the periaqueductal grey matter). 

Q4. What are the important questions incompletely solved?

Four questions incompletely solved are (a) its specificity regarding NP; (b) the role of sensory deafferentation; (c) the causal or consecutive nature of the dysfunction, and (d) the possible mechanisms leading to decreased local metabolism. 

Q5. What is the role of the ventromedial PFC in the control of unpleasant emotions?

In particular, the ventromedial PFC (orbitofrontal, perigenual) is considered as a region involved in the voluntary control of unpleasant emotions (Levesque et al 2003, Ohira et al 2006).