Pain that persists well beyond the healing of the injury, including surgical wounds, is a major clinical problem 1. Multiple mechanisms are likely responsible for persistent postinjury pain; some may be related to the transition of acute pain to chronicity, a concept that is supported by reports of the relationship between the intensity of acute postoperative pain and subsequent development of chronic pain after surgery 234. A long-lasting imprint of acute pain may contribute to its transition to chronic pain. In the rat model, acute pain and hyperalgesia following an inflammation disappeared within a few days, but enhanced sensitivity to a new noxious challenge (inflammation) may persist for a very long time 56. Our previous experiments with repeated injections of carrageenan into the hindpaws demonstrated that, well after recovery (2 to 4 weeks later) from hyperalgesia induced by the initial inflammation, repeated-crossover carrageenan injection causes a distant (in contralateral hindpaw) hyperalgesia that was absent after the initial injection. The acute pain and hyperalgesia following an injection of carrageenan disappears within a few days, but despite recovery the enhanced response to a new noxious challenge may persist for a very long time 57. This phenomenon was viewed as an indication of the long-lasting imprint of acute pain in the CNS 7.

It seems plausible that the transition to chronic pain involved modification of gene expression within the spinal cord, which might be a mechanism of long-term imprinting of pain hypersensitivity. However, the picture of gene expression alterations and their role in control of acute and chronic pain is far from clear 8910. Multiple pathways seems to be involved, and their elucidation will require global profile studies of gene expression using different models of acute and chronic pain with and without drug intervention. We have found no studies on changes in gene expression long after recovery from inflammatory hyperalgesia. The aim of this study was to characterize changes in gene expression when animals are long recovered from hyperalgesia but have maintained a propensity for enhanced hyperalgesic response. We hypothesized that following the recovery from inflammatory hyperalgesia there were persistent changes in spinal cord gene expression.

Carrageenan injection into a rat's hind paw induces inflammation and hyperalgesia (Fig. 1). Inflammation measured as increased paw volume reaches its maximum 3–24h after injection. The differences in paw volume between the injected and contralateral paws are minimal after 5 days and completely disappear in 2 weeks (Fig. 1B). Injection-induced hyperalgesia disappears at 24h (Fig. 1A). However, as was shown in previous experiments, after recovery from hyperalgesia a repeated-crossover injection of carrageenan (in the opposite hindpaw) produced a more exaggerated response, resulting in profound distant hyperalgesia in the originally injected paw even 28d later 7. In order to distinguish the acute changes induced by inflammation from long-term changes related to pain memory, we collected tissue samples both at 24h following injection, when the inflammation is still significant but there is no hyperalgesia, and at 28d, when as reported previously 7 there is no inflammation but enhanced response to pain stimulus, or "pain memory."

The statistical analysis of gene expression data is still a challenge. There is no standard approach to select differentially expressed genes. Often the combination of p-values and fold differences is used to select "interesting" genes for further analysis 10. However, to use fold changes for the central nervous system is not efficient, since the changes in gene expression are small and diluted by the large variety of unaffected neurons. In addition, fold changes are not statistically validated metrics and in many cases produce spurious results 11. In this study, we used the Rosetta Resolver statistical model and based our analysis only on p-value. We compared the number of differentially expressed genes at different p-values. For further analysis, we have used p < 0.01 as a cutoff. It was shown that selecting genes based on p < 0.01 results in more than 80% confirmation rate by real-time PCR 12 and provides a reasonable compromise between stringency and exploratory value of the analysis (Table 1). Using this method we have identified 629 differentially expressed genes in the 28d group and 512 genes in the 24h group out of a total of 31,042 reliably measured transcripts (see Additional file 1). These numbers include both annotated and uncharacterized genes (EST). Further analysis was limited to 356 full-length annotated genes that were differentially expressed compared with the control group at 28d after injection and 329 full-length annotated genes that were differentially expressed compared with the control group at 24h after injection (see Additional file 1). The low variability of our data permitted us to detect highly significant differences with relative changes as low as 10%.

In the next step we have characterized the functional groups based on gene ontology (GO) classification using DAVID tools (Table 2 and 3). There are several groups overrepresented only at 24h, when inflammatory response is near its peak (Table 2): protein synthesis/ribosomal protein group and proteins involved in the immune response. Chaperones involved in protein folding were also overrepresented at 24h but with smaller probability. The different groups were overrepresented at 28d (Table 3), such as the genes involved in cell-to-cell interaction, cell proliferation and neurogenesis, morphogenesis, and neuron differentiation–in other words, involved in the formation of new synapses and neurogenesis. DAVID-based GO analysis showed that genes responsible for regulation of apoptosis, phosphorylation, and acetylation were also overrepresented only at 28d group (Table 3). Several functional groups related to regulatory pathways were overrepresented both at 24h and 28d, for example, proteins responsible for ion binding (Table 2 and 3) but it may reflect their overrepresentation in neurons since we have used Affy chip as background for analysis (see Methods). Using SOM, we have grouped the known differentially expressed genes into eight clusters (Fig. 2). Clusters A and B contain the majority of selected genes (n = 233). Cluster A (n = 122) includes genes that increased at 24h after injection and then increased even further at 28d, whereas genes belonging to cluster B (n = 111) changed in the opposite direction: They decreased at 24h and further decreased at 28d. Clusters C and D are similar and include 51 genes that are practically not altered 24h after injection but then slightly increased (cluster C) or slightly decreased (cluster D) at 28d. Genes (n = 177) belonging to clusters E and H increased at 24h (cluster E) or decreased (cluster H) and remained at the same level up to 28d after injection. The last two clusters (G and F) include genes that increased (F) or decreased (G) at 24h and returned to the control level at 28d.

In order to see whether clusters G and F reflect changes in gene expression related to ongoing inflammation, we have selected the group (n = 16) of known mediators of inflammation. They have a similar profile with maximal changes at 24h following injection and practically expressed at the same level as the control group 28d following injection (Fig. 3).

As indirect validation of our result, among differentially expressed genes in both groups we identified (see methods) 67 genes (39 in the 28d group) previously reported to be involved in pain-related pathways, altered in different models of pain, or interacting with protein involved in pain. These genes were divided into several functional groups (Table 4). The first group includes membrane receptors and their ligands. The second contains ion channels, regulators of ion homeostasis, and calcium sensors (e.g., calmodulin) (Table 3). We selected 2 groups that contain signal transduction molecules: the MAPK pathway and proteins involved in cAMP signaling. The fifth group consists of growth factors known to modify pain sensitivity: BDNF 13 and its receptor (NtrkB), neuregulin1, and IGF binding proteins. The sixth group includes critical components of the arachidonic acid (AA) cascade, a known mediator of inflammatory pain.

Several genes ("Others", group 7) did not fit in any of these groups. Gene expression of this group (Table 4) has been reported to be altered in different pain models. Annexin 3 level was increased during episodes of migraine 14; microtubule-associated protein 1B was altered 7 days after chronic constriction injury (CCI) 15; synaptoporin, a major synaptic vesicle protein in Adelta- and C-fibers and co-localized with calcitonin gene-related peptide (CGRP) in sensory primary afferent neurons, increased after peripheral nerve injury 16.

Several genes were included in group 7 based on indirect evidence of their involvement in pain regulation. Tubulin isoform beta5 interacts with vanilloid receptor (TRPV1) and is probably a downstream effector of TRPV1 activation 17. Chaperonin subunit 4 (cct4) involved in folding tubulin and other cytosolic proteins is included, since mutation of this protein leads to early onset sensory neuropathy 18. Fyn proto-oncogene phosphorylates NMDA receptors in response to pain stimuli 19.

In order to verify the microarray result by independent methods, we randomly selected six of the differentially expressed genes and two non-differentially expressed genes to measure their expression by qPCR (Fig. 4). We found significant (p < 0.001) correlation between expression value obtained using microarray and the real-time PCR results at 28d following carrageenan injection.

The acute pain following a single injection of carrageenan disappears within a few days; however, the enhanced sensitivity to a new noxious challenge may persist for a prolonged time 567. The enhanced sensitivity may contribute to posttraumatic and postsurgical persistent pain. In experiments with repeated carrageenan injections in response to the first injection into a rat's hind paw, the induced hyperalgesia was profound on the side of injection and minimal or absent in the contralateral hind paw. However, repeated injection of carrageenan into the previously noninjected hind paw resulted in pronounced hyperalgesia in the other paw. The difference between distant hyperalgesia after the initial and repeated-crossover injection of carrageenan was used as a measure of the hyperalgesia-related memory. The enhanced hyperalgesia is present even 28 days after the initial injection 7.

In this study we found alterations of gene expression that lasted at least 28d and were quite different from transitory changes observed 24h after injection. The genes altered at 24h include those involved in immune reaction and protein synthesis (Table 2), whereas long-lasting changes observed 28 days after injection indicate different pathways, most notably those responsible for new synapse formation (Table 3).

The observed short-term upregulation of the immune response genes concurs with earlier reports that IL-6 2021 and TNF-a (tumor necrosis factor-a) 21 play an important role in inflammatory pain. The upregulation of these genes had previously been associated with the development of neuropathic pain as well 22. A similar result has been found following nerve injury and spinal cord injury (SCI). SCI induces a robust and significant increase in mRNAs of inflammatory cytokines such as TNF-a, interleukin-1β (IL-1β), and IL-6 at 1, 3, and 24 h post-injury 232425. We do not know whether the changes in gene expression are mediated by inflammation, pain, possible injury to the nerve, or a combination thereof.

We observed significant changes in expression of several key enzymes (3-alpha-hydroxysteroid dehydrogenase, phospholipase a2, prostaglandin d2 synthase, and COX2) involved in eicosanoid biosynthesis (Table 4, Arachidonic acid cascade and mediators of inflammation). This cascade is usually initiated by the activation of phospholipase A2 and the release of arachidonic acid (AA) 26. The AA is subsequently transformed by cyclooxygenase (COX) and lipoxygenase pathways to prostaglandins, thromboxane, and leukotrienes, collectively termed eicosanoids. Eicosanoid production is considerably increased during inflammation and inflammation-induced pain, and COX is the major target for nonsteroidal antiinflammatory drugs (NSAIDs). It is possible that activation of eicosanoids by inflammation/pain causes lasting changes in gene expression. NMDA receptors, which are prominently involved in activity-dependent synaptic plasticity and tonic pain 27, could mediate the upregulation of inflammatory factors. Inflammation-induced NMDAR activation involves phosphorylation of the NR1 and NR2B subunits in the spinal dorsal horn by fyn proto-oncogene 28. The downstream changes in Egr1, MAPK, and AC expression (see Table 4) can be also induced by activation of the NMDA receptor. As an alternative, activation of the neurotensin receptor can stimulate Egr1 expression and MAP kinase pathways 29.

We do not know whether activation of eicosanoids and MAP kinases is part of one inflammatory/pain pathway or the former is the result of inflammation and the latter is pain-induced. In any case, long-lasting changes in gene expression are unlikely to be induced by ongoing inflammation, since the majority of genes known to be involved directly in the inflammatory response have a profile with maximum changes at 24h and returning to control levels 28d following injection. No detectable changes of inflammatory cytokine genes, including those described above, were found 28d after injection. However, initial inflammatory response may contribute to long-term changes in gene expression and pain memory. A number of genes altered in the 28d group are involved in inflammation-induced potentiation of pain sensitivity. For example, dynorphin 3031, neuropeptide Y (NPY), and NPY (Y1) receptor 32 are induced in the spinal cord by peripheral inflammation. However, it is unknown whether activation of nociceptors is required.

Based on our expression data, carrageenan injection affected at least two signal transduction pathways, MAPK and cAMP/adenylyl cyclase (AC), known to regulate nociceptive signal perception and transmission. Several isoforms of MAPK are involved in regulation of acute and chronic pain both in neuron and glial cells 3334. cAMP mediates many aspects of pain transmission within neuronal cells. In particular, AC isoform 8 (AC8) that couples NMDA receptor activation to cAMP signaling pathways in neurons are important in the development of persistent pain 35. Activation of MAPK cascade might be mediated by neurotensin receptor via Egr1, which is known to be upregulated in response to persistent inflammatory pain 29 and stimulates Erk1/2 phosphorylation 36. It is likely that an interaction between the glutamate pathway and cAMP signaling is mediated by modulation of metabotropic glutamate receptor by RGS2 and RGS4 3738, which were downregulated 24h after carrageenan injection (Table 4). Based solely on our data, we cannot tell whether the trigger for long-term changes in gene expression is the initial inflammation or pain or their combination. It is also unlikely that all changes in gene expression observed in the 28d group are related to formation and maintenance of nociceptive memory. However, the prominent group (Table 3) of differentially expressed genes controls neuronal connectivity, synaptogenesis, and neurogenesis. It is known that long-term memory and plasticity in CNS depend on formation of new synapses that require synthesis of proteins responsible for cell-to-cell interaction–in particular cadherin-like protein. There is also evidence that formation of new neurons is important for memory formation and plasticity 39.

There is similarity in the development of long-term potentiation (LTP), long-term memory, and pain. Research indicates that central sensitization in the spinal cord has the identical mechanism to LTP 40. We have found that a number of genes altered at 28d were also altered in the hippocampus during LTP: brain-derived neurotrophic factor (BDNF), early growth response 1 (EGR1), CD9 antigen (CD9), neuropeptide Y receptor Y5 (NPY5R), and neuropeptide Y receptor (NPY1R) 41. Genes such as BDNF 1342, NtrkB (BDNF receptor), Egr1 29, neuropeptide Y receptor, and neuregulin1 are also known to be involved in inflammation-induced pain. For example, BDNF, a known modulator of memory, was found to be altered in rat pups one day after peripheral inflammation induced by injection of Freund's complete adjuvant 43. We do not believe, however, that pain memory and LTP are identical phenomena but rather share some basic mechanisms. They have different anatomical substrates, and most alterations found in our study were not observed in the hippocampus following memory formation. Based on our data, pain or inflammation or a combination thereof induces dramatic changes in gene expression. Most of these changes in gene expression subside when inflammation disappears, while some of them persist and even increase 28d later. Thus, part of the pain- and inflammation-induced changes in gene expression belong to pathway(s) that remain activated long after inflammation and acute pain disappear. We can only speculate which genes whose expression increased much more at 28d than at 24h belong to these pathways. Additional experiments are needed to narrow down the list of affected pathways. Based on the results of GO analysis, proteins involved in synaptogenesis, cell-cell interaction, and the formation of new neurons are overrepresented among differentially expressed genes at 28d after carrageenan injection. Thus, we propose that pathways related to synapse formation between newly generated neurons are particularly important for "pain memory." The tentative nature of this conclusion depends on the exploratory nature of our microarray data and analysis. The selection of genes is based on a limited number of replicates and multiple comparisons, which implies that a number of genes may be false positives. We confirmed the differential expression for a few selected genes by PCR, but more research is needed to complete verification of the microarray data.

We have observed long-lasting changes in gene expression following recovery from carrageenan-induced hyperalgesia. On the basis of these data, we propose that pathways related to synapse formation are involved in the formation of pain memory.

The study was approved by the Harvard Medical Area Standing Committee on Animals, Harvard Medical School. Experiments were performed on male Sprague-Dawley rats weighing 250–300 g that were housed with a 12-h light-dark cycle and food and water available ad libitum.

Gene expression in the lumbar part of the spinal cord was compared in three groups of rats: 28d after carrageenan injection, 24h after carrageenan injection, and a control group without injection.

Inflammation was induced by injection (26-gauge needle) of 0.1 ml of 2% carrageenan (Sigma Chemical Co., St. Louis, MO) in the plantar surface of the hind paw under halothane (2%) anesthesia. The rat paw volume was measured by a plethysmometer (Paw Volume Meter; Ugo Basile). The general procedure for measuring changes in behavioral responses for microarray experiments was as follows: For several days after arrival, rats were placed in the testing environment, and during the two days preceding the experiment, pressure threshold and hind paw volumes were measured. On the experimental day, basal values were determined twice with a 30-min interval; the average value of two readings was used as a baseline. The rats were assigned randomly (blocked randomization) to one of three groups. Carrageenan was injected into the right hind paw, and the rat was decapitated 24h or 28d after injection. Paw volume was measured on both sides 2h, 4h, 24h, and 28d after each of the carrageenan injections. We assumed that pain-related memory is most likely mediated by altered gene expression both in ipsilateral and contralateral parts of the lumbar enlargement of the spinal cord. In this study we have limited the measurement of gene expression to the ipsilateral portion that is most affected by carrageenan injection.

The author(s) declare that they have no competing interests.</