We applied a distinctive genome-wide search strategy to rat selection-lines that have similar genetic background, but contrasting pain phenotype. The aim was to discover miRNAs related to neuropathic pain while avoiding processes such as cell metabolism, regeneration and apoptosis that are also triggered by nerve injury but that do not differ between the lines . Overall we encountered only three miRNAs that were present in significant copy number and were regulated differentially by nerve injury in HA vs. LA rats. These were rno-miR-30d-5p and rno-miR-125b-5p which were expressed abundantly and rno-miR-379-5p which was expressed at modest copy number. Eight additional diff-reg miRNAs were found. But since they were rare, it is unlikely a priori that they are functionally important. This determination rests on the observation that most miRNAs for which a significant biological role has already been established are abundantly expressed. Furthermore, experimental use of decoy libraries has shown that in general, miRNAs expressed at 100 RPM or less are unlikely (<2%) to detectably suppress the expression of any target mRNA , let alone to have a prominent behavioral effect. Indeed, the 8 low-abundance miRNAs might not have been differentially regulated at all. Some or all of these hits probably met the significance criterion by chance (type 1 error). Since the ~3.5 × 106 miRNA reads in each pool represent only ~0.001% of the actual number of miRNA molecules loaded into the sequencer, stochastic effects can generate false positives, especially for low-abundance sequences . Consider that a random (or read-error) difference of only a few reads on a baseline RPM = 10 translates to a much larger fold-change than the same error on a baseline of RPM = 1000.
Genome-wide scans typically yield large numbers of hits requiring prioritization based on factors such as magnitude of fold-change ratio or occurrence in several different pain models . This is risky and possibly even counterproductive as it might highlight processes such as regeneration and metabolism that are generic to nerve injury but not necessarily to pain. The approach is also likely to obscure transcripts related to particular pain conditions but not others. Our approach sidesteps the need for secondary screens for obtaining a short-list. Indeed, we cast a broad net which included diff-reg miRNAs obtained with and without exclusion of outlier pool HA_SNL1, two different methods of calculating regulation and canonical miRNAs of unique sequence diff-reg isomiRs. Nonetheless, a very small number of diff-reg miRNAs were found and these turned out to be regulators of established major pro-nociceptive molecules. This affirms the intrinsic power of the strategy to prioritize pain-relevant transcripts .
We believe that the key factor was comparison of animals with very similar pedigree but strongly contrasting pain phenotype. Standard commercial strains show robust differences in constitutive miRNA expression  while in HA/LA rats we found only two miRNAs with a constitutive expression difference at baseline (sham-operated). Even these are questionable given their very low abundance (RPM ~ 10). Another factor was harvesting DRGs only 3 days postoperatively. Up- or down-regulation of miRNAs after nerve injury is progressively more robust over the first weeks postoperative. By 2–4 weeks a much larger fraction of expressed miRNAs are regulated [15, 38, 39]. Since differential pain phenotype in our rats is robust by 3 dpo  sequences that begin to show regulation only later are unlikely to be causative. We conclude that rno-miR-30d-5p, rno-miR-125b-5p, and perhaps rno-miR-379-5p are fundamental to the contrasting neuropathic pain phenotype in HA vs. LA rats. Since the contrast includes spontaneous pain as well as tactile allodynia  these miRNAs may be important players in neuropathic pain in general. Regulation of these miRNAs in the DRG following nerve injury has been noted previously (e.g. ), but they have not been specifically implicated as key factors in neuropathic pain.
Downstream targets of the differentially regulated miRNAs
The 8 rare diff-reg miRNAs, and the modestly expressed rno-miR-379-5p, had only 3 verified mRNA targets among them, none that hinted at a role in pain physiology. The two abundantly expressed diff-reg miRNAs had more, including 3 that code for well-known pro-nociceptive compounds, TNF, BDNF and STAT3 (Table 3). This represents notable enrichment of pain-related molecules. Since both miRNAs were more down-regulated in pain prone HA rats than in LA rats, and miRNAs suppress the expression of their mRNA targets, this regulation is expected to increase target gene product levels in HA vs. LA. Indeed, TNF and BDNF are upregulated and released by DRG neurons and glia in inflammatory and neuropathic pain models and are strongly pro-nociceptive [25, 26, 28, 29, 40, 41].
Interestingly, specific knockout of BDNF expression in small diameter neurons was found to have little effect on pain behavior 3 days postoperatively, indicating that BDNF in nociceptors does not play a significant role in neuropathic pain phenotype . However, BDNF expression in larger DRG neurons (Aβ-afferents) and glia is not eliminated in these knockout mice. Indeed, as noted, BDNF expression is markedly increased in large diameter afferents following axotomy [28, 29]. This reiterates the special role of activity in Aβ afferents as drivers of neuropathic pain . In addition, it suggests that regulation by miRNAs might be a particularly important factor in the phenotypic switching of Aβ touch-signaling neurons, a change that contributes importantly to their seemingly paradoxical involvement in pain perception in the event of neuropathy [5, 44].
STAT3, a transcription factor in the JAK-STAT pathway, is activated in sensory neurons and glia following noxious stimulation and nerve injury and is also associated with potentiation of pain [31, 45]. Thus, the selective downregulation of all 3 diff-reg miRNAs likely contributes to the prominent afferent hyperexcitabilty of HA rats  and hence to their marked neuropathic pain phenotype.
As noted above, Tnf, Bdnf and Stat3 have already been verified experimentally as targets of rno-miR-30d-5p and rno-miR-125b-5p, and functional assays involving primary afferents have already established all three as major players in pain physiology. Additional corroboration has therefore not been undertaken here, although the issue of whether miRNA-regulated expression is selective to particular types of DRG neurons, but not others, deserves follow-up. We conclude that neuropathy-induced regulation of these diff-reg miRNAs contributes to the control of individual variability in neuropathic pain phenotype. Further research is required to determine whether some of the other known or predicted gene targets of rno-miR-30d-5p and rno-miR-125b-5p might also contribute. In this regard it is noteworthy these targets in part relate to inflammatory processes and not strictly to processes of neuropathic (electrical) excitability. Indeed, it is clear that inflammatory mediators exacerbate ectopia in afferent neurons, blurring the line between neuropathic and inflammatory pain [1, 15, 26, 29, 41].
A key strength of whole-genome search strategies is their ability to reveal elements not previously associated with pain processing. TNF, BDNF and STAT3 were highlighted because of prior knowledge of their prominent role in pain. However, in vivo all of the genes targeted by our 3 miRNA hits are potentially down-regulated. We must imagine that the contrasting pain phenotype of HA vs. LA rats is orchestrated by differential regulation of the entire constellation of their target genes. The list of targets already validated, and those yet to be validated, constitutes a useful pool of pain-related genes. In this regard it should be noted that since RNA was extracted from whole DRGs, non-neural cells in the ganglia might also have contributed to the yield of diff-reg miRNAs.
Unique sequence miRNA isoforms (isomiRs) and other sncRNAs
Small sequence differences among isomiRs can affect binding affinity to mRNA targets [35, 46]. Thus, a particularly potent individual isomiR, or a small subset, could be responsible for a large proportion of the functional effect of the canonical miRNA. Indeed, in three instances the fold-change of one isoform reached statistical significance even though that of its canonical miRNA did not (Table 2). In each case the isomiR contributed only a small fraction of the overall RPM of the miRNA. The canonical miRNA of two of these, rno-101a-3p and rno-miR-130a-3p, had a variety of validated mRNA targets (n = 44, Table 3). Differentially regulated isomiRs merit further exploration. Likewise, diff-reg sncRNAs that are not miRNAs might be significant mediators of pain variability. Our reads level analysis identified a number of candidates. Many of these were parsed as tRNAs or rRNAs. Three of them were relatively abundant (Table 2) and showed signs of being processed rather reflecting random RNA degradation. These sequences also merit further study.
The contribution of miRNA regulation to neuropathic pain
Nerve injury is consistent in causing negative symptoms, hypesthesia and numbness, in the corresponding body part. Positive symptoms, paresthesias and pain, are much more variable indicating that the nerve lesion is not causative in itself. Among major limb amputees, for example, <25% report persistent severe phantom pain  while <0.01% of adults have trigeminal neuralgia among the ~17% who have the (putative) causative nerve lesion . Genetic polymorphism is believed to be a key co-factor determining whether or not nerve injury will lead to neuronal hyperexcitability and neuropathic pain. However, the widely used approach of simply identifying genes regulated by nerve injury is an inefficient route for discovering the polymorphisms and pathways that play a special role in pain processing. The problem is that thousands of mRNAs are regulated following nerve injury [21–23, 33]. It is very difficult to delineate which contribute to enhanced pain sensibility and which are related to other processes triggered by nerve injury. A systematic means is required to solve this problem. The strategy used here, examining differential effects of nerve injury in animals that are genetically similar but different in pain phenotype, represents a way forward. Remarkably few differentially regulated miRNAs and other sncRNAs were identified, and even fewer remained candidates after (tentative) exclusion on the basis of low copy number. Prioritizing candidate genetic elements using this strategy is likely to have general applicability to problems in which there is a variable link between the precipitating pathology and the resulting phenotype.