Developing nociceptor-selective treatments for acute and chronic pain

Abstract

Despite substantial efforts dedicated to the development of novel, non-addictive analgesics, success has been limited. Clinically available agents generally lack efficacy, creating pharmacoresistance, and are accompanied by undesirable side-effects. The traditional target-based drug discovery effort, while generating compounds with selectivity for single targets, has a high rate of attrition due to poor clinical efficacy. Here, we examine the problems associated with the current analgesic drug discovery model, and review evidence for a new approach for the identification of analgesic targets and compounds for the diverse forms of acute and chronic pain. Using human iPSC-derived nociceptors, genes selectively expressed in nociceptors, but not in other excitable cell types, can be identified, disease states can be modeled, phenotypic screens can identify compounds that selectively reduce nociceptor excitability and individual patient susceptibility to developing pain or responding to particular treatments can be identified. This human cell/genetics-based approach can identify nociceptor-specific and pain-disease selective compounds that modulate the function of multiple targets. This polypharmacology approach will enhance analgesic efficacy by broadening drug activity on complex pain mechanisms, while selective action on nociceptors will curb side-effects.

What is pain?

Although we use a single word, pain, to describe a broad range of unpleasant sensations that comprise feelings of soreness coupled with discomfort and unpleasantness, pain itself is complex and can be initiated by a range of quite distinct conditions. It is important to recognize this, as each pain condition may well require different therapeutic interventions(1, 2). There are aspects of pain that are truly physiological, our capacity to detect potentially damaging external stimuli: intense heat or cold, excessive mechanical force, and chemical irritants. This constitutes nociceptive pain, which is driven by the activation of high threshold nociceptor sensory neurons adapted to transduce such noxious stimuli into an inflow of sensory input into the central nervous system (CNS), where the acute pain the nociceptors evoke through activation of nociceptive circuits, helps us learn to avoid environmental danger(3–6). This pain is highly adaptive and essential for preventing damage in our daily interactions with the external world. Individuals who lack this damage warning system, such as those with congenital insensitivity to pain due to loss of function mutations in the Nav1.7 voltage-gated sodium channel or the TrkA receptor, typically damage their tongues and lips when they eat, their toes on walking, and fingers when exploring objects, and have no warning when they break a limb or have appendicitis, which reduces life expectancy(7, 8). Preserving nociceptive pain is essential, therefore, except during or post-surgery or immediately after major trauma.

Another adaptive pain is the inflammatory pain that occurs when tissue injury cannot be avoided or when inflammation occurs due to a pathogen invasion or pathological inflammation. Activation of the immune system in these cases leads to production of inflammatory mediators that act on nociceptors to both directly activate them(9–11) and sensitize them(12, 13), such that their threshold of activation drops so that low-intensity stimuli, such as light touch or movement of a joint, can now activate sensitized nociceptors, and these innocuous stimuli are now painful. This pain, at least in an acute setting, is adaptive in that the heightened pain and threshold drop promotes avoiding use of or contact with the inflamed area, which promotes healing and repair, however this pain typically needs to be reduced to a manageable level. In addition to the pain produced by the activation and sensitization of nociceptors, the augmented sensory inflow triggers use-dependent plasticity in the CNS, the phenomenon of central sensitization, that amplifies and spreads pain sensitivity beyond the primary area of damage/inflammation to neighboring secondary areas of non-inflamed areas, causing, for example, tenderness around a surgical wound or an inflamed joint, and prolonging the pain(14, 15).

For chronic inflammatory conditions, like rheumatoid arthritis, the ongoing inflammatory pain is not adaptive since the inflammation does not resolve and healing does not occur. In these situations, the combined chronic inflammation and pain constitute the clinical problem to be controlled, and the pain can be considered maladaptive or pathological(16). Finally, there are two categories of truly maladaptive pain, where the pain is not a reaction to some pathology but a disease state itself. The first of these is neuropathic pain, persistent pain due to lesions, most typically of the peripheral nervous system (PNS) (nerve injuries or neuropathies)(17, 18), but also of the CNS (after spinal cord injury or stroke)(19) that lead to pathological excitability in nociceptive circuits due to ectopic firing of injured neurons, a loss of the normal inhibitory circuits in the CNS - the phenomenon of disinhibition, neuroimmune interactions in the nerve and CNS, and structural and functional connectivity changes in the spinal cord and brain that alter circuit function and output(20–23).

The last major form of pain is dysfunctional or nociplastic pain, defined as chronic pain in the absence of noxious stimuli, active inflammation, or detectable damage to the nervous system(24–26). In individuals with this pain, there is abnormal functioning of the nervous system such that it amplifies and sustains essentially normal signals to a point at which they are perceived as being painful. The pain may localize to a particular organ, as in tension-type headaches, temporomandibular joint disease or irritable bowel syndrome, or may be diffuse as in fibromyalgia, a form of chronic widespread pain(27).

In addition to these quite distinct types of pain, we need to recognize that pain arising from the somatosensory system (skin/joints/muscle) may differ from that from visceral organs(28, 29), and that acute pain typically differs considerably from chronic pain in terms of mechanisms and response to analgesics(30). Furthermore, pain has several quite different manifestations, which reflect different mechanisms. Pain may arise spontaneously in the absence of a stimulus or detectable pathology or may need a stimulus to evoke it. Spontaneous pain is a major element of many clinical conditions, particularly neuropathic pain(31). Stimulus evoked pain includes exaggerated and prolonged pain in response to noxious stimuli, the phenomenon of hyperalgesia, as well as pain in response to normally innocuous mechanical or thermal stimuli, tactile or thermal allodynia(32). There is of course, a major difference between symptom suppression and disease modification in the setting of pain, the former requires ongoing treatment to reduce the experience of pain, while the latter has the potential to have long-lasting consequences, for example, interventions that can prevent the transition of acute to chronic pain or the development of axonal neuropathy, and hence neuropathic pain. To date, there are only symptom suppressor analgesic interventions.

We need to recognize the diverse nature and the multiple distinct specific mechanisms responsible for different pain conditions to be able to select those interventions with the greatest chance of efficacy. There is no point, for example, in blocking an inflammatory mediator or targeting a protein only induced by neuronal damage if a patient’s pain occurs in the absence of any inflammation or with a fully intact nervous system. We need to build an algorithm that enables identification of the most likely responders to particular therapies that act on distinct mechanisms, rather than relying on an empirical trial and error approach, as currently used. Such a precision pain medicine approach requires parallel efforts to 1. Understand and fully reveal all the mechanisms underlying different pains, 2. Define how to identify which mechanisms are operational in individual patients by careful phenotyping and utilization of biomarkers, and 3. The provision of a range of different therapies that interact specifically with each of the distinct mechanisms. To achieve the goal of developing analgesics with defined selective actions in particular patients will require an approach quite different from that conventionally used for the development of analgesic therapies. The nature of such a drug development approach and how to realize it using human stem cell-derived neuronal assays and screens, specifically for selectively targeting nociceptors, is the focus of this review.

Current pharmacological treatment

The current repertoire of analgesics (Figure 1) is limited to only a few of the defined types of acute or chronic pains and their underlying mechanisms, and overall have several major problems; generally low efficacy and multiple adverse effects, and for opioids, tolerance requiring increasing doses, dependency with withdrawal on discontinuation, and abuse liability(33–35).

Fig. 1: Site of action of current analgesics.

Clinically used analgesics act on diverse targets in the central and peripheral nervous system contributing to their lack of efficacy and numerous side-effects, highlighting the need for nociceptor-selective targets.

Available analgesics operate in the following ways: 1. Block electrical activity in neurons with non-selective sodium channel blockers – either as local anesthetics such as lidocaine or bupivacaine, or systemic sodium channel blockers, like carbamazepine, mexiletine, lacosamide, and amitriptyline(36). The lack of selectivity means that all excitable cells exposed to the blockers will be impacted such that the therapeutic index is limited, and adverse effects due to activity on non-nociceptor neurons are inevitable. 2. Block the action of inflammatory mediators on nociceptors – most notably cyclooxygenase inhibiting NSAIDs, which inhibit PGE2 synthesis(37, 38), as well as anti-TNF or anti-IL1 antibodies in particular inflammatory conditions(12, 39). Cox2 is induced in macrophages during inflammation and therefore Cox inhibitors are only effective in reducing the sensitization of nociceptors in those conditions where Cox activity is a major driver of this(40–42). In the context of postoperative pain relief, the NSAIDS diclofenac, ibuprofen and naproxen only provide at best 50% maximum pain relief for 4–6 hrs(43) but reduce opioid use with moderate success(44–46). Also, inhibiting PGE2 production during inflammation, although a widely used analgesic strategy, has several disadvantages on prolonged use, including bleeding, GI and CVS issues(47). Combination therapy using lower doses of NSAIDS together with acetaminophen has proven beneficial(48). TNF-α levels are elevated in inflammatory conditions and anti-TNF therapies can be used for treatment of rheumatoid arthritis, Crohn’s disease and psoriasis(49). Anti-TNF antibodies reduce cytokine production in joints, particularly, IL-1, IL-6, and TNF itself(50). Injectable therapies include soluble TNF receptors such as etanercept and PEG-sTNFR1, or TNF-α antibodies such as infliximab, adalimumab, and CDP-870. Small molecule inhibitors of TNF-α synthesis, such as p38 MAP kinase and PDE4 inhibitors are also used(51, 52). However, all these options are associated with side effects(51), with increased risk of congestive heart failure and resurgence of latent tuberculosis. Anti-IL-1 therapies are also used for the treatment of specific inflammatory conditions. IL-1 blockers such as anakinra, rilonacept, or canakinumab, provide relief from broad-spectrum inflammatory diseases including rheumatoid arthritis, osteoarthritis, osteomyelitis, traumatic joint injury, systemic juvenile idiopathic arthritis and gout. However, suppressing innate immune activity increases the risk of bacterial infections(53). 3. Reduce synaptic transmission from nociceptors to dorsal horn neurons, but the action of the drugs used for this purpose are not restricted to this, and therefore they block transmission quite broadly in the CNS, causing adverse effects. Such blockers include mu opiate receptor opioid agonists(54), alpha2delta calcium channel subunit agonists - the gabapentinoids(55, 56), the Cv2.2 inhibitor ziconotide(57, 58), and for migraine, CGRP receptor antagonists(59, 60). While most options for CGRP are monoclonal antibodies, a small molecule antagonist for its receptor is also proving to be equally effective(61). The expression of the alpha2delta calcium channel subunit increases substantially in DRG neurons after nerve injury(62, 63) explaining why GABApentin and pregabalin, which bind to it, have efficacy in neuropathic pain. Pregabalin is also used to treat nociplastic pain(64). 4. Promote inhibition in the CNS by mimicking the activity of inhibitory transmitters, again at present with no selectivity of action for nociceptive circuits, this includes dual amine uptake inhibitors like duloxetine(64–66), opioids(67), and GABA mimetics, like the benzodiazepines. Specific GABA-A receptor subtypes expressed in lamina II inhibitory interneurons of the spinal dorsal horn have been linked to analgesia. However, the sedative effects of the available non-selective allosteric modulators, such as benzodiazepines, dimmish therapeutic potential as analgesics(68–72). The highly conserved nature of the transmembrane domains of ion channel families, and GPCR receptor subtypes, pose a significant problem in identifying subtype-selective allosteric pockets for generation of subtype selective molecules. Efficacy-selective molecules rather than affinity-selective ones might offer a partial solution to this problem(73). 5. Finally, there are drugs that can reduce use-dependent plasticity in CNS: especially NMDAR antagonists like ketamine, which reduce synaptic plasticity(74), but once again the actions are not limited to changes in nociceptive circuits and there are in consequence adverse effects(75). Timing of treatment needs to be considered, since in some situations, pre-emptive treatment can be more effective than treating established pain(76, 77).

The two major problems with existing therapies are then, first, their non-selectivity (e.g. sodium channel blockers act on all excitable cell types, while NMDA and MOR receptors are widely distributed in the CNS). This reduces therapeutic index and means that analgesia is accompanied by undesired actions. These non-selective drugs have multiple non-pain-related actions (e.g. opioids reduce activity in nociceptive circuits but also generate euphoria, respiratory depression and constipation; gabapentinoids, reduce cognition and produce sedation, as well as a reduction in neuropathic hyperactivation, NMDA receptor antagonists have psychotropic actions, as well as reducing central sensitization. The second problem is the drugs available often only block one of many parallel pain triggering signals e.g. Cox inhibitors only block PGE2 production but there are other inflammatory mediators (IL1 or TNF, plus others) that will continue to act on nociceptors in the presence of Cox2 inhibition, substantially reducing efficacy. There is, therefore, an enormous need to both improve selectivity and broaden action to improve efficacy and reduce adverse effects.

There are targets that have some promise to fill some of these gaps that have moved toward clinical development. These include TRP channel antagonists (TRPV1 and TRPA1) to block nociceptor transduction(78, 79), inhibitors of acid-sensitive ion channels(80) or purinergic receptors(81, 82); anti-NGF antibodies to neutralize the pain-promoting actions of NGF produced by immune cells(83, 84); JAK/STAT pathway inhibitors as nociceptive cytokine signaling inhibitors(85), K channel openers to suppress neuronal excitability(86, 87); Nav selective blockers that act only on sodium channels expressed by nociceptors (Nav1,7 and Nav1.8 blockers)(88, 89); GABA receptor subunit-specific compounds(90, 91), inhibitors of tetrahydrobiopterin synthesis(92), angiotensin type 2 receptors(93), and attempts to modify opioid action on opiate receptors to retain analgesia but reduce euphoria/tolerance/dependence/respiratory depression(94–97). Some have been abandoned because of failure in the clinic and some are still progressing(34), but none have yet resulted in a new drug.

Current obstacles in pain treatment

The obstacles for the development of effective pain treatments that have not yet been overcome include then, the very limited range of targets/mechanisms that available analgesics act on, resulting in pharmacoresistance – where administered analgesic drugs are commonly ineffective in patients – current FDA approved drugs typically only are clinically effective (>50% reduction in pain) in a minority of chronic pain patients. A useful metric of this is the number needed to treat (NNT) which defines how many patients need to be given a particular treatment to see a clinically meaningful reduction in pain. Ideally, this should be one, but usually is from 5 to 20 for chronic pain conditions(98, 99). Apart from low efficacy across the pain spectrum most available analgesics also have a low persistence of activity when used chronically including both opioids and NSAIDS(100, 101).

Targeting nociceptors for cell selectivity

Like all neurons, nociceptors express dozens of different kinds of sodium, potassium and calcium ion channels that together regulate their excitability. However, the expression levels of particular ion channels in primary mouse nociceptors are substantially different from those in other kinds of neurons, even when compared with other primary sensory neurons(102). The unique pattern of expression of ion channels in nociceptors results in distinct intrinsic electrical properties compared to other neurons. It should be possible to exploit these differences to differentially inhibit/modulate the excitability of nociceptors, with minimal inhibition of other kinds of neurons. So far, such efforts have largely focused on a “magic bullet” approach of developing highly selective blockers for either Nav1.7 and Nav1.8 channels, which are predominantly expressed in nociceptors, but these efforts have been disappointing, so far. It may well be that selectively targeting any single ion channel will only have limited efficacy given that excitability is driven by activity in multiple channels and the capacity for compensation in response to inhibition of a single channel. It is noteworthy that almost all ion channel-targeted drugs currently effective in clinical practice are “dirty” in the sense that they inhibit many different ion channels; for example, carbamazepine and phenytoin used to control hyperexcitability in epilepsy not only target almost all subtypes of voltage-gated sodium channels with similar efficacy but also have effects on many kinds of calcium and potassium channels(103). The efficacy of these drugs was established without knowledge of their full target profile, but we now realize that the overall electrical activity of neurons has a complex non-linear dependence on the dozens of different ion channel types they express, present in different combinations and proportions in different kinds of neurons. With this knowledge, we can set out to analyze and model the overall excitability of different neurons with the goal of more rationally identifying that particular combination of targets required for optimal differential cell-selective effects. Then, it should be possible to deliberately screen for compounds with actions only on the identified multiple specific targets - a “magic shotgun” approach, similar to that proposed for improved GPCR-targeted drugs for mood disorders(104).

An additional consideration is that many ion channels have distinctive patterns of expression in different regions of nociceptors – peripheral axon terminals, axons in the nerve or dorsal root, and central axon terminals – than in the cell bodies in the DRG where electrical properties are commonly studied. Our knowledge of the expression of individual ion channels in nerve endings, axons, and central terminals, is unfortunately very limited, because even substantial channel densities are difficult to detect using immunocytochemistry. The most detailed knowledge of regional expression, in the case of sodium channels, is currently based on the differential sensitivity of transduction, action potential propagation and transmission to local TTX administration, revealing that specific populations of nociceptors differ widely in the extent to which TTX, or more specific sodium channel blockers, either inhibit transduction at terminals or the propagation of action potentials. This has led to the conclusion that because Nav1.7 and Nav1.8 channels are differentially important for the generation and the propagation of action potentials respectively, in various nociceptors “a dual-targeting strategy of peripheral Nav1.7 and Nav1.8 inhibition would be more effective in providing analgesia” than either alone(105). While this could be done by combining selective Nav1.7 and Nav1.8 inhibitors, an alternative approach would be to find single compounds with such dual-target action, and with no or minimal effects on other neuronal, cardiac, or skeletal muscle sodium channels, or even compounds with a broader profile on more ion channels.

Identification of the cell-region specific expression of all ion channels in nociceptors will be really valuable, particularly for the many types of potassium channels important for regulating the excitability of nociceptors. There are compounds that can both inhibit sodium channels and enhance several types of potassium channels, like riluzole, used for treating ALS(106, 107), so it is not far-fetched to search for and generate single agents that are targeted specifically at multiple sodium, potassium or calcium channels important for different aspects of nociceptor function, and the changes in these that occur during inflammation or after nerve lesions. Potassium channels with substantial expression in nociceptors include Kv3.4, BK, Kv7 and TREK channels(102, 108), and Kv7 enhancing compounds are effective in multiple pain models(108, 109). In general, the pharmacology of potassium channel enhancement is an underdeveloped area, but one of future promise for targeting neurons in a cell-selective way, especially because of the widely different patterns of potassium channel expression in different kinds of neurons. Overall, a polypharmacology approach towards acting on that set of ion channels that determine nociceptor activity in health or disease, should both increase efficacy over single target-selective drugs, and because of cell selectivity, reduce the adverse effects inherent in non-selective drugs. There are two ways potentially to do this, either identify compounds with the desired activity on the identified set of targets by parallel screening of each channel separately, or screen directly for compounds (single-agent or in a combinatorial fashion) that have the desired excitability modulating action on nociceptors but not on other excitable cells, using iPSC-derived neuronal phenotypic screens.

New approaches to analgesic development can increasingly be based on the optimal utilization of human stem cell-derived neurons for disease modeling, cell type and disease state-specific screening, and identification of patient susceptibility to particular disease states, like neuropathy, or responsiveness to particular treatments(110–117). Three therapeutic approaches may be useful to explore. First, ways to block inputs to nociceptors need to be identified, to enable reducing either the activation or the sensitization of nociceptors in health, inflammatory conditions or after nerve lesions. Second, there is a need to block action potential transmission selectively only in nociceptors, using drugs that silence these neurons with no or minimal activity on other cells, to block pain-triggering signals entering into the CNS. The strategy of using permanently charged sodium channel blockers that only enter neurons through activated large pore ion channels, which are primarily only expressed by nociceptors, generates long-duration selective silencing of nociceptors, but can only be administered topically(118, 119). Third, tools are needed to selectively block activity only in nociceptive circuits in the CNS, by either reducing excitation or increasing inhibition and preventing/reversing maladaptive plasticity only in these circuits in the spinal cord or brain(15, 120–122). However, given the similarity of nociceptive circuits to many others in the CNS, it is likely the third approach will be the most challenging, and we will therefore focus here only on the first two approaches, reducing the activation/sensitization of nociceptors and selectively silencing them.

Developing nociceptor specific therapies - Identifying targets

We propose two complementary approaches to identify novel pain targets. 1) Use differential gene and protein expression analyses to identify targets selectively expressed in nociceptors or those cells (e.g. immune cells or glia) linked to increased excitability in sensory neurons, and 2) use phenotypic screens of human iPSC-derived excitable cells against annotated libraries of compounds whose mechanism of action and targets are known, to identify those that selectively inhibit nociceptor excitability while sparing other excitable cells – such hits will reveal targets driving nociceptor-specific function (Figure 2a).

Figure 2:

Fig. 2a: Two complementary approaches to identify novel analgesic targets. Human iPSC-derived nociceptors can be used to identify novel analgesic targets. In one approach, differential gene expression in nociceptors is identified. In the second, an annotated library of compounds, whose primary targets of action are known, are screened for decreases in activity on multiple cell types. The annotated targets of compounds that selectively only inhibit nociceptor activity are potential pain targets. Fig. 2b: Identifying and validating candidate nociceptor-selective targets and pathways. Genes selectively expressed in iPSC-derived nociceptors can be identified (A). A similar comparison of primary mouse and human tissues will yield sets of genes selectively expressed in nociceptors for each species. These can be assembled into signaling pathways (B) predicted to affect nociceptor excitability selectively (C) and represent potential translatable target sets. Nociceptor-selective targets can be validated using small molecules that selectively modulate the function of targets or through CRISPR/Cas9 editing (D). The process can be repeated in the presence of inflammatory mediators, chemotherapeutic agents, or after axon injury to discover selective targets for each condition. Reporters of intracellular signaling can be used to garner information on pathways activated by disease perturbations.

The first approach will yield a list of gene transcripts and proteins selectively expressed in nociceptors and not in other neurons or excitable cells – using both human tissue samples and iPSC-derived cells (Figure 2b). Nociceptor selective proteins can then be assembled into causal networks that reveal signaling pathways that modulate excitability(123). From this approach, the best combination of network nodes predicted to diminish or block nociceptor activity can be determined and tested to establish if there is selective modulation of nociceptor excitability. Targeting selective pathways instead of a single protein could provide a polypharmacology approach to modulating nociceptor excitability(124–127).

Single-cell/nuclei RNA-seq enables the determination of the repertoire of genes expressed in nociceptors under physiological conditions, in response to disease-based manipulations, or as a result of the action of a drug. While some attempts have been made to identify targets enriched in primary human DRGs(128) relative to those expressed in other human cell types(129), single-cell profiling analyses are required to define exactly the human nociceptor transcript profile. Identifying nociceptor targets contributing to inflammatory and neuropathic pain will require access to DRG from patients with these conditions, which is a major challenge, especially for getting a temporal profile of induced changes. Data generated from human iPSC-derived neurons may provide an alternative means to identify inflammatory and neuropathic pain targets in a human “disease-in-a-dish” model system by detecting changes in gene expression under various pain-related conditions. A comparison of phenotypes and transcription profiles in nociceptors from individual patients in response to disease-related challenges may also assist in the identification of the drivers of disease susceptibility. Why do some patients with diabetes or on exposure to chemotherapy agents develop painful neuropathy and others do not? Phosphoproteomics can add an annotation to pathways of interest and help identify which nociceptor selective pathways actively impact excitability(130). While mouse DRG single-cell gene expression datasets may have some utility, there may be species differences both in cell identity and the response to inflammation/injury(102, 131–133). Comparing primary human and mouse nociceptor selective genes with iPSC-derived nociceptor selective genes will help determine how far iPSC-derived cells recapitulate pain-triggering pathways.

A phenotypic screen of bioactive compounds that yield hits that selectively modulate nociceptor excitability will enable deconvolution of a set of targets can then be and then tested in a primary mouse or human nociceptors (Figure 3).

Figure 3: Phenotypic screen-based discovery platform.

Schematic of discovery platform for novel pain targets using omics-based nociceptor selective targets and phenotypic screen-based compounds that decrease nociceptor excitability. The screen can be modified to discover targets that prevent peripheral sensitization by conducting phenotypic screens after nociceptors have been sensitized with inflammatory mediators. Validated nociceptor selective targets can be screened against diversity chemical libraries to identify scaffolds that modulate the activity of desired sets of targets. Structural and drug-like properties of these compounds can be enhanced and diversified using machine language assisted algorithms. New molecules designed from lead scaffolds can be tested for selective activity on nociceptor excitability by phenotypic screens and patch-clamp electrophysiology. Promising compounds with desired target modulation profile can be tested on preclinical animal models for efficacy, ADMET properties and safety.

Compounds that completely and selectively block action potential generation only in nociceptors will reveal targets for managing nociceptive pain. Compounds that decrease action potential firing only when nociceptors are treated with inflammatory mediators will reveal targets useful in managing inflammatory pain. Targets for neuropathic conditions can be identified by identifying compounds that diminish nociceptor excitability on laser cutting of axons, treatment with chemotherapeutic agents or other forms of axon injury. Genome-wide genetic editing screens in iPSC-derived neurons, could provide a more direct way to identify targets, providing the false positive and negative rate is small enough, and the genetic mutations do not affect the iPSC function or the differentiation into nociceptors. CRISPR/Cas9 is the method of choice for site-specific, inducible and reversible genetic manipulation(113, 134–139).

Once suitable targets are identified, their activity profile needs to be verified, either by using available small molecules selective for these targets or CRISPR/Cas9 gene KO or siRNA knockdown approaches, looking for nociceptor selectivity and disease normalization, with no activity in counter screens on other neurons/cells. Following such validation, heterologous expression systems can then be used to identify chemical entities that selectively modulate the function of the validated target in HEK or CHO cell lines. Large chemical diversity libraries can be screened against selected ion channel, GPCR or kinase targets to identify chemical scaffolds as starting material for medicinal chemistry campaigns for structure-activity relationship studies and ensuring drug-like properties. Target-based screening has the technical advantage that the screening assay is relatively straightforward and the signal typically robust (high Z-prime), a limitation though is that the activity of a target in a non-neuronal heterologous expression system may not match that in nociceptors because of a lack of post-translational processing or absence of interacting proteins that create a particular molecular architecture in native cells.

In vivo preclinical validation of putative targets is critical, if possible – if there is no species difference in target expression. Rodent pain models that are used for this need though to be accurate surrogates of human pain conditions and their response to specific interventions and measurements need to unbiased, specific and sensitive – something now possible using machine learning.

Using human genetic studies as the primary identifier of potential analgesic targets – where a loss of function mutation causes a clear pain phenotype(105), target-based screening has to date largely focused on developing single target-specific compounds for treating pain, e.g. Nav1.7 blockers(140). However, there is a high failure rate of target-based drug discovery in clinical trials, mainly due to lack of efficacy(141). Despite major efforts by multiple companies, Nav1.7 blockers have repeatedly failed in the clinic(142). It may be time to consider alternative strategies.

Cell and disease-based phenotypic screening using iPSC derived nociceptors

Unbiased phenotypic screens can be designed to identify compounds that act selectively only on a particular cell (e.g. silence only nociceptors) or change a specific disease-related phenotype (reduce inflammatory sensitization or neuropathic ectopic firing in nociceptors), something much more complex than interacting with a single target Figures 2 & 3). Nevertheless, recent advances in iPSC technology (Figure 4) now enable the design of disease-related phenotypic assays in physiologically relevant systems using human/patient cells(143). In addition, a hit in such assays may act on multiple targets and signal pathways, and such polypharmacology might well generate better efficacy in vivo than a compound that acts exclusively on one target.

Figure 4: Human iPSC technology in target discovery and translational applications.

Patient-derived somatic cells (e.g. skin or blood cells) can be reprogrammed into iPSCs using four defined transcription factors (OCT4, SOX2, KLF4, C-MYC). Once pluripotent stem cell lines are established, they can be expanded indefinitely and differentiated into nociceptors and other cell types of the human body relevant to pain (e.g. astrocytes, microglia and macrophages) enabling a broad range of applications for translational pain research including disease modeling, identification of novel targets, drug screening, cell therapy, gene therapy.

Currently there are several different types of image-based assays that can be used to identify compounds that act on nociceptors in high-content screens. The activity of the nociceptors can be measured using genetically encoded calcium indicators (GECIs) or voltage indicators (GEVIs). The primary advantage of measuring calcium signals as a surrogate for electrical activity, for example with GCaMP, is the strong magnitude of the signal, while a limitation is the slow kinetics of calcium entry(144). Furthermore, nociceptors are adapted to respond only to specific activation of their peripheral terminals by noxious stimuli, and therefore have low levels of spontaneous activity, which means that it is difficult to directly screen for nociceptor silencing compounds when there is little basal activity to inhibit. One way around this is to increase activity in the nociceptors, by for example enhancing sodium currents with veratridine(145), or by reducing potassium flux. Another more physiological way is incubating the iPSC-derived nociceptors with inflammatory mediators which sensitize TRPV1 channels, increase the current density of Nav1.8 and the trafficking of Nav1.7 channels to the membrane, mimicking inflammation-induced peripheral sensitization(146). Screening for a reduction of such inflammatory mediator triggered activity will identify compounds that either silence the nociceptors (e.g. by blocking action potential firing) or reduce the peripheral sensitization (e.g. by inhibiting kinases activated by the inflammatory mediators), and secondary screens are required to identify which specific action the compounds have. Alternatives to genetically encoded reporters are automated patch clamp and multi-well multielectrode array (MEA) recordings, which have though, a lower throughput screening capability and higher cost, but the direct measurement of membrane current or voltage makes them a useful option for secondary validation assays. Genetically encoded pH indicators (GEPIs) can identify compounds that block transmitter vesicle release from nociceptors. As GEPIs do not reveal the dynamics of specific neurotransmitters, genetically encoded transmitter indicators (GETIs) at the presynaptic or postsynaptic site can be used if specific release mechanisms (e.g. neuropeptide release) are being targeted(144). Automated neurite imaging systems in multi-well assay format can pick up compounds that rescue chemotherapy-induced axon degeneration or promote the regeneration of axons after axon injury(147). Finally, spatially multiplexed imaging of signaling reporter islands (SiRIs) as indicators of second messengers (Calcium and cCAMP) and kinases (PKA, PKC, pERK) in neuronal culture can unveil which signaling pathways drive changes in excitability. Such assays can reveal the signaling pathway of the hit compound and the dynamic changes caused by stimuli that cause sensitization/ectopic activity in nociceptors(148, 149).

A crucial element of phenotypic screening strategies is to run parallel counter screens. If the aim is to find compounds that act selectively on nociceptors, for example, then a clear demonstration of a lack of activity on other excitable cells is required, if the aim is to reduce spontaneous firing that drives neuropathic pain there must be no effect on non-injured nociceptors, if the goal is to develop a drug that reduces inflammatory pain, then disruption of immune cell interaction with nociceptors is required.

A direct comparison of target-based versus phenotypic strategies reveals that phenotypic screens are more successful in developing first-in-class drugs(150). Screening chemical libraries in miniaturized 384- and 1536-well plates and using different compound concentrations can establish valuable dose-response curves for thousands of small molecules and the use of multiple replicates can help minimize the number of false-positive and false-negative hits(151, 152). Similarly, taking advantage of innovative screening strategies, such as fast and precise low-volume liquid dispensing using acoustic sound energy, allows for the flexible testing of multiple small molecules in combinatorial drug screens(151, 153).

As an example of the promise of phenotypic screens Studer and colleagues(154) screened iPSC-derived neural crest precursors from patients with familial dysautonomia against 6,912 small molecule compounds and identified compounds that induce expression of IKBKAP, the gene affected in this rare disease, that among other symptoms leads to insensitivity to pain. While using undifferentiated cells might provide insights into particular diseases, the ultimate goal for drug screening is to use terminally differentiated neural cell types that are functional and express the relevant targets or phenotypes of interest. However, iPSC technology enables screening at different stages of cell differentiation or disease state, and it might be possible to elucidate key temporal features of pathological changes and perform chemical screens accordingly. For instance, precisely characterizing the transition from acute to chronic pain could lead to treatment options that prevent or reverse the development or maintenance of cellular pathology and aberrant cellular plasticity. While the standardized use of human iPSC-derived neural cell types generated by stepwise differentiation for high-throughput screening is still in its infancy, more efficient and scalable cell differentiation protocols should help overcome the current technical challenges.

Disease modeling using iPSC-derived neurons

Disease modeling and identification of underlying cellular and molecular disease mechanisms is an important and novel opportunity enabled by iPSC technology, constituting “disease-in-a-dish” or “patient-in-a-dish” for probing relevant phenotypes in cell types under precisely controlled laboratory conditions (Figure 4).

Establishing genotype-phenotype relationships and charting disease signatures that accurately reflect disease phenotypes represents a key strategy for translating basic research findings into clinical applications, especially in terms of phenotypic screening and target identification(150, 155). Genome engineering using CRISPR/Cas9 allows site-specific and precise manipulation of normal and disease alleles and vastly expands the translational toolbox when combined with iPSCs(134, 156). Of special interest for modeling monogenic diseases, and for target identification and validation, is the utilization of isogenic iPSC lines which are genetically matched with the parental cell line after gene editing. Use of isogenic iPSC lines increases confidence in that accurate comparisons between normal and diseased cellular phenotypes can be achieved, and can reduce technical and biological variability as potential experimental confounders(134, 157, 158).

Genome-wide association studies (GWAS) and study of rare familial disorders has provided new insight into the genetics and epigenetics of acute and chronic pain including susceptibility, interindividual differences, and has identified potential new targets(159–163). For instance, gain-of-function mutations of NaV1.7 lead to erythromelalgia, paroxysmal extreme pain disorder, and small fiber neuropathy(164, 165). By investigating iPSC-derived nociceptors of a family with the same NaV1.7 mutation (S241T) but different pain severity (mother with moderate pain and son with severe pain), Waxman and colleagues could correlate in vitro excitability of nociceptors with the clinical presentation. Interestingly, further analysis using whole-exome sequencing identified a variant in the potassium channel KCNQ2 that conferred the resilience to pain in the mother(166). This study shows that interindividual differences in pain phenotypes and the underlying molecular mechanism(s) can be identified using patient-derived nociceptors. Also taking advantage of iPSC-derived nociceptors from a patient with small fiber neuropathy, in vitro drug testing was able to predict that the anticonvulsant lacosamide would alleviate pain in the patient(114). Ex vivo testing of patient-derived cells can serve therefore, as a precision medicine approach to develop personalized treatments. However, while iPSC derived nociceptors preserve features on Nav1.7 gain of function mutations – such as enhanced activity and temperature dependence, such mutations in mice seem not to, presenting with no pain phenotype(167), a suggestion that mice may not be useful for modeling at least some pain conditions.

iPSC-derived cell types can be exploited for disease modeling in response to neurotoxicants, chemotherapeutic drugs, metabolites, inflammatory signals and pathogens. Chemotherapy-induced polyneuropathy (CIPN) is a common side effect of cancer treatment, including numbness and loss of proprioception as well as hyperalgesia or allodynia(168). To model CIPN, patient-derived iPSC neurons can be treated with paclitaxel, cisplatin, and vincristine(169, 170). Such studies provide invaluable information on mechanisms of neuronal damage and interindividual susceptibility differences and offer an opportunity for finding new therapeutic options.

iPSC derived neurons

Human pluripotent stem cells (hPSCs) are defined by their unlimited self-renewal capacity in the absence of senescence (i.e. high telomerase activity maintains telomere length and stability) and the potential to differentiate into all somatic cell types(171). Given their human origin and the ability to model disease using patient-derived disease-specific iPSC lines, these lines offer the advantage of providing the cell numbers required for phenotypic screens that are otherwise unattainable, together with a promise of personalized therapy. The two major types of hPSCs comprise embryonic stem cells (ESCs) and iPSCs. The first human ESC lines were derived from the inner cell mass of blastocyst-stage embryos over 20 years ago(172). However, because ESCs are directly derived from preimplantation human embryos, establishing new cell lines poses ethical challenges and limits the streamlined study of human genetics and diseases. In contrast, Yamanaka and colleagues(173, 174) established the iPSC lines by reprogramming skin fibroblasts to a pluripotent state by transient expression of the transcription factors OCT4, SOX2, KLF4, C-MYC.

Once cells are successfully reprogrammed to iPSCs, endogenous developmental genes (including the four transcription factors mentioned above) are activated and maintain the gene regulatory network of pluripotency. These cell lines can be created not only from skin fibroblasts but also from other somatic cells (e.g. blood cells, keratinocytes) of an healthy individual or patient, irrespective of age, genetic background, and disease type (e.g. common or rare, monogenic or polygenic). Importantly, iPSCs are highly similar to ESCs with regard to morphology, gene expression, self-renewal capacity, and functional differentiation into multiple cell types representing various developmental lineages including ectoderm (e.g. CNS, neural crest, skin), mesoderm (e.g. blood, bone, muscle), and endoderm (e.g. lung, gut, liver, pancreas).

While the first iPSC lines were established using retroviral and lentiviral methods, more advanced reprogramming techniques allow the generation of cell lines using non-integrating viruses or even virus-free methods (episomal DNAs, synthetic mRNAs, Sendai virus) thereby increasing safety and avoiding the risk of random genomic insertions and mutagenesis(175, 176). Neurons derived from human iPSCs are now widely used to model neurodegenerative, neuropsychiatric and neurodevelopmental diseases(177, 178).

Differentiating iPSCs into relevant cell types

Converting iPSCs into specialized functional cell types entails stepwise differentiation following the principles of developmental biology. By modulating cell signaling pathways with small molecules and recombinant proteins, the pluripotent cells are coaxed into lineage-committed precursor cells and then further patterned and differentiated into more specialized cell types that demonstrate gene expression signatures and functional properties similar to their in vivo counterpart (e.g. expression of marker proteins, cell-type-specific transcriptomes, intracellular signaling, response to natural ligands, electrophysiological properties). However, because most of our developmental biology knowledge is derived from animal models, not all available information on pathways controlling cell differentiation can be directly applied to human cells due to significant species-specific differences(128, 179–181). Specific protocols need, therefore, to be carefully developed and optimized for human iPSCs followed by comprehensive functional and molecular characterization of differentiated cell types to assess validity, maturation state, and usefulness for translational applications. Currently, a major challenge for utilizing iPSC technology for translational research and therapeutic development is the lack of highly efficient, reproducible, and scalable cell differentiation protocols. Other hurdles for iPSC application are cell-line-to-cell-line variability and technical variation of cell culture conditions that strongly influence the differentiation process(157, 182, 183). Nevertheless, progress has been made over the last decade in improving cell differentiation protocols. For instance, iPSCs can be converted into neural crest stem cells expressing the transcription factor SOX10, which then can be further differentiated into nociceptors by simultaneously manipulating specific cell signaling pathways. These iPSC-derived nociceptors express typical transcription factors (e.g. BRN3A, ISL1, PRDM12), neuropeptides (e.g. Substance P, CGRP), transmembrane receptors (e.g. TRPV1, TRKA, TRKB, P2RX3), sodium channels (e.g. Na_V1.7, 1.8, 1.9) and display cell-type-specific features of peripheral sensory neurons including response to specific ligands such as capsaicin, menthol, mustard oil, and ATP(113, 166, 179, 184–187). Besides generation of nociceptors, various other cell types relevant for pain research have been derived from iPSCs, including spinal cord neurons(188), Schwann cells(189, 190), microglia(191), astrocytes(192–194), GABAergic neurons(195), and cortical neurons(196).

In summary, iPSCs represent an inexhaustible source of human cells that allow studying of the cellular diversity and disease state of the PNS and CNS for translational pain research, and the utilization of pain-specific phenotypic assays for drug screens. As an alternative to iPSCs, it is also possible to directly convert fibroblasts into nociceptors by forced expression of specific transcription factors(117). However, such transdifferentiation protocols are inefficient and not scalable for high-throughput screening.

Organoids and complex tissue models

Under two-dimensional (2D) cell culture conditions, cells grow attached to flat cell culture plates and therefore do not fully recapitulate in vivo physiology. In contrast, 3D models such as organoids or mini-organs more closely model dynamic in vivo cell function and behavior, including cell-cell interactions and biomechanical forces within the appropriate extracellular matrix(197). The use of 3D models can also help enhance cell maturation, a common challenge for iPSC-derived cell types. Organoid models with relevance for pain research have been developed for skin(198), DRG(199) and thalamus(200). The study of complex tissues may provide future opportunities to interrogate nociceptive circuits and test neuroactive drugs. However, the reproducibility and consistency of randomly self-organizing organoid models need to be improved before they can be used as robust assays compatible with drug discovery. Other innovative technologies being developed are based on advances in bioengineering and biomaterials and could lead to standardized microphysiological assays, organs-on-chips, and bioprinted tissues(201, 202). These complex systems are also being developed as platforms for ADME (absorption, distribution, metabolism and excretion) of small molecules, which may provide human pharmacokinetic (PK) predictions and potentially eventually replace animal experimentation. Another model is the “nerve-on-a-chip”, which takes advantage of culturing Schwann cells with human motor or sensory neurons to establish long axonal projections for measuring electrophysiological properties, drug effects, and response to injury(203).

Scalable biomanufacturing of human iPSC-derived cells

Standardized and reproducible production of large quantities of human cells is required for genetic and chemical screening projects. Manual culture and differentiation of iPSCs are variable and labor-intensive, which limits their utilization in industrial-scale screening projects and high-throughput applications. Recent progress has been made in generating and culturing iPSCs using bioreactors, automated liquid handlers, and robotic cell culture systems(204–206). For instance, it is possible to automate all essential steps of iPSC culture and directed differentiation into neurons, cardiomyocytes, and hepatocytes(206). This approach allows culture of up to 90 iPSC lines in parallel, and that may help overcome the problem that investigation of only a few cell lines per study may not be sufficient for drawing general population-wide conclusions. Robotic cell culture supports the standardized manufacturing of billions of human cells, which can be cryopreserved and then used as needed. Scalable cell culture is compatible with batch production as well as continuous biomanufacturing. Overall, automation not only increases scientific rigor by reducing the investigator bias inherent in the manual iPSC culture, but also provides access to cellular material for functional genomics and high-throughput/high-content screening of chemical libraries, natural products, and biologics.

Cell therapy

One of the great promises of iPSC technology is the potential to develop personalized cell/regenerative medicine therapies (Figure 4). Autologous non-immunogenic cell therapies are of particular interest because, unlike allogeneic approaches, they do not require long-term immunosuppression or donors with a matching human leucocyte antigen (HLA) signature. Although various cell products generated from human ESCs and iPSCs are currently being tested in clinical trials(156), none of the current clinical trials target chronic pain. However, cell therapy may become a therapeutic option for certain pains, since human iPSC-derived GABAergic neurons injected into the spinal cord of mice with chronic neuropathic pain reduces the pain phenotype for up to two months post-transplantation(207). Earlier reports had documented that grafting of GABAergic precursors isolated from fetal rodent brains or human ESC-derived GABAergic neurons mitigates pain after spinal cord or peripheral injury by overcoming disinhibition(208, 209). These studies are consistent with the observation that intrathecal application of the GABA agonist baclofen provides analgesia in chronic pain patients(210). Central neuropathic pain may also benefit from grafting neuronal cells with other inhibitory neurotransmitter phenotypes (e.g. glycine), oligodendrocytes(211) and astrocytes(193, 194) but more research is needed to define the mechanisms of action and the optimal time window for cell therapeutic intervention. iPSC-based cell therapies may prove useful for the treatment of peripheral neuropathy and neuropathic pain, since the complex interplay of multiple cell types including sensory neurons, Schwann cells, satellite glia, and immune cells contribute to axon loss, demyelination, ectopic firing and pathological pain(212). The grafting of undifferentiated neural crest stem cells supports axonal regrowth and myelination after sciatic nerve injury(213, 214). Although progress has been made in differentiating iPSC-derived neural crest stem cells into Schwann cells(189, 190, 215), more efficient and clinically relevant protocols are needed to generate large quantities of cells with myelination capacity. Satellite glia are specialized support cells that closely envelop the cell bodies of sensory neurons in the DRG and are activated during inflammation and nerve injury(216). Despite their unique anatomical location and potential importance for sensory neuron function, human satellite glia are largely understudied, and derivation of satellite glia from iPSCs and their detailed characterization should change this. There is also increasing interest in immune cells such as microglia and DRG-resident macrophages, which contribute to the induction and maintenance of chronic pain(217, 218). Derivation of microglia-like cells from human iPSCs(191) should enable a better understanding of their basic biology, immunomodulatory function, and therapeutic potential. Altogether, production of all disease-relevant cell types from iPSCs (neuronal and non-neuronal) would allow systematic neurotransplantation experiments and testing of the utility of replacement of these cells in animal models of central and peripheral neuropathic pain. However, all stem cell-derived products will require extensive safety measures to ensure that grafted cells are karyotypically normal and properly differentiated otherwise in vivo tumor formation may occur. Indeed, the use of undifferentiated or proliferative cells is particularly risky as they can continue to grow upon transplantation and form teratomas or other tumors(219).

Gene Therapy

Newly emerging insights into genetic mechanisms and cellular pathologies provide opportunities to devise efficacious and safe therapies for pain by directly targeting DNA or RNA. The rationale for gene therapy includes replacing or correcting a disease-causing gene/mutation in a specific cell type, overexpression of neurotransmitters or ligands to modulate specific receptors, expression of transcription factors to drive change in function, neuroprotection, change of cellular identity or recruitment of endogenous cells for repair or replacement. Relevant technologies for gene delivery use replication-incompetent viruses such as herpes simplex viral vector (HSV) and adeno-associated viral vector (AAV), which have been used in the context of pain(220). Other available gene therapy strategies are RNA interference (RNAi) and antisense oligonucleotide (ASO) technology that can degrade or inhibit the processing of specific mRNA molecules, thereby modulating or silencing the expression of a protein of interest(221–223). Application of CRISPR/Cas9 methods can correct a genetic defect, overexpress genes, or fine-tune gene expression via inducible systems that increase the level of control and avoid/reverse unwanted effects. More research is warranted though, to better understand off-target effects, neurotoxicity, and long-term consequences(224). Specifically, the use of iPSC-derived cell types in the context of gene therapy could become a reality in two scenarios: first, gene therapy therapeutic strategies could be directly tested in relevant human cell types expressing the target gene, as part of a predictive in vitro platform (Figure 4). For instance, modulating the expression of specific ion channels in human nociceptors and characterizing the resulting functional consequences would provide valuable information before moving into clinical application. Second, the combined use of gene and cell therapies would expand the therapeutic armamentarium of personalized medicine. In theory, any iPSC-derived somatic cell type and any gene could be manipulated in vitro and then given back to a patient donor. Using chemogenetic or optogenetic cell engineering approaches, only the genetically modified cells could be selectively activated or inhibited in the patient. Neuromodulation could also be achieved using prodrugs that become pharmacologically active after being metabolized in vivo or on photostimulation(225–228).

Perspectives and translational opportunities

To develop effective analgesic drugs with no abuse liability it is important to acknowledge pain as a major and diverse systems-level challenge and identify optimal strategies across the preclinical and clinical spectrum. The wealth of new bioengineering technologies, multi-omics platforms, bioinformatics and machine learning and artificial intelligence will certainly aid in generating and integrating information derived from genomics, iPSC-derived human cells, and animal models, with a better representation of human disease, and therefore a better predictor of outcome in clinical studies. As a major new frontier in therapeutics development, iPSC technology can bridge the gap between genomic data and personalized medicine by enabling disease modeling and phenotype characterization, new target identification, drug screening and experimental validation (Figure 4). The development of such therapeutic strategies requires running “clinical trials in a dish” and using iPSC lines from large patient cohorts (including different genders, ethnic backgrounds, disease histories) to identify novel therapeutics and patient responders and non-responders. Mechanism-based scientific findings using human cellular models can, we argue, be efficiently translated into clinical applications that improve the management of chronic pain and avoid adverse effects, including dependence and addiction.

Crucial to translating preclinical findings to clinical use is to understand exactly which conclusions drawn from preclinical studies apply to clinical settings. Until recently preclinical studies have largely used rodent models, even though important species differences exist, and pain cannot be directly measured in animals. The introduction of human iPSC-derived models provides an opportunity to begin to overcome these difficulties, although challenges remain in establishing which preclinical parameters positively correlate with clinical efficacy. Phenotypic screens allow for the determination of the specific action of compounds on particular cells and defined disease phenotypes, as well exploitation of polypharmacology opportunities. Counter screens provide a means of removing compounds with non-specific/undesired effects early in the discovery process. Overall, prospects utilizing iPSC-derived neurons are now higher for the development of effective nociceptor selective silencers suitable for managing surgical or post-traumatic pain, and eliminators of either the immune-mediated sensitization that drives inflammatory pain hypersensitivity or of the development of ectopic activity in injured nociceptors that triggers neuropathic pain.