Discussion

In our study, we developed an automated, unbiased method to analyze opioid-induced MOR internalization (Fig. 1). We studied morphine, an opioid known to poorly induce MOR internalization (Abbadie and Pasternak, 2001), and fentanyl, an opioid that robustly induces MOR internalization (Koch et al., 2005). We demonstrated that morphine tolerance was completely inhibited by imatinib, a PDGFR inhibitor, and that imatinib did not alter the degree of MOR internalization caused by morphine after acute or chronic morphine administration (Fig. 3). We also showed that analgesic tolerance to fentanyl was completely eliminated by imatinib and that imatinib did not alter the degree of MOR internalization induced by fentanyl (Fig. 4). We also found that imatinib blocked the development of tolerance to three other clinically used opioids with varying pharmacological and internalization properties (Drewes et al., 2013) (Fig. 4). Taken together, our results indicate that 1) the development of opioid tolerance in the spinal cord occurs independently of changes in MOR internalization, 2) tolerance can be blocked without altering MOR internalization, and 3) PDGFR signaling mediates tolerance development for a wide range of clinically used opioids. However, our observations do not preclude the possibility that MOR internalization could play a role in the development of opioid tolerance in higher brain regions (Cox and Crowder, 2004; Williams et al., 2013). It is unclear why chronic administration of agonists known to induce internalization acutely did not significantly increase internalization in our study. There are two possibilities: First, we may have missed the peak time of internalization in the chronic study, because of either the delay induced by behavioral testing or the possibility that the kinetics of the internalization response could be altered by chronic drug administration. Second, it is also possible that chronic administration of DAMGO or fentanyl could decrease the internalization response. However, there is not clear data in support of this hypothesis (Koch and Hollt, 2008; Anselmi et al., 2013).

Studies investigating MOR internalization are painstaking and require a tremendous amount of personal effort. Automated methods have been recommended to improve reproducibility in preclinical science (McNutt, 2014; Begley and Ioannidis, 2015). A previous study developed an approach called quantitative automated microscopy to evaluate RTK-mediated signaling in pain sensitization (Andres et al., 2010). Although their analytical approach was somewhat different than ours, they identified two key concepts: namely, that the “eye-based” methods of identifying cells positive or negative for a marker (in our case, immunofluorescence) may not be accurate and that it is critical to define threshold criterion to distinguish signal from background (Andres et al., 2010). These concepts suggest that any manual evaluation of marker intensity could be suspect but do not provide a direct explanation for the contradictory findings of previous studies. Another possible confound is that, given the heterogeneity of cells in living tissue, large sample sizes will likely be needed to detect statistically significant changes in a population. Our automated analyses were based on sample sizes from approximately 250 to greater than 750 neurons per treatment group (Table 1). Previous important studies of the effects of internalization upon opioid tolerance evaluated 10-fold fewer cells, with the majority studying between 10 and 20 cells per treatment group (Trafton et al., 2000; He et al., 2002; Trafton and Basbaum, 2004; Silva-Moreno et al., 2012).

The role MOR internalization plays in opioid tolerance is complicated and controversial. Classically, opioid tolerance was thought to be due to agonist-induced desensitization and internalization of the receptor. After agonist binding, G-protein coupled receptor kinases are rapidly recruited to and phosphorylate the MOR. This increases the affinity of the receptor for β-arrestins, which uncouple the receptors from G proteins, leading to desensitization and receptor internalization (Williams et al., 2013). However, it is clear that tolerance can be induced by opioid agonists with varying abilities to induce receptor internalization. For example, morphine can cause profound tolerance without inducing receptor internalization. It has been suggested that this could be due to preferential engagement of β-arrestin 2 by morphine (Eisinger et al., 2002), whereas drugs with high intrinsic efficacy that induce strong MOR internalization interact with both β-arrestin 1 and β-arrestin 2 (Groer et al., 2011). However, morphine has been shown to induce MOR internalization in adrenergic medullary neurons (Drake et al., 2005) and the dendrites but not the cell bodies of nucleus accumbens neurons (Haberstock-Debic et al., 2003, 2005), suggesting that morphine does have the ability to cause MOR internalization depending on the expression levels and/or subcellular localization of relevant signaling molecules (e.g., G-protein coupled receptor kinases, β-arrestins) present in specific regions.

The concept of biased agonism is an evolution of these findings. It has been postulated that different opioid agonists induce unique conformational changes in the receptor, leading to preferential interactions with different subsets of cellular signaling molecules (Pradhan et al., 2012; Ehrlich et al., 2019), which could preferentially mediate different opioid effects. Importantly, Schmid et al. (2017) synthesized a series of MOR agonists that exhibited marked functional selectivity for either G protein– or β-arrestin–mediated signaling. The investigators found that β-arrestin–preferring compounds were more likely to induce respiratory depression, whereas bias toward G protein signaling decreased the amount of respiratory depression relative to analgesia, markedly improving the therapeutic index.

An assumption implicit in the classic model of opioid tolerance is that receptor internalization leads to receptor degradation in lysosomes and loss of functional surface receptors. However, studies have shown that internalized MOR receptors mainly undergo a cycle of dephosphorylation, reactivation, and recycling of active receptors into the cell membrane (Law et al., 2000; Tanowitz and von Zastrow, 2003). This concept underlies the hypothesis introduced by Whistler and colleagues (He et al. 2002)that sustained MOR signaling, rather than termination of signaling, causes opioid tolerance. They posited that the tolerance liability of a MOR agonist would be its relative activity versus its ability to induce endocytosis (“RA/VE”) (Kieffer and Evans, 2002). In this formulation, an effective agonist that poorly induced internalization (such as morphine) would induce profound tolerance, whereas an agonist that induced robust internalization would have a lower tolerance liability. To support this hypothesis, He et al. (2002) presented data showing that administering a small, subanalgesic dose of the strongly internalizing MOR agonist DAMGO with morphine markedly increased MOR internalization in the spinal cord and decreased tolerance. However, other studies were unable to replicate these findings (Bailey et al., 2003), and additional work indicated that chronic morphine treatment caused decreased, rather than sustained, MOR signaling (Bailey et al., 2004; Dang and Williams, 2005). A core assumption of the RA/VE hypothesis is that endocytosis has a direct relationship to tolerance. Our work, and that of other investigators, suggests that robust tolerance can occur regardless of the receptor internalizing properties of an opioid agonist. For example, Narita et al. (2006) showed that the highly internalizing agonist etorphine induced a number of receptor trafficking proteins, whereas morphine did not. In contrast, morphine, but not etorphine, induced elevations of glial fibrillary acidic protein. Treatment with the glial inhibitor pentophylline blocked morphine, but not etorphine, tolerance. Conversely, intrathecal injection of astrocytes and conditioned medium accelerated morphine tolerance but did not impact tolerance to etorphine. A review by Cahill et al. (2016) nicely summarizes the roles of neural circuit adaptations in the genesis of opioid tolerance.

In sum, the literature suggests that depending on the neural context, multiple mechanisms may be responsible for the reduction in opioid efficacy observed as tolerance develops to different opioid agonists. Our findings suggest that although opioid tolerance in the spinal cord is not dependent upon MOR internalization, both prior theories of tolerance development could be compatible with our findings. Given that imatinib blocked tolerance to a variety of internalizing and noninternalizing opioid agonists, it appears that PDGFR-β signaling is engaged by MOR activation regardless of the ligand’s internalization properties or signaling bias. This supports the idea that PDGFR signaling is a central mediator underlying opioid analgesic tolerance. Additional experiments will be required to test this hypothesis.

It is important to note that chronic opioid administration induces sexually dimorphic responses. Observed differences can be inconsistent between studies and may depend upon the species or animal strain studied, drug dosing, route of administration, testing paradigm used, and hormonal state of the animals (Dahan et al., 2008). For example, systemic administration of morphine to rats resulted in more tolerance development in male than female rats (Loyd et al., 2008; Zhang et al., 2012). However, spinal administration of morphine to mice induced greater tolerance to a systemic challenge of morphine in female animals (Hopkins et al., 2004), whereas intracerebroventricular injection induced the same magnitude of tolerance in male and female mice (Kest and Hopkins, 2001). Gonadal hormones may also play an important role in tolerance development, as castration has been shown to eliminate sex differences in tolerance (Mousavi et al., 2007). An important direction for future studies will be to determine whether our observations that tolerance is blocked by PDGFR inhibitors and does not appear to require receptor internalization also apply to female animals.

The PDGFR-β is a member of the RTK family (Heldin and Lennartsson, 2013). GPCRs are known to transactivate RTKs to induce downstream signaling (Shah and Catt, 2004). The MOR has previously been shown to activate the RTK’s epidermal growth factor receptor (EGFR) and fibroblast growth factor receptor in vitro (Belcheva et al., 2001, 2002), and we previously demonstrated MOR activation of the PDGFR-β (Wang et al., 2012). It is also known that RTKs can directly feed back and affect GPCR signaling (Delcourt et al., 2007; Di Liberto et al., 2019). Previous work has shown direct effects of EGFR signaling upon the MOR (Clayton et al., 2010). These findings further support the possibility of selective RTK mediation of opioid effects.

The RTK family is quite large, and there are several RTKs that are closely related to the PDGFR (Lemmon and Schlessinger, 2010). This raises the question of whether tolerance is a selective consequence of PDGFR signaling or whether other related RTKs could also modulate tolerance. We recently reported that vascular endothelial growth factor inhibition blocked morphine tolerance (Lopez-Bellido et al., 2019). We have also observed that inhibition of EGFR signaling also blocks tolerance (Puig et al., 2020). Other investigators have shown that inhibition of signaling by the fibroblast growth factor receptor (Fujita-Hamabe et al., 2011) and the Ephrin receptor (Ephrin B) (Liu et al., 2011) also blocks morphine tolerance. Taken together, these results indicate that opioid tolerance can be caused by MOR signaling through several related RTKs. Additional studies will be needed to further define the generalizability of this phenomenon.

In conclusion, by utilizing an automated, unbiased analysis system, we have established that spinal opioid tolerance is not dependent upon MOR internalization. We have also shown that inhibition of PDGFR-β signaling blocks tolerance development in the spinal cord for a wide variety of opioid agonists with diverse properties, indicating that PDGFR-β signaling could be a central mechanism underlying opioid tolerance. These findings suggest the exciting possibility that opioid tolerance can be directly eliminated without altering opioid analgesia. This could potentially eliminate the dose escalation caused by tolerance, thereby dramatically increasing the safety and efficacy of opioid treatment.