In liver hepatocytes, cholesterol is metabolized via cytochrome P450 (CYP) to primary BAs (Fig. 1A). Primary BAs are subsequently metabolized to other BAs, after which they are conjugated to either glycine or taurine for more efficient biliary secretion. In the intestine, bacteria metabolize BAs to form secondary BAs, which can return to the bloodstream and liver through enterohepatic circulation (Fig. 1A). We recently identified MRGPRX4 as a bilirubin receptor that may contribute to cholestatic pruritus (27). We performed a counter screen of MRGPRX4 and Mrgpra1 against additional bile constituents using G_⍺15-expressing HEK cells. G_⍺15 is a permissive G_α protein capable of coupling GPCRs to intracellular calcium stores via phospholipase C (PLC) (30). As measured by intracellular Ca²⁺ changes, numerous BAs, both primary and secondary, activated MRGPRX4 (Fig. 1 B and C). We tested BA activity against 12 closely related mouse Mrgprs (SI Appendix, Fig. S1A) as well as sequence-similar human receptors, MRGPRX1–X3. No mouse receptors, including the previously identified mouse ortholog Mrgpra1 or sequence-similar human receptors, were activated by BAs (SI Appendix, Fig. S1 B–E). Based on this result, we concluded that BAs specifically activated MRGPRX4 among the Mrgpr family members.

Open in a separate window

Fig. 1.

BAs activate MRGPRX4, a human sensory neuron-expressed GPCR. (A) The molecular stereostructures of primary and secondary BAs and their metabolic precursor, cholesterol. Major structural differences are highlighted in red. R1 denotes the potential presence of an α-hydroxyl substituent in some BA derivatives. R2 denotes additional substituents that vary among BAs. (B–H) Ca²⁺ imaging of HEK293 cells stably expressing MRGPRX4. (B) Representative Fura-2 fluorescence heat map images of HEK cells showing changes in intracellular [Ca²⁺] induced by DCA (10 µM). (Scale bar, 20 µm.) (C and F) Concentration–Ca²⁺ response curves of BAs against MRGPRX4. Data are a representative experiment of three independent replicates performed in triplicate, depicted as mean ± SEM (C) Concentration–Ca²⁺ response curves of unconjugated BAs. OA, oleic acid. (F) Concentration–Ca²⁺ response curves of tauro-conjugated BAs. (D, E, G, and H) Ca²⁺ imaging traces. A total of 10 µM DCA was added where indicated by black bars. Mean ± 95% CI depicted. n = 15. (E, G, and H) Cells were preincubated with either 10 µM sarco/endoplasmic reticulum Ca²⁺-ATPase inhibitor thapsigargin (E), 10 µM of the G_αq inhibitor YM254890 (G), or 10 µM PLC inhibitor U73122 (H) for 30 min before imaging.

Ca²⁺ imaging of MRGPRX4-expressing HEK293 cells displayed robust Ca²⁺ flux in response to 10 μM deoxycholic acid (DCA) (Fig. 1 B and D). Thapsigargin, a noncompetitive inhibitor of the sarcoendoplasmic reticulum Ca²⁺ ATPase, completely inhibited BA-associated Ca²⁺ influx (Fig. 1E), indicating that MRGPRX4-induced Ca²⁺ signal was dependent on intracellular Ca²⁺ stores. Previous studies have identified MRGPRX4 as being G_αq linked (27, 29). In line with this, DCA-associated Ca²⁺ flux appears to be similarly G_αq linked, since both the G_αq inhibitor, YM254890, and the PLC inhibitor, U73122, inhibited Ca²⁺ signal (Fig. 1 G and H).

Of the BAs tested, DCA and ursodeoxycholic acid (UDCA) displayed the highest potency in terms of MRGPRX4-associated Ca²⁺ signal (Fig. 1C and SI Appendix, Fig. S1F). Both DCA and UDCA had EC₅₀s of ~5 μM, similar to their activity toward previously published receptors like TGR5 (17). Unlike TGR5, however, MRGPRX4 exhibited a larger range of potency for various BAs. For example, cholic acid (CA) displayed much lower potency against MRGPRX4 (EC₅₀ ~430 μM), well outside physiologic range. In contrast, the EC₅₀ of CA toward TGR5 is ~10 μM (17).

Primary bile acids can be conjugated to either glycine or taurine to form water-soluble bile salts. We tested three tauro-conjugated bile acids, tauro-DCA (TDCA), tauro-CDCA (TCDCA), and tauro-cholic acid (TCA) against MRGPRX4. All conjugated bile acids were capable of activating MRGPRX4 at concentrations similar to the respective unconjugated BA (Fig. 1F and SI Appendix, Fig. S1G).

MRGPRX4 is primarily expressed in sensory neurons of the DRG. To test whether BAs interact with MRGPRX4 in neurons, we generated humanized mice expressing MRGPRX4. We crossed Tg(Mrgpra3-GFP-Cre) (A3-Cre) animals with newly generated ROSA26^{loxSTOPloxl-MRGPRX4} (lsl-MRGPRX4) animals (Fig. 2A). A3-Cre animals express Cre recombinase under control of the Mrgpra3 promoter, which labels ~5% of the DRG sensory neurons that specifically encode for itch (20). In lsl-MRGPRX4 animals, expression of MRGPRX4 is under Cre-loxp control. Progeny that inherit both A3-Cre and lsl-MRGPRX4 (+X4) would be expected to express MRGPRX4 in mouse Mrgpra3-containing itch neurons (Fig. 2A). Cre-null but transgene-positive mice (Cre-, lsl-MRGPRX4) were studied as comparative controls.

Open in a separate window

Fig. 2.

The +X4 humanized mouse sensory neurons are more readily activated by BAs. (A) Diagram showing mating strategy to generate humanized (+X4) mice. Mrgpra3-Cre expression is restricted to mouse DRG. Approximately 5% of DRG would be expected to express receptor. qPCR (B and C) RT-PCR of RNA collected from both control (A3-Cre negative) and +X4 (A3-Cre positive) mice expressing Rosa26^lsl-MRGPRX4. (B) Mean ± 95% CI depicted. n = 3 representative samples. ***P < 0.001, Student’s t test. (C) Arrow indicates the expected band based on primer design. Four independent samples collected from different mice are depicted. (D and H) Ca²⁺ imaging traces of Ctrl and +X4 DRG neurons. Mean ± 95% CI depicted. n = 10 neurons. After a 10-s baseline, either 10 µM DCA (D) or 100 µM UDCA (H) was applied where indicated by black bars. After 1 min, 1 mM CQ was added. (E and I) Percent activation of Ctrl and +X4 DRG neurons from either 10 µM DCA (E) or 100 µM UDCA (I). (F and J) Percent overlap of CQ-activated Ctrl and +X4 DRG neurons (~5% of total) activated by both DCA and CQ (F) and both UDCA and CQ (J). (E, F, I, and J) *P < 0.05; **P < 0.01; χ² test. A neuron was considered to be activated if F > 0.2 for at least 30 s. (G and K) Venn diagrams of DCA- (G), UDCA- (K), and CQ-activated neurons as a percent of total DRG neurons.

The +X4 animals expressing A3-Cre and lsl-MRGPRX4 have high levels of MRGPRX4 mRNA in sensory neurons as assayed by both qPCR and RT-PCR (Fig. 2 B and C). Based on previously determined potencies, we selected DCA and UDCA for further experimentation. As assayed by Ca²⁺ imaging, DCA activated a subpopulation of small-diameter (~20 μm) +X4 DRG neurons (Fig. 2D and SI Appendix, Fig. S2 A and B). At the tested concentration, TGR5 receptor present in both control (Ctrl) and +X4 neurons would be activated. Indeed, DCA activated 3.4% of control neurons, likely due to activation of TGR5 (Fig. 2E). Approximately 25% of control neurons activated by DCA were also activated by chloroquine (CQ), a ligand for Mrgpra3 (Fig. 2 F and G). Based on the genetic strategy employed for +X4 generation, we hypothesized that DCA would activate a higher percentage of Mrgpra3, CQ-sensitive, sensory neurons in +X4 animals. DCA activated 6.4% of +X4 sensory neurons (Fig. 2E). In addition, DCA activated significantly more CQ-sensitive, Mrgpra3-expressing, neurons in +X4 DRG (Fig. 2 F and G). Based on these data, we concluded that +X4 humanized sensory neurons expressed functional MRGPRX4 receptor that could be activated by DCA.

TGR5 has been previously identified as being expressed in Mrgpra3⁺ itch neurons (10). To better disambiguate MRGPRX4 and TGR5 activity, we selected UDCA as a ligand for additional experimentation. All BAs tested against MRGPRX4 are reported to activate TGR5 (31). However, UDCA, the second most potent activator of MRGPRX4, is a significantly less potent TGR5 agonist (17). Based on these data, we predicted that lower concentrations of UDCA would activate more +X4 neurons than control neurons. Indeed, UDCA activated a similar percentage of +X4 DRG neurons compared with DCA but activated a significantly smaller percentage of control neurons (Fig. 2 H and I). Additionally, UDCA activated only 10% of CQ-sensitive control neurons while activating 54% of +X4 neurons (Fig. 2 J and K).

To test the in vivo relevance of bile acid agonism of MRGPRX4, we injected bile acids into control and +X4 animals and assessed acute itch. As before, the genetic strategy of MRGPRX4 humanization results in +X4 animals expressing MRGPRX4 receptor in Mrgpra3⁺ itch neurons. Therefore, we hypothesized that a true, in vivo ligand–MRGPRX4 interaction would activate Mrgpra3⁺ neurons and produce itch behavior. Compared with control, +X4 animals scratched significantly more in response to injection of pathophysiologic concentrations of DCA, UDCA, CDCA, and TDCA at the nape (Fig. 3 A–D). This increase was specific to BAs, as histamine and chloroquine, two canonical pruritogens, stimulated similar levels of itch in control and +X4 animals (SI Appendix, Fig. S3 A and B). Additionally, injection of cholestatic patient plasma, containing BAs and numerous other pruritogens (SI Appendix, Fig. S3C), elicited greater itch in +X4 animals compared with control animals (Fig. 3D). This difference between Ctrl and +X4 animals was not observed upon injection of control plasma (Fig. 3D). Compared with normal plasma, patient plasma elicited greater itch in control animals (Fig. 3D). This could result from numerous pruritogens acting at established itch receptors in mice, such as bilirubin at Mrgpra1 or BAs at TGR5.

Open in a separate window

Fig. 3.

Mice expressing MRGPRX4 scratch more after injection with unconjugated and conjugated BAs. (A) Scratching bouts associated with injection of DCA. A total of 50 µL of 1 mM DCA was injected into the nape of Ctrl and +X4 mice. Ctrl n = 13; +X4 n = 6. (B) Scratching bouts associated with injection of TDCA. A total of 50 µL of 1 mM TDCA was injected into the nape of Ctrl and +X4 mice. Ctrl n = 9; +X4 n = 5. (C) Scratching bouts associated with injection of UDCA. A total of 50 µL of 2 mM UDCA was injected into the nape of Ctrl and +X4 mice. Ctrl n = 6; +X4 n = 6. (D) Scratching bouts associated with injection of CDCA. A total of 50 µL of 2 mM CDCA was injected into the nape of Ctrl and +X4 mice. Ctrl n = 6; +X4 n = 6. (E) Scratching bouts associated with injection of human plasma from either normal or cholestatic patients (Chol) into Ctrl and +X4 mice. A total of 50 µL of plasma was injected into the nape of mice. For normal plasma: Ctrl n = 5; +X4 n = 7. For Chol plasma: Ctrl n = 5; +X4 n = 5. (A–D) Mean ± SEM depicted. Each open circle represents an individual mouse. (A–D) *P < 0.05; **P < 0.01; ***P < 0.001; two-tailed unpaired Student’s t test.

Having established that unconjugated and conjugated BAs elicited greater acute itch in +X4 animals, we sought to determine whether this observation remained true in a mouse model of cholestatic itch. To induce cholestasis, we administered α-napthyl isothiocyanate (ANIT) to mice. ANIT is a well-established model that elicits intrahepatic cholestasis in mice (27, 32). We treated control and +X4 animals with 25 mg/kg ANIT daily for 5 d while assessing spontaneous itch and weight 24 h after each treatment (Fig. 4A). Over the course of the experiment, cholestatic itch in +X4 animals was significantly different from control animals (Fig. 4B). The majority of this difference was observed at early timepoints. Specifically, +X4 animals scratched significantly more than baseline just 24 h after a single dose of ANIT (Fig. 4B). Subsequent treatment did not significantly increase pruritus. In control animals, itch was not significantly elevated until the third day of treatment (Fig. 4B). After day 3, control mice did not scratch more with subsequent treatments. During days 3–5 of treatment, itch between control and +X4 animals was indistinguishable (Fig. 4B). ANIT does not appear to directly activate MRGPRX4, since ANIT treatment did not elicit itch in either WT or +X4 animals immediately after treatment (SI Appendix, Fig. S4A). Control and +X4 animals displayed no differences in weight change from ANIT treatment. Both groups lost significant weight over the 5-d treatment course (Fig. 4C). Control and +X4 animals had similar degrees of cholestatic injury after either 1 or 5 d of ANIT treatment, as assessed by plasma levels of BAs, total bilirubin, albumin, aspartate aminotransferase (AST), alkaline phosphatase (ALP), and alanine aminotransferase (ALT) (Fig. 4 D and E and SI Appendix, Fig. S4 B–E).

Open in a separate window

Fig. 4.

The +X4 humanized mice display increased cholestatic itch concordant with serum BA elevation. (A) Experimental flowchart of cholestatic model. ANIT was given daily per os for 5 d. After 24-h treatment, spontaneous itch was assessed for 1 h. (B) Spontaneous scratching bouts from ANIT-treated animals. Itch was assessed for 1 h. Scratching bouts from Ctrl and +X4 were significantly different. P < 0.05 by ANOVA. For day 0: Ctrl n = 5, +X4 n = 7; for day 1: Ctrl n = 14, +X4 n = 11; for day 2: Ctrl n = 13, +X4 n = 9; for day 3: Ctrl n = 11, +X4 n = 9; for day 4: Ctrl n = 10, +X4 n = 9; and for day 5: Ctrl n = 7, +X4 n = 9. (C) Percent weight change from baseline of Ctrl and +X4 mice during ANIT treatment. For day 1: Ctrl n = 5, +X4 n = 7; for all other days: Ctrl n = 4, +X4 n = 6. (D) Plasma BA levels from Ctrl and +X4 mice at indicated days of ANIT treatment. For day 0: Ctrl n = 4, +X4 n = 4; for day 1: Ctrl n = 6, +X4 n = 9; for day 5: Ctrl n = 6, +X4 n = 7. (E) Total plasma bilirubin levels at indicated days of treatment. For day 1: Ctrl n = 3, +X4 n = 3; for day 5: Ctrl n = 5, +X4 n = 9. (B–E) Mean ± SEM depicted. Circles represent individual mice. n.s., not significant; *P < 0.05; **P < 0.01; Student’s t test.

Based on the observed time course of itch behavior, we hypothesized that humanized +X4 mice scratched more at earlier timepoints due to a ligand of MRGPRX4, such as BAs, being elevated during those periods. Indeed, at day 1, ANIT treatment resulted in significant elevation of plasma BAs, similar to levels observed on day 5. Importantly, plasma bilirubin, albumin, and liver enzyme levels were not elevated at day 1 (Fig. 4E and SI Appendix, Fig. S4 B–E). This distinction suggested that the increase in itch on days 1–2 in +X4 animals resulted from an increase in BAs and not bilirubin, another ligand of MRGPRX4. We hypothesized that +X4 cholestatic itch is similarly elevated compared with control at later timepoints (days 3–5) due to signaling from the intact bilirubin itch receptor Mrgpra1 in Ctrl animals and other potential pruritogen–receptor interactions.

In this study, we demonstrate that numerous unconjugated and conjugated BAs are ligands for MRGPRX4, a human sensory neuron-expressed GPCR. BAs activated humanized (+X4) mouse sensory neurons at higher percentages, indicating that BAs are capable of interacting with neuronally expressed receptor. Humanized mice scratched more in response to both acute injection of BAs and a chronic model of cholestatic itch. Cholestatic itch in humanized mice rose in conjunction with plasma BA concentration. Our results strongly support that MRGPRX4 is a receptor for BAs. Given the receptor’s expression pattern in humans, it is possible that MRGPRX4 mediates a BA component of cholestatic itch.

Despite being structurally similar, BAs displayed a wide range of affinity for MRGPRX4. CA, the least potent BA, has 3α, 7α, and 12α hydroxyl groups. In contrast, the higher potency BAs, DCA and UDCA, have only a 3α hydroxyl and a 12α or 7β hydroxyl group, respectively. Based on this, we hypothesize that BA potency for MRGPRX4 is structurally determined by hydroxyl groups at the 7 and 12 position carbons, and their effects on hydrogen bonding, sterics, and water solubility. Cholesterol has no detectable activity toward MRGPRX4 and is most clearly structurally different from BAs in that cholesterol exhibits more planarity among its four rings due to a double bond between carbons 5 and 6.

UDCA, one of the more potent determined agonists of MRGPRX4, has been shown to reduce itch in a significant number of patients suffering from pruritus associated with intrahepatic cholestasis of pregnancy and primary biliary cholangitis (33–35). Our data suggest that UDCA could promote itch via MRGPRX4. There are many possible explanations for this apparent discrepancy. Notably, UDCA may act as an antipruritic agent by decreasing hepatocellular injury, inflammation, and fibrosis (33–35). UDCA treatment also does not appear to increase total serum BA levels despite being a BA. Rather, while serum UDCA levels increase, UDCA either decreases or leaves total serum BAs unchanged. The molecular mechanisms by which UDCA improves cholestatic injury could result from a combination of effects, such as improving hepatocellular function, immunosuppressive effects through unknown mechanisms, and improvements in BA flow through the digestive system. Improvement of cholestasis in the absence of total BA serum level elevation would be expected to improve pruritus and is consistent with our data.

The role of BAs in eliciting cholestatic itch in patients is complex. Cholestyramine, a BA chelating resin, aids in BA secretion and is frontline therapy for cholestatic pruritus. In a double-blind trial, cholestyramine was shown to be effective for treating itch. However, more recently, colesevalam, a BA-binding resin many times more effective than cholestyramine, displayed contradictory trial results (36). This discrepancy in the effectiveness of serum BA-modulating therapy is buttressed by clinical observations that serum BA levels do not correlate with pruritus. Not all patients with pruritus have elevated serum BAs, and not all patients with elevated serum BAs exhibit itch. For example, serum BA levels are maximally elevated during liver failure, but these patients often report little to no itch.

Based on these data, we hypothesize that cholestatic itch is multifactorial, with numerous pruritogens present in bile that likely activate a variety of neuronal receptors. It is also possible that a single pruritogen may activate multiple receptors or that multiple bile components activate the same receptor. For example, we recently identified that MRGPRX4 is activated by bilirubin at pathophysiologic concentrations observed in cholestasis (27). With this is mind, our data support a model whereby both BAs and bilirubin contribute to cholestatic itch via agonism at the same receptor, MRGPRX4. In this model, even without taking additional pruritogen–receptor interactions into account, serum levels of a single marker would not be expected to perfectly correlate with itch. Therefore, to properly interpret correlations relying on serum metabolite values, multiple components would likely need to be considered in relation to itch. Additional hypotheses for this lack of correlation include possible desensitization and internalization of MRGPRX4 during late-stage liver failure and/or populational polymorphisms in MRGPRX4 sequence that alter receptor affinity. Further research is required to address these possibilities. However, despite these open questions, our data suggest that MRGPRX4 may be a potential therapeutic target for cholestatic itch.