Abstract
Chronic itch remains a highly prevalent disorder with limited treatment options. Most chronic itch diseases are thought to be driven by both the nervous and immune systems, but the fundamental molecular and cellular interactions that trigger the early development of itch, and the acute-to-chronic itch transition, remain unknown. Here, we show that skin-infiltrating neutrophils are key initiators of itch behaviors in atopic dermatitis, the most prevalent chronic itch disorder. Neutrophil depletion significantly attenuated itch-evoked scratching behaviors in a mouse model of atopic dermatitis. Neutrophils were also required for several key hallmarks of chronic itch, including upregulation of inflammatory cytokines, skin hyperinnervation, enhanced expression of itch signaling molecules, and upregulation of activity-induced genes and markers of neuropathic itch. Finally, we demonstrate that induction of CXCL10, a ligand of the CXCR3 receptor that promotes itch via activation of sensory neurons, is neutrophil-dependent and that CXCR3 antagonism attenuates chronic itch.
Introduction
Chronic itch is a debilitating disorder that affects millions of people worldwide (Dalgard et al., 2015; Matterne et al., 2011; Mollanazar et al., 2016). It is a symptom of a number of skin diseases and systemic disorders, as well as a side effect of a growing list of medications. Like chronic pain, chronic itch can also be a disease in and of itself (Berger et al., 2013; Stumpf and Ständer, 2013). Unlike acute itch, which can facilitate removal of crawling insects, parasites, or irritants, persistent scratching in chronic itch disorders damages skin, leading to secondary infection, disfiguring lesions, and exacerbation of disease severity (Mack and Kim, 2018; Yosipovitch and Bernhard, 2013). One of the most common chronic itch disorders is atopic dermatitis (AD; commonly known as eczema), which affects fifteen million people in the US alone. Severe AD often leads to the atopic march, where chronic itch and inflammation progress to food allergy, allergic rhinitis, and asthma.
Little is known about the underlying mechanisms that drive chronic itch pathogenesis. As such, studies of human chronic itch disorders have sought to identify candidate mechanisms of disease progression. A number of studies have identified biomarkers and disease genes in human atopic dermatitis (AD) itch lesions (BS et al., 2012; Ewald et al., 2017; Guttman-Yassky et al., 2009; Jabbari et al., 2012; Li et al., 2014; Suárez-Fariñas et al., 2015). Indeed, a recent study compared the transcriptomes of healthy skin to itchy and non-itchy skin from psoriasis and AD patients, revealing dramatic changes in expression of genes associated with cytokines, immune cells, epithelial cells, and sensory neurons (Nattkemper et al., 2018). However, due to the difficulty in staging lesion development and obtaining staged samples from patients, there is currently no temporal map of when individual molecules and cell types contribute to chronic itch pathogenesis. Furthermore, the use of human patient data does not allow for rigorous mechanistic study of how disease genes contribute to chronic itch. To this end, we used a well-characterized inducible animal model of itch to define where, when, and how these genes identified from patient data contribute to chronic itch pathogenesis.
We employed the MC903 mouse model of AD and the atopic march (Dai et al., 2017; Li et al., 2009a; 2006; Moosbrugger-Martinz et al., 2017; Zhang et al., 2009) to provide a framework in which to identify the molecules and cells that initiate the development of atopic itch. The MC903 model is ideal for this approach because of its well-defined time course of lesion development and scratching onset (Li et al., 2006; Morita et al., 2015; Oetjen et al., 2017; Zhang et al., 2009). While MC903 has been widely used to study the chronic phase of AD (Kim et al., 2013; 2014; Li et al., 2006; 2009b; Naidoo et al., 2018; Oetjen et al., 2017; Zhang et al., 2009), we hypothesized that MC903 could also be used to define the early mechanisms underlying the development of chronic itch, beginning with healthy skin. We performed RNA-seq of skin at key time points in the model and complemented this approach with measurements of itch behavior and immune cell infiltration. The primary goal of our study was to identify the inciting molecules and cell types driving development of chronic itch. To that end we found that infiltration of neutrophils into skin is required for development of chronic itch behaviors. Additionally, we demonstrate that early hyperinnervation of skin, and upregulation of itch signaling molecules and activity-induced genes in neurons are also dependent on neutrophils. Finally, we identify CXCL10/CXCR3 signaling as a key link between infiltrating neutrophils and sensory neurons that drives itch behaviors.
MC903 triggers rapid changes in expression of skin barrier, epithelial cell-derived cytokine, and axon guidance genes
Although a variety of AD- and chronic itch-associated genes have been identified, when and how they contribute to disease pathogenesis is unclear. Using RNA-seq of MC903-treated skin, we observed distinct temporal patterns by which these classes of genes were differentially expressed across the first eight days of the model (Figure 1A-B, Figure 1-Figure Supplement 1A). Proteases (Klk6, Klk13, among others) and skin barrier genes (Cdhr1) changed as early as six hours after the first treatment, before mice begin scratching (Figure 1A). Increased protease activity in AD skin is thought to promote breakdown of the epidermal barrier and release of inflammatory cytokines from keratinocytes (Steinhoff et al., 2003; Yosipovitch, 2004). One such cytokine, thymic stromal lymphopoetin (TSLP) is a key inducer of the Type 2 immune response, which is characteristic of human AD and the MC903 model, via signaling in CD4+ T cells, basophils, and other immune cells (Briot et al., 2009; Demehri et al., 2009; Gao et al., 2010; Kim et al., 2013; Li et al., 2009a; Zhang et al., 2009). Beginning at day two, before any significant itch-evoked scratching (Figure 1A), immune cell infiltration (Figure 1E-G, Figure 1-Figure Supplement 2A-C), or skin lesions (data not shown) were observed, we saw increases in Tslp, as well as several other epithelial-derived cytokines, including the neutrophil chemoattractant genes Cxcl1, Cxcl2, Cxcl3, and Cxcl5 (Figure 1D). To ask whether upregulation of these chemokine genes was dependent on protease activity, we treated human keratinocytes with the Protease Activated Receptor 2 agonist SLIGRL. SLIGRL treatment triggered increased expression of several of these chemokine genes, including IL-8, the human ortholog of mouse Cxcl1/Cxcl2, and CXCL2 (Figure 1-Figure Supplement 3A). These increases occurred after a few hours of exposure to SLIGRL, suggesting that increased protease activity can rapidly trigger increases in neutrophil chemokines in skin, similar to what we observe in MC903-treated mouse skin.
Unexpectedly, we observed early changes in a number of transcripts encoding neuronal outgrowth factors (Ngf, Artn) and axon pathfinding molecules (Slit1, Sema3d, Sema3a), some of which are directly implicated in chronic itch (Hidaka et al., 2016; Kou et al., 2012; Tominaga et al., 2008; Tominaga and Takamori, 2014); Figure 1-Figure Supplement 4A), prior to when mice began scratching. We thus used immunohistochemistry (IHC) of whole-mount skin to examine skin innervation at this time point. We saw increased innervation of lesions at day two but not day one of the model (Figure 1H-I, Figure 1-Figure Supplement 5A). This increase came as a surprise because such changes had previously only been reported in mature lesions from human chronic itch patients (Haas et al., 2010; Kamo et al., 2013; Nattkemper et al., 2018; Oaklander and Siegel, 2005; Pereira et al., 2016; Schüttenhelm et al., 2015; Tominaga et al., 2009). Our findings suggest that early hyperinnervation is promoted by local signaling in the skin and is independent of the itch-scratch cycle.
Neutrophils are the first immune cells to infiltrate AD skin
At day five, mice exhibited robust itch behaviors and stark changes in a number of AD disease genes (Figure 1A-B). For example, loss-of-function mutations in filaggrin (FLG) are a major risk factor for human eczema (Palmer et al., 2006). Interestingly, Flg2 levels sharply decreased by this time point. In parallel, we saw a continued and significant elevation in neutrophil and basophil chemoattractant genes (Cxcl1,2,3,5, and Tslp, Figure 1D). Using flow cytometry, we determined that neutrophils were the most highly increased cell type in the skin at day five (Figure 1E). It was not until day eight that we observed the classical AD-associated immune signature (Gittler et al., 2012) in the skin, with upregulation of Il4, Il33 and other Th2-associated genes (Figure 1B, Figure 1D). We also observed increases in the T cell chemoattractants Cxcl9, Cxcl10, and Cxcl11 (Figure 1D), which are thought to be hallmarks of chronic AD lesions in humans (Mansouri and Guttman-Yassky, 2015; Oetjen and Kim, 2018). Neutrophils and a number of other immune cells that started to infiltrate on day five were significantly elevated in skin by day eight, including basophils (Figure 1F), CD4+ T cells (Figure 1G), inflammatory monocytes (Figure 1-Figure Supplement 2A), and mast cells (Figure 1-Figure Supplement 2B), but not eosinophils (Figure 1-Figure Supplement 2C).
CD4+ T cells are ubiquitous in human AD lesions (Guttman-Yassky and Krueger, 2017) and have been shown to play a key role in promoting chronic AD itch and inflammation (Oetjen et al., 2017). More specifically, they are implicated in IL4Rα-dependent sensitization of pruriceptors in the second week of the MC903 model (Oetjen et al., 2017). Thus, we were quite surprised to find that itch behaviors preceded significant CD4+ T cell infiltration. Therefore, neutrophils drew our attention as potential early mediators of MC903 itch. While neutrophil infiltration is a hallmark of acute inflammation, it remains unclear whether neutrophils contribute to the pathogenesis of chronic itch. Moreover, neutrophils release known pruritogens, including proteases and/or histamine, and secrete inflammatory lipids and cytokines that sensitize and/or activate pruriceptors (Dong and Dong, 2018; Hashimoto et al., 2018). Increased levels of PGE2 and the neutrophil-specific lipid LTB4 have also been reported in skin of AD patients (Fogh et al., 1989). Indeed, by mass spectrometry, we observed increases in several of these inflammatory lipids, PGD2 and PGE2, as well as LTB4 and its precursor 5-HETE (Figure 1-Figure Supplement 6A). Thus, we next tested the requirement of neutrophils to itch in the MC903 model.
Neutrophils are required for early itch behaviors in the MC903 model of AD
We first asked whether neutrophils were required for MC903-evoked itch by depleting neutrophils, the earliest and largest population of infiltrating immune cells in this chronic itch model. Depletion of neutrophils by daily injections of an anti-Gr1 (aGr1) antibody (Ghasemlou et al., 2015; Kim et al., 2009; Sivick et al., 2014) dramatically attenuated itch-evoked scratching through the first eight days of the model (Figure 2A). Consistent with a key role for neutrophils in driving chronic itch, our depletion strategy significantly and selectively reduced circulating and skin infiltrating neutrophils on days five and eight, days on which control, but not depleted mice, scratched robustly (Figure 2B; Figure Supplement 1A-C). In contrast, basophils and CD4+ T cells continued to infiltrate the skin following aGr1 treatment (Figure 2C-D), suggesting that these cells are not required for early MC903 itch.
We next used the cheek model of acute itch (Shimada and LaMotte, 2008) to ask whether neutrophil recruitment is sufficient to trigger itch. As expected, we observed significant and selective recruitment of neutrophils to cheek skin within 15 minutes after CXCL1 injection (Figure 2-Figure Supplement 2A-B). CXCL1 injection also triggered robust scratching behaviors (Figure 2E) on a similar time course to neutrophil infiltration (Figure 2 - Figure Supplement 2B). Thus, we depleted neutrophils with aGr1 to determine whether neutrophils were required for CXCL1-evoked acute itch. Indeed, aGr1-treatment significantly reduced numbers of skin-infiltrating neutrophils (Figure 2-Figure Supplement 3A) and resulted in a dramatic loss of CXCL1-evoked itch behaviors (Figure 2C). This effect was specific to neutrophil-induced itch as injection of chloroquine, a pruritogen that directly activates sensory neurons, triggered robust scratching in aGr1-treated animals (Figure 2-Figure Supplement 4A). Given that CXCL1 has been shown to directly excite and/or sensitize sensory neurons (Deftu 2017,2018), it is possible that the CXCL1 mechanism in itch may also involve neuronal pathways. However, our results show that CXCL1-mediated neutrophil infiltration is sufficient to drive acute itch behaviors, and that neutrophils are necessary for itch in the MC903 model.
We also examined the MC903-evoked itch behaviors of mice deficient in Crlf2, the gene encoding the TSLP receptor (TSLPR-/- mice; (Carpino et al., 2004). TSLPR is expressed by both immune cells and sensory neurons and is a key mediator of atopic dermatitis in human and mouse models. Surprisingly, MC903-treated TSLPR-/- mice displayed robust scratching behaviors during the first eight days of the model (Figure 2F). In contrast to our results in aGR1-injected mice, TSLPR-/- mice displayed robust neutrophil infiltration (Figure 2G), but a complete loss of basophil and CD4+ T cell infiltration into the skin (Figure 2H-I). These results suggest that basophils and T cells are not required for early itch and further support an inciting role for neutrophils. Previous studies have shown that TSLP drives the expression of Type 2 cytokines and related immune cells that promote itch and inflammation in mature AD skin lesions and in the second week of the MC903 AD model (Briot et al., 2009; Demehri et al., 2009; Li et al., 2009a; Zhang et al., 2009). Consistent with a later stage role for TSLP signaling in AD, we did observe a significant reduction in itch-evoked scratching in TSLPR-/- mice in the second week of the model (Figure 2F). Thus, our data support a model in which neutrophils are necessary for initiation of AD and itch behaviors early in the development of AD, while TSLPR signaling mediates the recruitment of basophils and CD4+ T cells to promote later stage itch and chronic inflammation.
MC903 drives rapid and robust changes in the peripheral and central nervous systems
But how do neutrophils drive MC903 itch? Itchy stimuli are detected and transduced by specialized subsets of peripheral somatosensory neurons. Thus, to answer this question we first profiled the transcriptional changes in somatosensory neurons in the MC903 model, which were previously unstudied. In general, little is known regarding neuronal changes in chronic itch. Our initial examination of early hyperinnervation and changes in axon guidance molecules in skin suggested that neurons are indeed affected early on in the MC903 model, before the onset of itch-evoked scratching behaviors. In contrast to the skin, where we saw many early transcriptional changes, we did not see any significant transcriptional changes in the trigeminal ganglia (TG) until five days after the first treatment, and in total only 84 genes were differentially expressed through the eighth day (Figure 3A-B). These hits included genes related to excitability of itch sensory neurons (Dong and Dong, 2018; Usoskin et al., 2015), neuroinflammatory genes (Takeda et al., 2009) and activity-induced or immediate early genes (Figure 3A). Interestingly, we observed enrichment of neuronal markers expressed by one specific subset of somatosensory neurons that are dedicated to itch (Il31ra, Osmr, Trpa1, Cysltr2, and Nppb (Solinski et al., 2019; Usoskin et al., 2015), termed “NP3” neurons. Similar to what has been reported in mouse models of chronic pain, we observed changes in neuroinflammatory (Bdnf, Nptx1, Nptx2, Nptxr) and immune genes (Itk, Cd19, Rag, Tmem173). However, these transcriptional changes occurred just a few days after itch onset, in contrast to the slow changes in nerve injury and pain models that occur over weeks (Scholz and Woolf, 2007), indicating that neuropathic changes may occur sooner than previously thought. These changes occurred in tandem with the onset of scratching behaviors (Figure 1C), suggesting that the early molecular and cellular changes we observed by this time point may be important for development of itch-evoked scratching.
The changes we observed in immune-related genes in the TG were suggestive of infiltration or expansion of immune cell populations, which has been reported in models of nerve injury and chronic pain, but has never been reported in chronic itch. To validate our observations, we used IHC to ask whether CD45+ immune cells increase in the TG. We observed a significant increase in immune cell counts at day eight but not day five (Figure 3C-F, Figure 3-Figure Supplement 1A-D). Because we observed such dramatic expression changes in the TG on day eight of the model, we postulated that the CNS may also be affected at this time point. Thus, we performed RNA-seq on spinal cord segments that innervate the rostral back skin of MC903-treated mice. To date, only one study has examined changes in the spinal cord during chronic itch. They showed that upregulation of the STAT3-dependent gene Lcn2 occurred three weeks after induction of chronic itch and was essential for sustained scratching behaviors (Shiratori-Hayashi et al., 2015). Surprisingly, we saw upregulation of Lcn2 on day eight of the MC903 model and, additionally, we observed robust upregulation of immediate early genes (Fos, Junb, Figure 3G), suggesting that MC903 itch drives activity-dependent changes in the spinal cord as early as one week after beginning treatment. Together our findings show that sustained itch and inflammation can drive changes in the PNS and CNS much sooner than previously thought, within days rather than weeks after the onset of scratching. We next set out to ask how loss of neutrophils impacted the molecular changes observed in skin and sensory neurons in the MC903 model, and which of these changes might contribute to neutrophil-dependent itch.
Neutrophils are required for upregulation of select itch- and atopic-related genes, including the itch-inducing chemokine CXCL10
To ask how neutrophils might promote itch in the MC903 model, we examined the transcriptional changes in skin and sensory ganglia isolated from non-itchy neutrophil-depleted animals and from the TSLPR-/- mice, which scratched robustly. In both the neutrophil-depleted and TSLPR-/- skin, several AD-associated cytokines were similarly dysregulated, including Il33 (Figure 4A, Figure 4-Figure Supplement 1A). Expression of epithelial-derived cytokines and chemokines Tslp, Cxcl1, Cxcl2, Cxcl3, and Cxcl5 were unaffected by either loss of TSLPR or neutrophil depletion (Figure 4B), suggesting these molecules are produced by skin cells even when the MC903-evoked immune response is compromised. Among the hundreds of MC903-dependent genes we examined, only a handful of genes were uniquely affected by aGr1 treatment. One such gene was Cxcl10, a chemokine known to be released by skin epithelial cells, neutrophils, and other myeloid cells (Hashimoto et al., 2018; Ioannidis et al., 2016; Kanda et al., 2007; Koga et al., 2008; Michalec et al., 2002; Padovan et al., 2002; Tamassia et al., 2007). Cxcl10 expression was increased in TSLPR-/- but not neutrophil-depleted skin (Figure 4B, Figure 4-Figure Supplement 1A). CXCL10 has been previously shown to drive acute itch in a model of allergic contact dermatitis via CXCR3 signaling in sensory neurons (Qu et al., 2015), and is elevated in skin of AD patients (Mansouri and Guttman-Yassky, 2015). Expression of Cxcl9 and Cxcl11, two other CXCR3 ligands that are elevated in AD but have an unknown role in itch, was also decreased in AD skin of neutrophil-depleted mice (Figure 4B).
CXCR3 signaling is necessary for MC903-evoked chronic itch
We hypothesized that neutrophil-dependent upregulation of CXCL10 activates sensory neurons to drive itch behaviors. Consistent with this model, neutrophil depletion attenuated the expression of activity-induced immediate early genes (Vgf, Junb) in the TG, suggestive of neutrophil-dependent sensory neuronal activity in the MC903 model (Figure 4C, Figure 4-Figure Supplement 1B). We found that neutrophils also contributed to other sensory neuronal phenotypes in the model. For example, we observed that expression of Lcn2, a marker of neuropathic itch, and activity-induced genes Fos and Junb were not increased in spinal cord isolated from neutrophil-depleted animals after MC903 treatment, indicating that neutrophil-dependent scratching behaviors may indeed drive changes in the CNS (Figure 4D). We also observed that neutrophil-depleted animals displayed no skin hyperinnervation at day two (Figure 4E). This result was surprising because we did not observe significant neutrophil infiltration at this early time point, but these data suggest that low numbers of neutrophils are sufficient to mediate these early effects.
To test our model wherein CXCL10 activation of CXCR3 drives neutrophil-dependent itch, we first asked whether this CXCR3 ligand is in fact released in MC903-treated skin. We performed ELISA on cheek skin homogenate and found that CXCL10 protein was increased in MC903-treated skin, but not in skin from neutrophil-depleted mice (Figure 4F). To test whether CXCR3 signaling directly contributes to MC903-evoked itch, we asked whether acute blockade of CXCR3 signaling using the antagonist AMG 487 (Qu et al., 2015) affected scratching behaviors in the MC903 model. We found that the CXCR3 antagonist strongly attenuated scratching behaviors on days five, eight, and twelve (Figure 4G), with the greatest effect at day eight. In contrast, CXCR3 blockade did not attenuate scratching behaviors in naive mice injected with the pruritogen chloroquine (Figure 4G), demonstrating that CXCR3 signaling contributes to chronic itch but is not required for scratching in response to acute pruritogens. Thus, we propose that neutrophils promote chronic itch via upregulation of CXCL10 and subsequent activation of CXCR3 (Figure 5).
Discussion
There is great interest in unraveling the neuroimmune interactions that promote acute and chronic itch. Here, we show that neutrophils are essential for the early development of MC903-evoked itch. We further show that the acute recruitment of neutrophils to the skin is sufficient to drive itch behaviors within minutes of infiltration. While neutrophils are known to release a variety of pruritogens, their roles in itch and AD were not studied (Hashimoto et al., 2018). Only a few studies have even reported the presence of neutrophils in human AD lesions (BS et al., 2012; Koro et al., 1999; Mihm et al., 1976; Shalit et al., 1987). Neutrophils have been implicated in psoriatic inflammation and inflammatory pain (Carreira et al., 2013; Cunha et al., 2008; Pinho-Ribeiro et al., 2018; Sumida et al., 2014), where they are thought to rapidly respond to tissue injury and inflammation (Kim and Luster, 2015; Kolaczkowska and Kubes, 2013), but they have not been directly linked to itch. A previous study showed that the neutrophil chemoattractant Cxcl1 was elevated in a mouse model of dry skin itch (Wilson et al., 2013a). While tissue damage can trigger upregulation of neutrophil chemoattractants and neutrophil infiltration, Cxcl1 expression was upregulated both in scratched and unscratched itchy skin, suggesting that upregulation of CXCL1, and subsequent neutrophil infiltration, is not simply a consequence of tissue damage or the itch-scratch cycle.
TSLP signaling triggers development of a Type 2 immune response, which is crucial for the progression of atopic disease. For example, Th2 cytokines, most notably IL-4 and IL-13, are thought to drive the transition from the acute to chronic phase of human AD, and IL4Rα signaling was recently shown to mediate inflammation and itch in the second week of the MC903 AD mouse model (Oetjen et al., 2017). Interestingly, the same study showed that IL-4 sensitizes the population of itch sensory neurons that responds to TSLP. Based upon their observation, and because TSLP is induced early on in the MC903 model, we expected TSLP to contribute to early MC903 itch. We were surprised to find that TSLPR signaling is dispensable for itch behaviors in the first week of the model, and is only required for itch in the second week. In light of our observation that neutrophils are required for early MC903 itch, and that TSLPR signaling is necessary for increased IL-4 production, we hypothesize that TSLPR signaling plays a key role in later, IL-4-dependent itch, whereas early itch is independent of IL-4. Our findings align well with a recent study showing that TSLP and IL4 act on overlapping sensory neuron populations and that neuronal IL4Rα promotes itch during the second week of the MC903 model (Oetjen et al., 2017).
Given the large magnitude of the itch deficit in the neutrophil-depleted mice, we were surprised to find fewer expression differences in MC903-dependent, AD-associated genes between neutrophil depleted and non-depleted mice than were observed between WT and TSLPR-/- mice. One of the few exceptions were the Th1-associated genes Cxcl9/10/11 (Brunner et al., 2017; Ewald et al., 2017). We found that induction of these genes and CXCL10 protein was completely dependent on neutrophils. While our results do not identify the particular cell type(s) responsible for neutrophil-dependent CXCL10 production, a number of cell types present in skin have been shown to produce CXCL10, including epithelial keratinocytes, myeloid cells, and sensory neurons (Flier et al., 2001; Hashimoto et al., 2018; Ioannidis et al., 2016; Kanda et al., 2007; Koga et al., 2008; Michalec et al., 2002; Padovan et al., 2002; Steain et al., 2011; Tamassia et al., 2007). We also observed striking differences in neutrophil-dependent gene expression in the spinal cord, where expression of activity-induced genes and the chronic itch gene Lcn2 were markedly attenuated by loss of neutrophils. Our observations newly implicate neutrophils in setting the stage for the acute-to-chronic itch transition by triggering molecular changes necessary to develop a chronic, itchy lesion.
Additionally, we demonstrate a novel role of CXCR3 signaling in MC903-induced itch. The CXCR3 ligand CXCL10 contributes to mouse models of acute and allergic itch (Liu et al., 2016; Qu et al., 2015); however, its role in chronic itch was previously unknown. Our results show that CXCL10 upregulation is dependent on neutrophils and that CXCR3 blockade dramatically attenuates itch behaviors. Our findings are in alignment with a recent study showing that dupilumab, a new AD drug that blocks IL4Rα, a major downstream effector of the TSLP signaling pathway, does not significantly reduce CXCL10 levels in human AD lesions (Hamilton et al., 2014). Taken together, this suggests that the TSLP-IL-4 and neutrophil-CXCL10 pathways are not highly interdependent, and supports our findings that Il4 is robustly upregulated in the absence of neutrophils. Additionally, targeting IL4Rα has been successful in treating itch and inflammation in some, but not all, AD patients (Simpson et al., 2016). We propose that biologics or compounds targeting neutrophils and/or the CXCR3 pathway may be useful for AD that is incompletely cleared by dupilumab monotherapy. Drugs targeting neutrophils are currently in clinical trials for the treatment of psoriasis, asthma, and other inflammatory disorders. For example, MDX-1100, a biologic that targets CXCL10, has already shown efficacy for treatment of rheumatoid arthritis in phase II clinical trials (Yellin et al., 2012). While rheumatoid arthritis and AD have distinct etiologies (Scott et al., 2010), our work indicates that CXCL10 or CXCR3 may be promising targets for treating chronic itch. Our findings may be applicable to other itch disorders such as psoriasis and ACD where neutrophil chemoattractants and/or CXCL10 are also elevated. Overall, our data suggest that neutrophils incite early itch and inflammation in the development of AD through several mechanisms, including: 1) directly triggering itch upon infiltration into the skin (e.g. by release of known sensitizing/itchy lipids such as PGE2 and LTB4, Figure 4) and, 2) indirectly triggering itch by altering expression of known pruritogens (e.g. induction of Cxcl10 expression; (Flier et al., 2001; Hashimoto et al., 2018; Ioannidis et al., 2016; Kanda et al., 2007; Koga et al., 2008; Michalec et al., 2002; Padovan et al., 2002; Tamassia et al., 2007). These direct and indirect mechanisms for neutrophil-dependent itch may explain why neutrophils have a dramatic effect on scratching behaviors on not only day eight but also day five of the model, when neutrophils are recruited in large numbers, but CXCR3 ligands are not as robustly induced.
More generally, our study provides a framework for understanding how and when known human disease genes contribute to the distinct stages of chronic itch pathogenesis. Our analysis of MC903-evoked transcriptional changes suggests we may be able to extend findings in the model not only to atopic dermatitis, but also to related disorders, including specific genetic forms of atopy. For example, we provide evidence that the MC903 model may also model the filaggrin loss-of-function mutations, which are a key inciting factor in human heritable atopic disease (Palmer et al., 2006; Schuttelaar et al., 2009). There are many rich datasets looking at mature patient lesions (BS et al., 2012; Ewald et al., 2017; Guttman-Yassky et al., 2009; Jabbari et al., 2012; Li et al., 2014; Suárez-Fariñas et al., 2015) and datasets for mature lesions in other mouse models of chronic itch (Liu et al., 2016; Oetjen et al., 2017; Swindell et al., 2011; 2017). Our study adds a temporal frame of reference to these existing datasets to probe the function of human disease genes in AD in greater detail. Furthermore, we have mapped the time course of gene expression changes in primary sensory ganglia and spinal cord during chronic itch development. We show that the MC903 model recapitulates several hallmarks of neuropathic disease on a time course much shorter than has been reported for chronic itch, or chronic pain. Nervous system tissues are extremely difficult to obtain from human AD patients, and thus little is known regarding the neuronal changes in chronic itch disorders in both mouse models and human patients. Our findings can now be compared to existing and future datasets examining neuronal changes in chronic pain, diabetic neuropathy, shingles, neuropathic itch, psoriasis, and other inflammatory disorders where neuronal changes are poorly understood but clearly contribute to disease progression. The early changes we see in skin innervation, sensory ganglia, and spinal cord dovetail nicely with recent studies examining neuroimmune interactions in other inflammatory conditions (Baral et al., 2018; Blake et al., 2018; Pinho-Ribeiro et al., 2018), which all implicate early involvement of sensory neurons in the pathogenesis of inflammatory diseases.
Methods
Mouse studies
All mice were housed in standard conditions in accordance with standards approved by the Animal Care and Use Committee of the University of California Berkeley (12 hr light-dark cycle, 21 °C). Wild-type C57Bl6 mice were obtained from Charles River or Jackson Laboratories or raised in-house. TSLPR-/- mice were kindly provided by Dr. Steven Ziegler (Crlf2tm1Jni; (Carpino et al., 2004) and were backcrossed onto C57Bl6. All experiments were performed under the policies and recommendations of the International Association for the Study of Pain and approved by the University of California Berkeley Animal Care and Use Committee. Genotypes were assessed using standard PCR.
MC903 model of atopic dermatitis
MC903 (R&D Systems) was applied to the mouse cheek (20 μl of 0.2 mM in ethanol) once per day for 1-12 days. 100% ethanol was used as a vehicle control. 3-7 days prior to the beginning of treatment, mice were singly housed and their cheeks were shaved. All MC903 studies were performed on 8-12 week old age-matched mice. Behavior, RNA-seq, flow cytometry, and immunohistochemistry were performed on days 1, 2, 3, 5, 8 and/or 12. Behavioral scoring was performed while blind to experimental condition and mouse genotype. Itch behavior was scored for the first 30 min. For AMG 487 experiments in the MC903 model, 50 μL 3.31 mM AMG 487 (Tocris) or 20% HPCD-PBS vehicle was injected subcutaneously one hour prior to recording behavior (Qu et al., 2015). Both bout number and length were recorded.
MC903 RNA isolation and sequencing
At Days 1 (six hours post-treatment), 2, 5, and 8 post-treatment, mice treated with MC903 or vehicle were euthanized via isoflurane and cervical dislocation. Cheek skin was removed, flash-frozen in liquid nitrogen, and cryo-homogenized with a mortar and pestle. Ipsilateral trigeminal ganglia were dissected and both skin and trigeminal ganglia were homogenized for three minutes (skin) or one minute (TG and spinal cord) in 1 mL RNAzol RT (Sigma-Aldrich). Thoracic spinal cord was dissected from mice treated with 40 μL MC903 or ethanol on the shaved rostral back and homogenized in 1 mL RNAzol. Large RNA was extracted using RNAzol RT per manufacturer’s instructions. RNA pellets were DNase treated (Ambion), resuspended in 50 μL DEPC-treated water, and subjected to poly(A) selection and RNA-seq library preparation (Apollo 324). Single-end read sequencing (length = 50 bp) was performed by the QB3 Vincent G. Coates Genomic Sequencing Laboratory (UC Berkeley) on an Illumina HiSeq4000. See Supplementary Table 1 for number of mice per experimental condition and number of mapped reads per sample.
MC903 RNA sequencing analysis
Reads were mapped to the mm10 mouse genome using Bowtie2 and Tophat (Langmead and Salzberg, 2012; Langmead et al., 2009). For a given time point, replicate measurements for each gene from treated and control mice were used as input for DESeq (R) and genes with padjusted < 0.05 (for skin and spinal cord) or padjusted < 0.1 (for trigeminal ganglia) for at least one time point were retained for analysis (Anders and Huber, 2010). For the skin dataset, we collated a set of AD-related immune cell markers, cytokines, atopic dermatitis disease genes, neurite outgrown/axonal guidance genes, and locally expressed neuronal transcripts, and from this list visualized genes that were significantly differentially expressed for at least one time point. For the trigeminal ganglia dataset, we plotted all genes that were significantly differentially expressed for at least one time point. Genes from these lists were plotted with clustering using heatmap2 (R).
Custom gene groups
Genes were clustered into functional groups and significance was evaluated using a permutation test. Briefly, we first tabulated the absolute value of the log2 fold change of gene expression (between MC903 and EtOH) of each gene in a given group of n genes in turn, and then we calculated the median of these fold change values, ztrue. We then drew n random genes from the set of all genes detected in the samples and computed the median log2 fold change as above using this null set, znull. Repeating the latter 10,000 times established a null distribution of median log2 fold change values; we took the proportion of resampled gene groups that exhibited (ztrue ≥ znull) as an empirical p-value reporting the significance of changes in gene expression for a given group of n genes.
Flow Cytometry
Skin samples were collected from the cheek of mice at the indicated time points with a 4- or 6-mm biopsy punch and minced into smaller pieces with scissors. Myeloid cells were digested for 1h at 37 °C using Liberase (Roche). At the end of the digestion, cells were washed in FACS buffer (PBS with 0.5% FCS and 2 mM EDTA). Cells were stained with LIVE/DEAD fixable stain Aqua (Invitrogen), then blocked with anti-CD16/32 (UCSF Core) and stained with fluorophore-conjugated cKit-Biotin (ACK2) (secondary stain with SA-FITC), anti-CD11b-PB (M1/70), Ly6C-PerCP (HK1.4), CD49b-PECy7 (DX5), CD45.2-APCCy7 (104), FceRI-PE (MAR-1), Ly6G-AF700 (1A8, all from eBiosciences). 10ml of counting beads (Invitrogen) were added after the last wash to measure absolute cell counts. CD4+ T cells were digested for 30 minutes at 37°C using Collagenase VIII (Sigma). At the end of the digestion, cells were washed in RPMI buffer (RPMI, 5% FCS, 1% penicillin-streptomycin, 2 mM L-glutamine, 10 mM HEPES buffer, 1 mM sodium pyruvate). Cells were blocked with antiCD16/32 (UCSF Core) and stained with anti-CD45-APC780 (30-F11; eBiosciences), Cd11b-PECy7 (M1/70; BD Biosciences), B220-PECy7 (RA3-6B2; Tonbo Biosciences), Cd11c-PECy7 (N418; eBiosciences), CD3-FITC (145-2C11; eBiosciences), CD8-BV785 (53-6.7; Biolegend), CD4-PE (GK1.5; BD Biosciences), 647-gdTCR (GL3; Biolegend). 10ml of counting beads (Invitrogen) were added after the last wash to measure absolute cell counts, and samples were resuspended in DAPI LIVE/DEAD (Invitrogen). Single cell suspensions were analyzed on an LSR II or LSR Fortessa (BD Biosciences), and data were analyzed using FlowJo (TreeStar, v.9.9.3) software. One skin sample was excluded from analysis due to clear blood contamination.
Human keratinocyte RNA sequencing
Normal human epidermal keratinocytes from juvenile skin (PromoCell #C-12001) were cultured in PromoCell Keratinocyte Growth Medium 2 and passaged fewer than 5 times. Cells were treated for three hours at room temperature with 100 μM SLIGRL or vehicle (Ringer’s + 0.1% DMSO). Total RNA was extracted by column purification (Qiagen RNeasy Mini Kit). RNA was sent to the Vincent J. Coates Sequencing Laboratory at UC Berkeley for standard library preparation and sequenced on an Illumina HiSeq2500 or 4000. Sequences were trimmed (Trimmomatic), mapped (hg19, TopHat) and assigned to transcripts using htseq-count. Differential gene expression was assessed using R (edgeR).
IHC of whole-mount skin
Staining was performed according to Hill et al. (Hill et al., 2018). Briefly, 8-week-old mice were euthanized and the cheek skin was shaved and tape-stripped. The removed skin was fixed overnight in 4% PFA, then washed in PBS (3X for 10 min each). Dermal fat was scraped away with a scalpel and skin was washed in PBST (0.3% Triton X-100; 3X for two hours each) then incubated in 1:500 primary antibody (Rabbit anti beta-Tubulin III: Abcam #ab18207) in blocking buffer (PBST with 5% goat serum and 20% DMSO) for 6 days at 4°C. Skin was washed as before and incubated in 1:500 secondary antibody (Goat anti-Rabbit Alexa 594; Invitrogen #R37117) in blocking buffer for 3 days at 4°C. Skin was washed in PBST, serially dried in methanol: PBS solutions, incubated overnight in 100% methanol, and finally cleared with a 1:2 solution of benzyl alcohol: benzyl benzoate (BABB; Sigma) before mounting between No. 1.5 coverglass. Sectioned and whole mount skin samples were imaged on a Zeiss LSM 880 confocal microscope with OPO using a 20x water objective. Image analysis was performed using a custom macro in FIJI. Briefly, maximum intensity z-projections of the beta-tubulin III channel were converted to binary files that underwent edge-detection analysis. Regions were defined by circling all stained regions. Region sizes and locations were saved.
IHC of sectioned trigeminal ganglia
TG were dissected from 8 to 12 week old adult mice and post-fixed in 4% PFA for one hour. TG were cryo-protected overnight at 4°C in 30% sucrose-PBS, embedded in OCT, and then cryosectioned at 14 μm onto slides for staining. Slides were washed 3x in PBST (0.3% Triton X-100), blocked in 2.5% horse serum + 2.5% BSA PBST, washed 3X in PBST, blocked in endogenous IgG block (1:10 F(ab) anti-mouse IgG (Abcam ab6668) + 1:1000 Rat anti-mouse CD16/CD32 (UCSF MAB Core) in 0.3% PBST), washed 3X in PBST and incubated overnight at 4°C in 1:1000 primary antibody in PBST + 0.5% horse serum + 0.5% BSA. Slides were washed 3x in PBS, incubated 2 hr at RT in 1:1000 secondary antibody, washed 3X in PBS, and then incubated 30 min in 1:2000 DAPI-PBS. Slides were washed 3x in PBS and mounted in Fluoromount-G with No. 1.5 coverglass. Primary antibodies used: Mouse anti-CD45 (eBioscience #14-054-82) and Chicken anti-Peripherin (Abcam #39374). Secondary antibodies used: Goat anti-Chicken Alexa 594 (ThermoFisher #A11042) and Goat anti-Mouse Alexa 488 (Abcam #150117). DAPI (ThermoFisher #D1306) was also used. Imaging of TG IHC experiments was performed on an Olympus IX71 microscope with a Lambda LS-xl light source (Sutter Instruments). For TG IHC analysis, images were analyzed using automated scripts in FIJI (ImageJ) software. Briefly, images were separated into the DAPI, CD45, and Peripherin channels. The minimum/maximum intensity thresholds were batch-adjusted to pre-determined levels, and adjusted images were converted to binary files. Regions were defined by circling all stained regions with pre-determined size-criteria. Region sizes and locations were saved. All ImageJ macros are available upon request. All scripts are available upon request.
Neutrophil depletion
Neutrophils were acutely depleted using intraperitoneal injection with 250 μg aGR1 in PBS (clone RB6-8C5, a gift from D. Portnoy, UC Berkeley, or Biolegend), 16-24 hours before behavioral and flow cytometry experiments. Depletion was verified using flow cytometry on blood collected post-mortem. For longer depletion experiments using the MC903 model, mice were injected (with 250 μg aGR1 in PBS or PBS vehicle, i.p.) beginning one day prior to MC903 administration and each afternoon thereafter through day 7 of the model, and blood was collected via saphenous venipuncture at days 3, 5, or by decapitation at day 8 to verify depletion.
CXCL10 ELISA measurements in skin
Neutrophil-depleted or uninjected mice were treated with MC903 or ethanol for 7 days. On day 8, 6mm biopsy punches of cheek skin were harvested, flash-frozen in liquid nitrogen, cryo-homogenized by mortar and pestle, and homogenized on ice for three minutes at maximum speed in 0.5 mL of the following tissue homogenization buffer (all reagents from Sigma unless stated otherwise): 100 mM Tris, pH 7.4; 150 mM NaCl, 1 mM EGTA, 1 mM EDTA, 1% Triton X-100, and 0.5% Sodium deoxycholate in ddH2O; on the day of the experiment, 200 mM fresh PMSF in 100% ethanol was added to 1mM, with 1 tablet cOmplete protease inhibitor (Roche) per 50 mL, and 5 tablets PhosSTOP inhibitor (Roche) per 50 mL buffer. Tissues were agitated in buffer for two hours at 4°C, and centrifuged at 13,000 rpm for 20 minutes at 4°C. Supernatants were aliquoted and stored at −80C for less than one week after isolation. After thawing, samples were centrifuged at 10,000 rpm for five minutes at 4°C. Protein content of skin homogenates was quantified by BCA (Thermo Scientific) and homogenates were diluted to 2 mg/mL protein in PBS and were subsequently diluted 1:2 in Reagent Diluent (R&D Systems). CXCL10 protein was quantified using the Mouse CXCL10 Duoset ELISA kit (R&D Systems; #DY466-05) according to manufacturer’s instructions. Plate was read at 450 nm and CXCL10 was quantified using a seven-point standard curve (plus blank and buffer controls) and fitted with a 4-parameter logistic curve.
Acute itch behavior
Itch and acute pain behavioral measurements were performed as previously described (Morita et al., 2015; Tsunozaki et al., 2013; Wilson et al., 2013b). Mice were shaved one week prior to itch behavior. Compounds injected: 1 μg carrier-free CXCL1 (R&D systems) in PBS, 3.31 mM AMG 487 (Tocris, prepared from 100 mM DMSO stock) in 20% HPCD-PBS, 50 mM Chloroquine diphosphate (Sigma) in PBS (Chloroquine, CXCL1), along with corresponding vehicle controls. Acute pruritogens were injected using the cheek model (20 μL, i.d.) of itch, as previously described (Shimada and LaMotte, 2008). AMG 487 (50 μL) or vehicle was injected s.c. into the rostral back skin one hour prior to recording of behavior (Qu et al 2015). Behavioral scoring was performed as described above.
Lipidomics
Skin was collected from the cheek of mice post-mortem with a 6-mm biopsy punch and immediately flash-frozen in liquid nitrogen. Lipid mediators and metabolites were quantified via liquid chromatography-tandem mass spectrometry (LC-MS/MS) as described before (Moltke et al., 2012; Sapieha et al., 2011). In brief, skin was homogenized in cold methanol to stabilize lipid mediators. Deuterated internal standards (PGE2-d4, LTB4-d4, 15-HETE-d8, LXA4-d5, DHA-d5, AA-d8) were added to samples to calculate extraction recovery. LC-MS/MS system consisted of an Agilent 1200 Series HPLC, Luna C18 column (Phenomenex, Torrance, CA, USA), and AB Sciex QTRAP 4500 mass spectrometer. Analysis was carried out in negative ion mode, and lipid 30 mediators quantified using scheduled multiple reaction monitoring (MRM) mode using four to six specific transition ions per analyte.
Statistical analyses
Different control experimental conditions (e.g. uninjected versus PBS-injected animals) were pooled when the appropriate statistical test showed they were not significantly different (Supplementary Table 2). For all experiments except RNA-seq (see above), the following statistical tests were used, where appropriate: Student’s t-test, one-way ANOVA with Tukey-Kramer post hoc comparison, and two-way ANOVA with Tukey Kramer or Sidak’s post-hoc comparison. Bar graphs show mean ± SEM. Statistical analyses were performed using PRISM 7 software (GraphPad). For all p values, *=0.01<p<0.05, **=0.001<p<0.01, ***=0.0001<p<0.001, and ****=p<0.0001.
Conflict of interest statement
The authors declare no conflict of interest.
Acknowledgements
The authors would like to thank members of the Bautista and Barton labs for helpful discussions on the data. We are grateful to S. Ziegler (Ben Aroya Research Institute) for the gift of the TSLPR-/- mouse. We also thank M Pellegrino and L. Thé for pilot studies on human keratinocyte transcriptome analyses and S. Ziegler (Ben Aroya Research Institute) for the gift of the TSLPR-/- mouse. D.M.B. is supported by the NIH (AR059385, NS07224 (also to R.B.B)) and the Howard Hughes Medical Institute. G.M.B. is supported by the NIH (AI072429, AI063302, AI104914, AI105184) and the Burroughs Wellcome Fund. J.D. was supported by a Long-Term Fellowship from the Human Frontier Science Program (LT-000081/2013-L). K.G. is supported by NIH grant EY026082. This work used the Vincent J. Coates Genomics Sequencing Laboratory at UC Berkeley, supported by NIH S10 OD018174 Instrumentation Grant.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.
- 13.↵
- 14.↵
- 15.↵
- 16.
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.
- 96.↵
- 97.↵
- 98.↵