Emerging insights in sarcoidosis: moving forward through reverse translational research

Abstract

Sarcoidosis is a chronic granulomatous disease of unknown etiology that primarily affects the lungs. The development of stage IV or fibrotic lung disease accounts for a significant proportion of the morbidity and mortality attributable to sarcoidosis. Further investigation into the active mechanisms of disease pathogenesis and fibrogenesis might illuminate fundamental mediators of injury and repair while providing new opportunities for clinical intervention. However, progress in sarcoidosis research has been hampered by the heterogeneity of clinical phenotypes and the lack of a consensus modeling system. Recently, reverse translational research, wherein observations made at the patient level catalyze hypothesis-driven research at the laboratory bench, has generated new discoveries regarding the immunopathogenic mechanisms of pulmonary granuloma formation, fibrogenesis, and disease model development. The purpose of this review is to highlight the promise and possibility of these novel investigative efforts.

INTRODUCTION

Sarcoidosis is a multisystem, chronic granulomatous disease of unknown etiology that primarily affects the lungs (1). Although many patients experience stable lung disease (1, 2), some develop progressive pulmonary fibrosis that variably responds to immunosuppression, accounting for a significant proportion of the morbidity and mortality attributable to sarcoidosis (2, 3). Approximately 5% of all patients progress to stage IV disease (4), which is associated with significantly worse survival relative to the general population (5). The development of respiratory complications, namely pulmonary hypertension and chronic respiratory failure, were the leading causes of death among this patient population (5). Thus, further investigation into the active mechanisms of disease pathogenesis and tissue remodeling might illuminate fundamental mediators of injury and repair while providing new opportunities for clinical intervention. However, progress in sarcoidosis research has been hampered by 1) the heterogeneity of clinical phenotypes that results in a protean array of disease manifestations and trajectories that has yet to be fully defined (6) and 2) the lack of a consensus modeling system that replicates the tightly formed, T-helper type 1 (Th1) predominant granulomas (7) that are the defining feature of disease. Hence, the development of novel, innovative methods to study sarcoidosis has the potential to greatly advance our understanding of this enigmatic disease.

Translational research is defined as extending discoveries made in the laboratory setting (T0–T1) or clinical setting (T2–T4) into patient relevant interventions (8). “Reverse translational research” is a more recent concept, denoting the process by which observations made from patients inspire hypothesis-driven research at the laboratory bench or in the clinic (9). This novel research paradigm has fostered drug repurposing studies (10), ushered in the era of precision medicine (11), and expediated diagnostic and therapeutic efforts in managing public crises, with the most recent example being the coronavirus disease 2019 (COVID-19) pandemic (9, 12). In the field of sarcoidosis, patient-level studies have catalyzed the discovery of novel biomarkers (13) and therapies (14) that have generated new ideas regarding the immunopathogenic mechanisms of pulmonary granuloma formation, fibrogenesis, and model development. The purpose of this review is to highlight the promise and possibility of these reverse translational efforts.

PULMONARY SARCOIDOSIS GRANULOMA FORMATION

The cardinal histopathological feature of sarcoidosis is the presence of noncaseating granulomas composed of tightly formed aggregates of macrophages, epithelioid cells, and lymphocytes (1). The formation of these granulomas is initiated by the presence of a yet-to-be identified antigen, which leads to the differentiation of monocytes into antigen-presenting cells, mainly macrophages and dendritic cells (15). Macrophage aggregates evolve into epithelioid cells, whereas fusion of macrophages and dendritic cells form multinucleated giant cells (15). Within this central core of the granuloma, these cells secrete soluble mediators that foster Th1 differentiation of T lymphocytes, resulting in an outer rim composed of Th1 CD4+ cells, fibroblasts, and collagen (1, 16). These lymphocytes amplify immune responses through the expression of various Th1-associated soluble mediators that are instrumental to granuloma development and/or maintenance (17). Extensive study of the bronchoalveolar lavage fluid (BALF), an accepted surrogate of the lung milieu (18), obtained from patients with sarcoidosis has shed valuable insight into the heterogeneity of these soluble mediators as cytokines such as interleukin-2 (IL-2), IL-12, IL-18, tumor necrosis factor-α (TNF-α), and interferon-γ (IFN-γ) have been found in high concentrations (17). Taken together, the formation of the sarcoidosis granuloma is believed to be a spatiotemporal process of failed antigen clearance and persistent immune activation.

Investigation into sarcoidosis-relevant cytokines has catalyzed the successful repurposing of therapies initially developed for other applications. One result of this “bedside to the bench” approach targets the Janus kinase (JAK)-signal transducer and activator of transcription (STAT) pathway, a mediator of IFN-γ, IL-2, IL-6, and IL-12 signaling that is enriched in lung tissue and peripheral blood from patients with sarcoidosis (19, 20). Small series and case reports have demonstrated efficacy in treating cutaneous sarcoidosis with the JAK inhibitor tofacitinib (19–21). Although larger clinical trials are underway, interim molecular analysis revealed the novel finding of constitutively active expression of IFN-γ, IFN-α, and IL-6 in treatment-resistant sarcoidosis (20), raising the exciting possibility of this cytokine profile as a mediator of, and potential biomarker for, refractory disease. Interestingly, this study also demonstrated variable expression of TNF-α, which has been implicated in sarcoidosis granuloma formation and associated with corticosteroid-resistant disease (17, 22). Although results of clinical studies with TNF-α inhibitors have been promising (23), some reports have suggested a pathological association with TNF-α inhibition (24), a finding that both illustrates the complex immunopathogenesis of sarcoidosis and highlights the need to better understand the basic biology of granuloma formation.

The aforementioned studies have been recently augmented by the identification of new mediators such as mechanistic target of rapamycin complex 1 (mTORC1) (25, 26), which in animal modeling and cell culture work has shown to regulate hypertrophic macrophage proliferation and aggregation in experimentally induced granulomas (26). Excitingly, a connection to human disease was implied by the detection of mTORC1 expression in sarcoidosis granulomas (26), and in several case reports demonstrating successful treatment with mTOR inhibitors such as sirolimus and rapamune (27, 28). When viewed in combination, these studies provide an exciting clinical rationale for investigators to return to the laboratory bench and better understand the respective contributions of JAK-STAT, TNF-α, and mTOR to the clinical heterogeneity of this disease, with the ultimate goal of developing phenotype-driven management approaches.

FIBROTIC PULMONARY SARCOIDOSIS

Although most patients with sarcoidosis experience a relatively benign disease course, the presence of stage IV disease, or pulmonary fibrosis, often portends significant morbidity and mortality (4). Although the mechanisms driving this transformation are poorly understood, chronic granulomas are believed to engender fibrotic remodeling seen in stage IV sarcoidosis (16). The maintenance of these granulomas may be related to several factors, including high levels of serum amyloid A (SAA) (29) and the impaired function of regulatory T cells (Tregs) (30). SAA, an acute phase reactant produced in various inflammatory conditions, has been shown to enhance granuloma formation in experimental models through a Th1-like response with elevated levels of TNF, IL-18, and IL-10 (29). Similarly, Tregs typically function to suppress and terminate immune responses after resolution of injuries or antigen. Although enrichments in these cells have been reported in the blood, BALF, and granulomas of patients with sarcoidosis, in vitro studies have revealed defective immune suppression that contributes to the maintenance of the chronic inflammatory state (30, 31). Further mechanistic and functional study of these divergent molecular pathways are needed to elucidate their contribution(s) to this convergent disease state.

Additional T-cell anomalies implicated in stage IV disease include a pronounced Th2 response (1, 32, 33). This observation is of particular interest because the Th1 cytokines that are proposed to influence granuloma formation in acute disease also possess antifibrotic functions: for example, IFN-γ inhibits fibroblast extracellular matrix deposition and, together with IL-12, suppresses profibrotic transforming growth factor-β (TGF-β) activity (34). In contrast, Th2 responses driven by IL-5 and IL-13 are associated with the development of fibrosis. This concept is highlighted by the finding that IL-5 is increased in the serum of patients with fibrotic pulmonary sarcoidosis compared with those with nonfibrotic disease (33), along with the well-established role of IL-13 as a potent profibrotic factor that increases TGF-β production and myofibroblasts differentiation (16). Whether these Th1 responses are in concert with or independent of Th2 responses remains an area of active investigation that could shed new insights into the immunopathogenesis of this devastating phenotype.

Moreover, contributions of Th2 cytokines may involve the alternative, or “M2” polarization of macrophages (35). The role of macrophage polarization in fibrotic processes has been an area of great interest, and although there has been an increasing evidence favoring alternatively activated, or “M2” macrophage involvement in fibrosis, there have also been varying findings on the profibrotic properties of both the classically activated (“M1”) and M2 cell populations. The cytokine and chemokine environment observed in fibrosis reflects this overlap, as both M1 and M2 signature proteins are noted to be present (36). M1 macrophages are typically described as a proinflammatory cell population and have been shown to be increased in comparison with M2 macrophages in bleomycin mouse models of pulmonary fibrosis (37). When these mice were treated with an aspirin-triggered lipoxin analog that reversed bleomycin-induced pulmonary fibrosis, a corresponding decrease in the M1 population was observed (37). It has been further suggested that activation of these M1 macrophages by CCL17, a CC chemokine receptor 4 ligand, plays a role in inducing tissue injury that leads to fibrosis in the bleomycin model (38). On the other hand, suppression of M2 macrophage polarization in bleomycin models by Schisandra chinensis fructus (39) and microcystin-leucine arginine (40) resulted in inhibition of fibrosis through the reduction of TGF-β/Smad signaling. Furthermore, BALF studies from patients with interstitial lung disease demonstrated enhanced expression of M2 markers with samples from the fibrotic remodeling stages of sarcoidosis showing increased production of M2 cytokines, such as TGF-β and CCL18, that have been implicated in fibrotic progression (35). Interestingly, in studies involving lung tissue, relative to the lung granulomas arising in tuberculosis, CD163+ M2-like cells are enriched in pulmonary sarcoidosis, where they were most abundant in patients with stage IV disease (41). Despite the difficulties in comparing an idiopathic disease such as sarcoidosis with granulomatous processes arising in response to a specific exposure, these findings further suggest that M2 polarization accompanies the Th2 shift in sarcoidosis and promotes fibroproliferation (41). Work using disease-relevant models will be required to understand both the paradoxical nature of the immune response in fibrotic sarcoidosis and how these observations can be leveraged in the clinical setting.

This goal of better understanding fibrotic sarcoidosis has fostered efforts at precision medicine. For example, a 2010 landmark microarray study identified 334 differentially expressed genes in the lungs of patients with progressive fibrotic sarcoidosis when compared with individuals with self-limited disease. Pathway enrichment analysis revealed that patients with progressive fibrosis exhibited increased expression of genes associated with leukocyte and lymphocyte activation and differentiation, cytokine production, immune system development, immune defense response, intracellular signaling, cell cycle, wound response, and healing, although surprisingly, transcripts associated with fibroblast activation and extracellular matrix production were not differentially detected (42). These findings are complemented by a recent proteomics study which identified 121 differentially expressed proteins in the BALF of patients with progressive sarcoidosis when compared with those with stable disease (43). These proteins, which were linked to 27 different canonical pathways, revealed a milieu associated with aberrant immune responses (43), including a potential role for IL-8 signaling that has been implicated as a mediator of mesenchymal progenitor cell proliferation and activation and macrophage migration to fibroblastic foci in idiopathic pulmonary fibrosis (44). The emergence of these modern, high throughput assays have catalyzed precision medicine efforts, leading to ground-breaking insights into the fibrotic lung microenvironment.

These studies are paralleled by a longitudinal study of the whole blood transcriptome, in which differential expression of genes related to CXCL9 and T-cell receptor (TCR) signaling were associated with the development of chronic disease (45). Specifically, the authors found that among the subjects with chronic sarcoidosis, the expression of genes associated with CXCL9 was significantly elevated whereas those for TCR were significantly decreased; these findings persisted at 6 and 12 mo (45). Given the potential roles of CXCL9 in the recruitment of immune cells to areas of inflammation and TCR in immune activation (46), these findings suggest the paradoxical coexistence of inflammatory and anergic components as a concept that warrants further investigation in chronic and fibrotic disease. Most recently, the application of single-cell RNA sequencing has yielded previously unknown insights by demonstrating persistent activation of classical monocytes, innate immunity, and naïve CD4+ T cells coupled with regulatory T-cell dysfunction and enrichment of genes and pathways associated with, among others, TGF-β signaling (47). When viewed in parallel, these multiomics studies gleaned from well-characterized patient cohorts using high throughput approaches have accelerated discovery through the generation of novel hypotheses regarding the immunopathogenic mechanism(s) of this disease.

One challenge in studying sarcoidosis is that its status as a rare disease characterized by an unusual degree of phenotypic heterogeneity impedes the generalizability of single-center studies which are usually small and geographically confined. To overcome this barrier, NIH-sponsored, multicenter studies that have become indispensable for generating new hypotheses for the study of this disease (48, 49). One recent example is the Genomic Research in Alpha-1 Antitrypsin Deficiency and Sarcoidosis (GRADS) study, which enrolled and characterized subjects with various forms of sarcoidosis across nine US institutions from 2013 to 2015 (3). Using cells obtained from 215 BALF samples from this cohort, Vukmirovic et al. (50) conducted bulk RNA sequencing for the purposes of transcriptome-wide gene expression profiling. Here they identified four unique sarcoidosis endotypes, which included hilar lymphadenopathy with increased acute T-cell immune response, extraocular organ involvement with phosphatidylinositol-3-kinase (PI3K) activation pathways, chronic and multiorgan disease with increased immune response pathways, and multiorgan disease with increased IL-1 and IL-18 immune and inflammatory responses (50). We subsequently conducted exploratory analysis on their now publicly available data set by comparing RNA-sequencing data from untreated sarcoidosis subjects with stage IV disease (n = 12) versus their untreated counterparts with stage II or stage III disease (n = 40). We identified 178 genes that exhibited a fold change exceeding 2.0 with a Wilcoxon ranked-sum P value < 0.05. Unbiased Gene Ontology (GO) enrichment analysis of these 178 genes via MetaCore revealed a significant association with the signal transduction pathways shown in Table 1. This analysis is but one representation by which high throughput “omics” technologies applied to multicenter studies can be used as a hypothesis generating, bedside to bench approach. Continued development of well-phenotyped multicenter cohorts will be instrumental to advance precision medicine efforts in chronic sarcoidosis and its association with pulmonary fibrosis.

Table 1.

Top 10 Gene Ontology defined processes significantly enriched in the bronchoalveolar lavage fluid of patients with stage IV sarcoidosis

GO Processes	P Value	False Discovery Rate
Positive regulation of cell communication	4.846E-07	2.030E-04
Positive regulation of intracellular signal transduction	4.192E-07	2.030E-04
Negative regulation of adenylate cyclase activity	3.639E-07	2.005E-04
Regulation of synaptic glutamatergic transmission	2.611E-07	1.644E-04
Renal system development	1.896E-07	1.393E-04
Kidney development	1.179E-07	1.039E-04
Regulation of vesicle-mediated transport	1.041E-07	1.039E-04
Regulation of exocytosis	5.552E-08	8.158E-05
Multicellular organismal process	1.904E-08	4.197E-05
Regulation of regulated secretory pathway	2.498E-09	1.101E-05

GO, Gene Ontology.

MODELS OF SARCOIDOSIS

Addressing the knowledge gaps in sarcoidosis granuloma formation and its associated fibrosis will also require modeling systems that are practicable, reproducible, and disease-state relevant. Given the absence of a well-established unifying model for sarcoidosis, granuloma models have been used frequently instead to recapitulate the disease process. These models have mainly focused on granuloma formation as the field is lacking in models of granulomatous fibrosis (51). Several in vitro platforms have demonstrated varying degrees of efficacy in modeling granuloma formation. For example, early cell culture studies using purified monocytes, macrophages, and/or CD4+ T cells isolated from the blood or BALF of patients with sarcoidosis have provided some insight into the function of specific cell populations in a reductive environment (52). However, despite their ability to decipher some aspects of human immunology, these monocellular studies do not simulate the complexity of granuloma formation. Accordingly, the more recent multicellular systems using antigen-driven stimulation of peripheral blood mononuclear cells (PBMCs) obtained from patients with sarcoidosis impart increased significance to the disease being studied (53, 54). Two such models exist: one using purified protein derivative (PPD)-coated beads (53) and another involving microparticles derived from Mycobacterium abscessus (MAB) cell walls (54). Both models engender multicellular granuloma-like complexes comprised of abundant macrophages and lymphocytes (53) that alternately include Th1 (PPD) or Th17 cells (MAB). Although not directly comparable with sarcoidosis, which stems from a yet-to-be identified antigen, the benefits of these models include their recapitulation of cytokine and chemokine responses, which in both settings revealed significantly reduced IFN-γ release by sarcoidosis cells compared with control cells (53, 54). Interestingly, gene expression analysis of PPD-induced granulomas implicated IL-13, M2 macrophages, and Th2 responses (53, 55), whereas the MAB model displayed enrichment of Th1 and Th17 pathways (54). Although these models are promising in their ability to simulate some aspects of granuloma formation by sarcoidosis PBMCs, the contradictory findings highlight the need for a unifying model of this complex and, at times, intractable disease.

More recently, the application of organ-on-a-chip technology has facilitated a model that most accurately replicates the cellular microenvironment in human pulmonary sarcoidosis (56). In a departure from traditional two-dimensional (2-D) culture systems, this three-dimensional (3-D) microfluidic chip model recreated the architecture and components of the lung’s air-blood barrier, including epithelial cells and vascular endothelial cells, while also allowing for perfusion of the environment to mimic exposures. Here, mature granulomas formed from sarcoidosis PBMCs following exposure to MAB-based microparticles were transferred to the air-lung interface in the chip model (54). The advantages of this model in providing the structural aspect and mimicking the physiological stressors of the multicellular environment distinguish it from 2-D systems by capturing the dynamics involved in granulomatous pulmonary sarcoidosis and deciphering the discrete cellular contributions to this process.

These ex vivo models are both separate from and complementary to in vivo systems which most commonly involve presensitization of mice to putative antigens followed by repeated administration of the same antigens. The most widely studied methods have included administration of Propionibacterium acnes (PA) and mycobacterial antigens such as catalase-peroxide, superoxide dismutase A, and trehalose 6,6′-dimycolate (57). Noninfectious models include use of multiwall carbon nanotubes and beryllium, and mice deficient in genes for either apolipoprotein E and tuberous sclerosis complex 2 (57). Although these models, as well as others under development, appear to capture some aspects of human disease, the granulomas produced by these systems poorly reflect the cellular environment and local inflammatory milieu observed in sarcoidosis (7). In addition, these models are, for the most part, unable to simulate fibrotic development, thus limiting the study of the mechanisms that may drive progression of chronic granulomatous inflammation to fibrosis. Jiang et al. (58) attempted to address this knowledge gap by developing a granulomatous mouse model with pulmonary fibrosis using repeated intratracheal PA challenges. As seen in prior PA granuloma models, this model demonstrated granuloma formation following an initial PA challenge. However, the authors were able to induce the development of fibrotic lesions through the application of a second booster dose of PA intratracheally at 28 days, suggesting a potential relationship between chronic granulomatosis and fibrotic progression resulting from repeated antigenic stimulation (58). Although this model exhibits the exciting possibility for the study of fibrotic development in relation to granulomatous inflammation, further investigation is needed to model this association. Although reverse translational efforts have resulted in these as well as other sarcoidosis models (Table 2), given the existing limitations of current models, improved surrogates of human biology are required to operationalize the “bedside to bench” concept of reverse translational research.

Table 2.

Reverse translational models of sarcoidosis

Model	Description	Reference
Concanavalin A	Monocytes from patients with sarcoidosis cultured in media containing the supernatant of concanavalin A-stimulated peripheral blood mononuclear cells (PBMCs) lead to multinucleated giant cell formation	Mizuno et al. (52)
Mycobacterium bovis bacillus Calmette–Guérin (BCG)	PBMCs from patients with sarcoidosis cultured with BCG extract-coated sepharose beads induce granuloma-like cellular aggregates	Taflin et al. (30)
Purified protein derivative (PPD)	PBMCs from patients with sarcoidosis exposed to polystyrene beads coated with PPD of Mycobacterium tuberculosis induce granuloma-like cellular aggregates	Crouser et al. (53)
Mycobacterium abscessus (MAB)	PBMCs from patients with sarcoidosis exposed to microparticles derived from MAB cell walls induce granuloma-like cellular aggregates	Zhang et al. (54)
Lung-on-a-chip	Granulomas formed from the MAB model are added to a microfluidics system with a bilayer of normal bronchial epithelial cells and human microvascular endothelial cells	Calcagno et al. (56)

Ultimately, whether an infectious or inflammatory model is adopted, differences between mice and humans hamper disease relevance. This interspecies gap is being addressed by the development of “humanized mice” which have for over three decades been a source of intense research and development. First described in 1983, humanized mice contained a mutation in their DNA-dependent protein kinase gene (Prkdc), resulting in the severe combined immunodeficiency phenotype (Scid), which lacks both T and B lymphocytes, that is required for xenografting of human cells and tissues (59). However, widespread use of this mouse was limited by low engraftment rates, and as the mice aged, there was increased spontaneous (“leaky”) production of mouse B and T cells (59). These issues were abrogated to some extent when Scid mice were crossed with the nonobese diabetic (NOD) mouse (Scid-NOD) (59). However, the most significant breakthrough in this field resulted from the generation of mice deficient in the γ chain of the IL-2 receptor (IL-2RG^null), which is necessary for the binding and signaling of IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 (60). The IL-2RG^null mutation forms the basis of the three most commonly used immunodeficient mouse strains today: NOG, NSG, and BRG (61–63). NOG and NSG strains are generated through combinations with Scid-NOD mice, whereas the BRG strain is the product of combination with Scid-NOD mice that also contain mutations to the recombination-activating gene 2 (Rag-2). These three strains all lack B, T, and NK cells and to date, are most frequently used to recapitulate the human immune system in mice, offering the unprecedented opportunity to perform studies of human translational immunology in a unique in vivo model.

One obstacle with these humanized mice is that murine cytokines are not fully recognized by their respective human receptors, a limitation that is addressed by the use of transgenes or knock-in recombination has further improved differentiation of human immune cells (64). For example, the MISTRG mouse strain is a BRG mouse with human genes for macrophage colony-stimulating factor (M-CSF), IL-3, granulocyte-macrophage colony-stimulating factor (GM-CSF), signal regulatory protein α (SIRPα), and thrombopoietin (TPO) knocked into their respective mouse loci (65). These cytokines promote the development and function of myeloid cells and lymphocytes derived from human CD34+ progenitor cells, including macrophages and CD4+ lymphocytes (65). These mice have demonstrated the potential to be a faithful xenotransplantation model for hematologic malignancies such as myelodysplastic syndrome (66); for the study of sarcoidosis, these mice present the intriguing possibility for reverse translational research through the generation of patient-derived humanized mice using bone marrow biopsy-obtained CD34+ cells from patients with sarcoidosis. Such humanized mice would have the potential to be a new sarcoidosis model distinct from the current array of granuloma formation-centered models.

CONCLUSIONS

Exciting developments at the laboratory bench offer great promise for the study of poorly understood heterogeneous clinical phenotypes that define sarcoidosis (18). Observations gleaned from clinical trials and multicenter patient cohorts have generated new hypotheses related to granuloma formation and fibrotic progression. Application of next generation technologies and deployment of improved ex vivo platforms, combined with advances in in vitro modeling and humanized mice, will enable further study of disease pathogenesis and treatment. These efforts across the research spectrum will allow for reverse translation of clinical observations from the bedside to bench, and back again.

GRANTS

Research reported in this publication was supported by National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes of Health. A.L. was supported by NHLBI Grant No. T35HL007649 and the Richard K. Gershon Fellowship. L.S. was supported by the Parker B. Francis Foundation and American Lung Association Catalyst Award. C.S.D.C. was supported by NHLBI Grant No. R01HL126094, Veterans Affairs VA Merit Award (BX004661), and Department of Defense DOD Grant No. PR181442. E.L.H. was supported by NHLBI Grant Nos. R01HL109233, R01HL125850, and R01HL152677 and grants from the Foundation for Sarcoidosis Research, Gabriel and Alma Elias Research Fund, and the Greenfield Foundation. C.R. was supported by a Chest Foundation Grant, Boehringer-Ingelheim IPF/ILD Discovery Award, and NHLBI Grant No. K08HL151970-01.

DISCLAIMERS

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Heart, Lung, and Blood Institute or the National Institutes of Health.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.L., L.S., X.Y., C.S.D.C., and C.R. prepared figures; A.L., L.S., X.Y., C.S.D.C., E.L.H., and C.R. drafted manuscript; A.L., L.S., X.Y., C.S.D.C., E.L.H., and C.R. edited and revised manuscript; A.L., L.S., X.Y., C.S.D.C., E.L.H., and C.R. approved final version of manuscript.