Background
Rheumatoid arthritis (RA) is a chronic inflammatory disease of unknown etiology that affects about 1% of the population in industrialized countries [
1]. It is associated with disability, pain and significantly affects quality of life [
2]. If left untreated, RA can ultimately lead to joint destruction, systemic bone loss, increased risk of fractures and other comorbidities [
3,
4]. The pathogenesis of RA comprises a complex inflammatory response, involving macrophages, synoviocytes, T cells, B cells, proinflammatory cytokines and autoantibodies, causing joint damage and resulting in erosion of bone and cartilage [
5,
6]. Currently, there are well established therapeutic options for treating RA, namely Non-Steroidal Anti-Inflammatory Drugs (NSAIDs) and Disease Modifying Anti-Rheumatic Drugs (DMARDs) that include biologic agents targeting cytokines, T lymphocytes and B cells [
7‐
9]. More recently, targeting kinases with small molecule inhibitors for inflammatory disorders has been an area of intense focus for research [
10]. This has led to the approval of the oral DMARD Tofacitinib, a pan-Janus kinase inhibitor, for the treatment of moderate to severe RA [
11,
12].
The synovial joints of RA patients have higher levels of several proinflammatory cytokines and chemokines, the most predominant of which are: TNF, IL-1β, IL-6 and MCP-1 [
13‐
16]. TNF is a proinflammatory cytokine that plays a central role in the pathogenesis of RA, resulting in destruction of bone and cartilage [
17,
18]. Therapies targeted toward neutralizing TNF have shown substantial efficacy in RA patients and in preclinical models as well [
19,
20]. Although anti-TNF therapy is a preferred strategy for the treatment of RA, some patients do not respond to anti-TNF treatment, whereas others lose the initial response over time [
21]. One of the reasons for the loss of efficacy could be attributed to the generation of anti-drug antibodies (ADA) to the anti-TNF agent, that might reduce or neutralize the therapeutic effect [
22].
In RA, the release of numerous proinflammatory mediators such as cytokines IL-6, IL-1β and TNF, result in an increased sensitivity to pain [
23,
24]. It has also been shown that thresholds for pain and pressure are decreased in the affected joints of patients with RA [
25]. The manifestation of pain is a result of both excitatory and inhibitory signals that are processed by higher brain centers [
26]. Most anti-rheumatic therapeutics are effective in controlling inflammation, however further investigation is required, in order to identify novel anti-rheumatic agents that can simultaneously inhibit inflammation and pain [
27]. The impact of treatment for rheumatoid arthritis is typically evaluated using American College of Rheumatology (ACR) scores, of which one of the critical components is the evaluation of pain in addition to inflammation [
28]. Furthermore, Heiberg et al. [
29] suggested that an improvement in pain reduction was an area of high priority and managing pain can significantly improve the quality of life for RA patients. The reduction of inflammation alone is only a partially effective strategy in the treatment of RA, as patients still present with chronic joint pain [
23]. Consequently, management of pain is critical for any effective treatment paradigm for RA [
27].
Animal models of Rheumatoid Arthritis have played a major role in our understanding of the mechanisms of disease pathophysiology and have supported drug discovery leading to identification of novel therapies [
30‐
32]. Preclinical models of arthritis share many immunological, clinical and histological characteristics with human RA, however, none of them capture all the facets of the human disease [
33]. Several preclinical models of arthritis such as Adjuvant Induced Arthritis (AA) and Collagen Induced Arthritis (CIA) are widely used in drug discovery [
30]. These models are poly-arthritic, involving multiple joints and the disease phenotype is chronic and progressive unlike the flares and remissions observed in RA [
34]. Streptococcal Cell Wall (SCW) is a rodent model of arthritis that effectively captures repeated remission and flaring phenotype, similar to RA [
35]. A single intra-peritoneal injection of SCW extract PeptidoGlycan-PolySaccharide (PG-PS 10s) induces inflammation in peripheral joints with repeated phases of self-reactivating flares resembling RA [
36]. However, the recurrence of reactivation is unpredictable and often difficult to control, hence this model was modified by Schwab et al. [
37], in order to synchronize the flares. The modified SCW model is induced by a local intra-articular (i.a.) injection of SCW extract PG-PS 100p in the hind tarsus (flare 1) followed by a systemic intravenous (i.v.) challenge (flare 2). The model is characterized by a mono-arthritic multi-flare phenotype of two distinct remissions and flares. Inflammation is limited only to the sensitized joint with no detectable involvement of other joints, unlike other preclinical models of arthritis [
38]. Demonstrating efficacy in animal models of pain is an important step in identifying novel anti-rheumatic agents that can effectively target inflammation and pain in the clinic [
23]. Therefore, in addition to inflammation we have evaluated paw withdrawal threshold as a surrogate for pain. We applied the established von-Frey assay, previously described for mechanical pain assessment in preclinical animal models and in RA patients alike [
25,
39]. Although, evaluation of inflammation and pain have been previously reported in other preclinical models of arthritis [
40‐
42], to our knowledge, this is the first report investigating the clinically relevant readouts such as pain and inflammation simultaneously in the SCW model. Furthermore, we extended the model by inducing an additional flare by re-challenging the rats with an additional systemic intravenous injection of SCW (flare 3). This leads to a more persistent inflammation and mechanical pain in the previously sensitized joint. In addition, we evaluated the impact of anti-TNF, corticosteroid and analgesic therapy using etanercept, dexamethasone and buprenorphine to understand the translatability of this model in a clinical setting. The mechanisms leading to pathogenesis in the model were further delineated by histopathological evaluation, cytokine profiling, cell phenotyping and bioluminescence imaging of the arthritic joint.
Here we report distinct temporal profiles of inflammation and mechanical pain in the rat SCW model that have not been previously described. Moreover, we show that TNF could potentially be a key driver of inflammation and could in part contribute to the onset of pain in this model.
Methods
Animal use and care
Female Lewis rats (6–8 week old; Harlan Laboratories, Indianapolis, Indiana, USA) were acclimated for 5 days prior to the experiments and were housed under standard conditions. Female Lewis rats were used in all of these experiments due to their established susceptibility to various mediators of inflammation [
43]. These experiments were conducted in accordance with federal animal care guidelines and all procedures were reviewed by the Institutional Animal Care and Use Committee (IACUC) of Merck Inc.
Induction and assessment of the SCW model
For the induction of SCW arthritis we modified the protocol that was originally described by Schwab et al. [
37]. For initial model development studies the rats were allocated to four different groups and were administered with either saline (non-arthritic; negative control), 2.5 μg, 5 μg or 10 μg of SCW extract PeptidoGlycan-PolySaccharide (PG-PS) 100p (BD Biosciences, Franklin Lakes, NJ, USA) by intra-articular (i.a.) injection into the tarsal joint to induce flare 1. On day 21, three weeks after the initial i.a. injection, rats were challenged intravenously (i.v.) with either saline (non-arthritic; negative control) or 100 μg PG-PS 100p (SCW) to induce flare 2. Based on the data from our initial studies the 5 μg dose of SCW via i.a. injection followed by 100 μg of SCW via i.v. injection was used for all subsequent experiments. In later experiments, an additional third flare (Flare 3) was induced by re-challenging the rats on day 42 with an i.v. injection of 100 μg of SCW. Inflammation and mechanical pain were assessed by measuring ankle diameter (surrogate for inflammation) using precision mechanical calipers and withdrawal threshold (surrogate for pain) using electronic von-Frey assay through the course of study.
Withdrawal threshold: electronic von-Frey for assessing mechanical pain
Electronic von-Frey (Somedic Sales AB, Horby, Sweden) analysis was performed using methods described previously [
39]. The rats were placed on an elevated grid rack under individual polycarbonate cages allowing easy access to the plantar foot surface. The tip of the von-Frey probe was brought up gently to touch the center of the paw plantar surface, and pressure was gradually increased perpendicularly at a rate of approximately 5 g/sec. The von-Frey probe records increasing pressure (grams) values, until a paw withdrawal reaction by the rat was observed. The mean of two consecutive responses per rat was recorded and used for data analysis.
Dosing paradigm in flare 2
SCW induced arthritic rats were randomly assigned to specific treatment groups and their baseline ankle diameter and withdrawal threshold values were recorded. Test article interventions were by oral gavage or subcutaneous injection either in a Prophylactic (P) or Therapeutic (T) regimen. In the prophylactic treatment regimen, compounds were administered once daily for 10 days starting on day 20 (1 day prior to SCW intravenous challenge) ending on day 29. The therapeutic treatment regimen entailed compound administration once daily for 8 days starting on day 22 (1 day post SCW intravenous challenge) ending on day 29. Non-arthritic and SCW control rats received vehicle (PEG 400:10% Tween 80 [1:9]) orally. Etanercept (Enbrel; Amgen Inc., Thousand Oaks, CA, USA) was purchased from Myoderm Limited, Norristown, PA, USA and was reconstituted in bacteriostatic water as per manufacturer’s instructions. Etanercept (subcutaneous (s.c.); 0.25 or 1 mg/kg/day) was administered either in prophylactic or therapeutic regimens. Rats in the dexamethasone group (Sigma Aldrich, St. Louis, MO, USA) received dexamethasone (per oral (p.o.); 0.3 mg/kg/day) suspended in (PEG 400:10% Tween 80 [1:9]). Buprenorphine (Sigma Aldrich) groups received buprenorphine (p.o.; 0.05 mg/kg/day) suspended in saline in the therapeutic regimen. The dosing of compounds for all groups was stopped on day 29.
Dosing paradigm in flares 2 and 3
In certain experiments an additional flare (Flare 3) was induced in all rats following flare 1 and flare 2. The rats were randomly assigned to two cohorts, cohort 1 and cohort 2, prior to compound administration. In cohort 1, SCW induced rats were assigned to one of the following treatment groups: etanercept (s.c.; 1 mg/kg/day), human IgG1 isotype control (s.c.; 1 mg/kg/day) or dexamethasone (p.o.; 0.3 mg/kg/day) and were dosed in the prophylactic regimen from day 21 to day 29 in flare 2. Subsequently, these rats had a drug washout period of 14 days and were treated with same doses of etanercept (s.c.; 1 mg/kg/day), human IgG1 isotype control (s.c.; 1 mg/kg/day) and dexamethasone (p.o.; 0.3 mg/kg/day) from days 41 to 51 (flare 3). In cohort 2, the rats were treated with etanercept (s.c.; 1 mg/kg/day), human IgG1 isotype control (s.c.; 1 mg/kg/day) or dexamethasone (p.o.; 0.3 mg/kg/day) in flare 3 only. Non-arthritic and SCW control rats received vehicle (bacteriostatic water) as a subcutaneous injection.
In vivo BioLuminescence imaging (BLI)
Bioluminescence Imaging was performed using IVIS Spectrum imaging system (Perkin Elmer, Waltham, MA, USA) following methods described previously [
44,
45]. Drug naïve non-arthritic controls and SCW induced rats were injected subcutaneously with a single injection of 200 mg/kg of Luminol (Sigma Aldrich) suspended in phosphate-buffered saline (PBS). Image analysis was performed using Living Image 4.0 (Perkin Elmer, Waltham, MA, USA), and average radiance (photons/second) was measured by placing a circular region of interest (ROI) centered over the SCW or non-arthritic hind tarsal joint with a second ROI placed over the contralateral tarsal joint for comparison.
Histopathology
All the rats designated for histomorphologic assessment (post-flare 2 only) survived to scheduled study termination. At necropsy, left hindlimbs were excised distal to the hip, and fixed in 10% Neutral Buffered Formalin. The knee (with attached distal femur and proximal tibia) and hind paw (with ankle, distal tibia, tarsal bones, metatarsal and phalangeal joints) were decalcified in Immunocal (Decal Chemical Co, Suffern, NY, USA) for approximately 20 hours (knee) or 51 hours (hind paw), trimmed longitudinally (midsagittal) and placed back in decalcification solution for 2 hours (hindpaws only). After washing, the knee and hind paws from each rat were processed, sectioned, and stained with Hematoxylin & Eosin for subsequent microscopic evaluations. A veterinary pathologist scored the sections of knee and hind paws for inflammation, pannus formation, cartilage destruction, periosteal bone formation, and/or bone resorption, using severity grades: 1 = very slight, 2 = slight, 3 = moderate, 4 = marked, 5 = severe. The histomorphologic assessment was subsequently submitted for peer review by a second veterinary pathologist.
Cytokine analysis
Ankle tissue (whole joints including synovium, bone and surrounding tissues) were excised post euthanasia and flash frozen in liquid nitrogen. The ankle tissue was pulverized and treated in a co-mixture of cell extraction solution, phosphatase and protease inhibitors. The tissue was homogenized and the supernatant was processed for cytokine expression, using Milliplex MAP 27-plex rat cytokine/chemokine magnetic bead panel (EMD Millipore, Billerica, MA, USA) on a Luminex FlexMAP 3D instrument (Luminex, Austin, TX, USA) following manufacturer instructions. Values for samples below the lower limit of the standard curve were set at the value of the lowest standard for analysis. All samples were run in duplicate or triplicate and the values are reported for technical replicates.
Cellular phenotyping
Lymph nodes or ankle tissues were pooled for each group, prior to flow cytometry analysis. Cells from the lymph nodes were mechanically harvested to generate a single cell suspension in media. Ankle tissue was minced and digested with 0.44 U/mL Liberase enzyme and 9 U/mL dnase and made into a single cell suspension. All the antibodies (clone name in parenthesis) were purchased from BD Biosciences unless noted otherwise. The entire lymph node and ankle tissues were digested in 1800uL of the buffer, 80 uL of which was analyzed by flow cytometry to ascertain the cell population distribution in the lysate. The absolute numbers were quantified by back calculating to account for the whole lysate. Cells were blocked with anti-CD32 (D34-485) Fc block, followed by addition of either of two staining antibody cocktails: cocktail 1 containing CD45 V450 (OX-1), CD45RA FITC (OX-33), CD3 APC (1F4), CD4 pecy7 (OX-35), CD8 percp (OX-8), and cocktail 2 containing CD45 V450, CD172 PE (OX-41), anti-granulocyte FITC (HIS48), CD4 peCy7, CD163 Alexa Fluor 647 (ED2) (AbD Serotec, Kidlington, Oxford, UK). Total T cells were derived from CD45+, CD3+, CD45RA- cells from the cocktail 1 staining condition. This population was further analyzed for CD4+ and CD8+ cells. Neutrophils were derived from CD45+, anti gran hi, CD172+, CD163- cells from the cocktail 2 staining condition. T cell and neutrophil counts were analyzed relative to the starting leukocytes counts (CD45+ cells) for each assay.
Compound exposure (PK) of etanercept
Circulating levels of etanercept in rat serum were ascertained on the Gyrolab xP instrument (Gyros AB, Uppsala, Sweden) equipped with Bioaffy 200 CD. The capture antibody used was Biotinylated-anti-TNFaRII mouse IgG2a (Clone # 22235) (R and D Systems, Minneapolis, MN, USA) and the detection antibody used was DyLight 650-conjugated anti-human IgG1 Fc Rabbit monoclonal antibody (Clone # H26-10) (Abcam, Cambridge, MA, USA). Standard curve was prepared in 50% naïve female lewis rat serum in Rexxip A (Gyros AB) with the linear range from 0.457 to 1000 ng/ml. Etanercept was quantified with Gyrolab Evaluator software.
Data analysis
Data were analyzed and plotted using Graphpad Prism 5 (GraphPad Software Inc., La Jolla, CA, USA). Percentage inhibition for individual treatment groups are calculated by normalizing Area Under the Curve (AUC) for paw swelling and withdrawal threshold by fitting SCW vehicle groups to 100% and non-arthritic controls to 0%. AUC normalization was performed using the formula: ((Treatment – Non-Arthritic)/(SCW – Non-Arthritic)) * 100. AUC was calculated over time for flare 2 from days 21–29 and for flare 3 from days 42 to 51. Statistical significance (P value <0.05) was determined by two way analysis of variance for inflammation and mechanical pain analysis or by one way analysis of variance for cytokine and Bioluminescence imaging analysis followed by Bonferroni post-tests. Comparisons were made for all drug treated groups versus non-arthritic treated with appropriate vehicle (represented by *) or SCW treated with appropriate vehicle (represented by ^) groups. All values are expressed as mean ± SEM, unless otherwise noted. Relative change of ankle diameter and withdrawal threshold in SCW injected rats as compared to non-arthritic controls is represented by Δ.
Discussion
RA is characterized by a complex interplay of various pathogenic mechanisms leading to inflammation and pain [
47]. Animal models of arthritis can be effective tools to investigate these mechanisms and delineate pathways that might help predict the successful outcome of novel therapeutics in the clinic. The previously described mono-arthritic SCW model in rats captures the relapsing and remitting flares of the disease, similar to RA [
35]. Typically, the model is characterized by two flares that resolve over time [
37,
48]. Intra-articular injection of SCW antigen to local joint results in mono-arthritis with inflammation limited only to the sensitized joint, even after systemic challenge with the antigen. The first flare induced by an intra-articular injection of SCW results in mild paw swelling that peaks 24 hrs post sensitization (flare 1) and resolves over 72 hours. The second flare induced by an intravenous challenge with SCW (typically three weeks later) results in a pronounced onset of paw swelling reaching peak on day 3 after i.v. challenge (flare 2) and resolves over a 5 day period. Previous studies by Schimmer et al. [
46], showed that the early phase of the model can be triggered by Th2 cells and neutrophils. They have also demonstrated that neutrophils were involved in addition to T cells in the reactivation of flares. In addition, the model is dependent on multiple proinflammatory cytokines including TNF and IL-1, as assessed by specific anti-cytokine therapy and gene expression analysis [
49,
50]. In order to capture the chronic disease phenotype, we have developed an extended version of the model with the induction of an additional flare (flare 3). The third flare appears to exhibit a chronic disease phenotype not previously described. The mono-arthritic SCW model is a robust model with reported incidence of arthritis in 90-100% of the rats [
34]. In line with literature, we were successful in inducing arthritis in 100% of the animals in all the three flares.
Histological assessment demonstrated that inflammation, pannus formation, degeneration of cartilage, periosteal bone formation, and/or bone resorption in SCW rats was limited to the sensitized hind paw, there were no histomorphologic findings in the proximal hind limb joint tissues and the contralateral hind limb and hind paw. We also observed that the gross histological damage in the model was mild to moderate compared to CIA or AA models and this is in agreement with current literature [
34,
48]. An additional attractive feature of this model, is that the disease severity and progression is less severe when compared to other rodent models of arthritis [
34,
38]. We observed that even in the extended flare 3, the disease was mild and the rats maintained a healthy status throughout the course of the study.
Neutrophils have been shown to play a major role in both flare 1 and 2. Depleting neutrophils by an anti-neutrophil antibody has been shown to be efficacious in the model [
49]. Our results show the presence of neutrophils in flare 1 as determined by flow cytometry and in flare 2 as assessed by histology. In addition to neutrophils, T cells also have been shown to contribute to the pathogenesis of disease [
38]. 24 hr post induction of both flare 1 and 2, we observed an up regulation of T cells in the ankle joint. The infiltrating T cells in the local joint belonged to the CD4
+ subpopulation, however CD8
+ T cells were not detected at this time-point. These findings were also corroborated by histological analysis of samples collected on day 29. A two fold increase in T cells was observed in the draining lymph node, 24 hr post induction of flare 2 compared to flare 1 and non-arthritic controls (Table
1). We further characterized the production of multiple proinflammatory cytokines and showed increased levels of IL-1β, IL-6, MCP-1, CINC-1 in the local joint. The role of TNF, IL-1β, IL-6, MCP-1 and CINC-1 in the pathogenesis of SCW has been demonstrated previously [
50,
51]. Moreover, neutralizing antibodies to TNF, IL-1α/β and MCP-1 have been shown to inhibit inflammation, thereby further confirming the role of these cytokines in the model [
46,
49]. Although, we expected to detect TNF in the arthritic joint homogenates, this cytokine could not be quantified perhaps due to its transient kinetics or the sensitivity of the assay. The presence of IL-1β and IL-6 in the arthritic joint of SCW rats and the role of these proinflammatory cytokines in inducing pain in addition to inflammation in RA, led us to investigate the kinetics of pain in our model [
15,
24,
51].
In RA, pain experienced by patients involves the complex interplay of nociceptive and inflammatory processes [
23]. Changes in joint pathology often result in an increased sensitization of primary sensory neurons and central sensitization, due to the changes in ascending or descending modulatory pathways [
52]. Physical changes in the rheumatic joint (tissue edema, biochemical changes, inflammatory mediators, nociceptor activation) all can cause a decrease in the threshold for pain [
53]. Furthermore, pain is an important clinical feature of RA and its assessment is incorporated in the ACR scoring paradigm [
28]. The mono-arthritic SCW model allows us to concurrently assess pain and inflammation following drug intervention, facilitating the discovery of novel anti-rheumatic agents. Simultaneous assessment of inflammation and pain has been previously described in the polyarthritic mouse CIA model [
54]. However, the distinctive mono-arthritic phenotype of the SCW model offers an advantage over the poly-arthritic models, as we were able to monitor mechanical pain and inflammation in the sensitized joint, with the contralateral paw serving as an internal negative control.
To further characterize our model, we profiled a panel of therapeutic agents such as, etanercept, dexamethasone and buprenorphine. We selected these agents based on their established efficacy in other preclinical models of arthritis [
55‐
57]. The specific doses used for each drug in these studies were selected based on multiple dose response studies conducted internally (data not shown). In flare 2, treatment with anti-TNF agent, etanercept resulted in a significant inhibition of paw swelling in both prophylactic and therapeutic regimens. Interestingly, prophylactic treatment with etanercept showed a modest decrease in mechanical pain, which was statistically significant. However, therapeutic treatment of etanercept had no effect on pain. Our data suggests that TNF may play a bigger role in the onset and progression of paw swelling, whereas other cytokines such as IL-1β, IL-6, or other mechanisms in addition to TNF, could also contribute to pain. Currently, efforts are underway to further understand the roles of these mechanisms in the multi-flare model pathophysiology. Therapeutic treatment with corticosteroid dexamethasone was more effective in inhibiting paw swelling and mechanical pain compared to etanercept. Prophylactic administration of analgesic buprenorphine significantly inhibited pain but had no effect on inflammation. Taken together, these results suggest that inflammation and pain can be distinguished and evaluated separately in the SCW model. Histomorphological evaluation of the local joints corroborated with the efficacy data, showing reduced severity of inflammation and pannus formation after treatment with etanercept and dexamethasone (data not shown). Our internal data also show similar efficacy profiles of etanercept and dexamethasone in the rat CIA model on inflammation (data not shown).
We further extended the model to capture the chronic phase of the disease (RA), by inducing an additional flare via a second systemic antigen challenge of SCW, characterizing both inflammation and pain. We observed that inflammation and pain in flare 3 did not resolve by study termination (10 days post induction of flare 3) resulting in what appears to be a chronic phenotype. Interestingly, the cytokine expression profile in the flare 3 was similar to that observed in flare 2 with an up-regulation of IL-1β, IL-6, MCP-1 and CINC-1.
The efficacy of etanercept in the chronic phase of the model was also investigated. As observed in flare 2, prophylactic treatment in flare 3 with etanercept resulted in a significant inhibition of paw swelling, with a modest impact on pain which was statistically not significant. This data suggests that TNF may play similar role in flare 3. However, further investigation is required to delineate additional mechanisms that may be involved in the regulation of pain in this model.
Interestingly, the cohort of SCW treated rats that had received etanercept treatment in flare 2, when treated with the same dose of etanercept in flare 3, did not respond to the treatment. The lack of efficacy observed in flare 3 was attributed to the reduced systemic exposure of etanercept in these rats most likely due to the immunogenicity. Immunogenicity toward certain biologics, especially anti-TNF agents has been demonstrated in rodents as well as in the clinic [
58,
59]. In addition, multiple dosing of the agent may also contribute to the generation of anti-drug antibodies [
59]. Administration of anti-TNF alone in RA patients can elicit autoantibodies resulting in an enhanced clearance and loss of efficacy of the agent [
60]. Species specificity may also contribute to the immunogenicity of biologic agents, as observed with certain chimeric antibodies [
61,
62]. Interestingly, co-administration of methotrexate (MTX), along with certain anti-TNF agents, can reduce the immunogenicity and improve the efficacy of these agents in the clinic [
63]. Although the mechanisms involved in the immunogenicity of biologic agents are not fully elucidated, anti-TNF in combination with MTX is considered to be a gold standard therapy for moderate to severe RA patients [
64]. Hence, it will be interesting to test etanercept and MTX combination therapy compared to etanercept treatment alone in flare 3. The flaring mechanism in the SCW model allows for drug washout periods in between compound administration. This might provide useful preclinical insights on potential immunogenicity mechanisms that may be relevant in a clinical setting.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
KC, RF, RS, GR, JS, MZ, MJC, AB, EL, SH, SZ, JZ-H, LYM, RLM, DS, WZ, AM and MC are full-time employees of Merck & Co Inc. and/or own stock and/or stock options in Merck & Co Inc. C-SC is a former employee of Merck & Co Inc. and/or owns stock and/or stock options in Merck & Co Inc.
Authors’ contributions
KC, RF, C-S C, J Z-H, WZ, AM and MC conceived the study design and participated in its coordination. KC, RF, GR, RS, JS, MZ, AB, EL and MJC contributed to study execution. KC, RF, GR, RS, SH, SZ and DS were involved in data acquisition. Analysis and interpretation of data were performed by KC, RF, J Z-H, WZ, AM, and MC. KC, RF, RLM, LYM, DS, AM and MC conducted the manuscript preparation and critical review. Statistical analyses were done by KC and RF. All authors contributed to revising the manuscript critically for important intellectual content, and read and approved the manuscript for publication.