Temporally complex inflammatory networks in an animal model reveal signatures for interstitial cystitis and bladder pain syndrome phenotype

Interstitial cystitis and bladder pain syndrome (IC/BPS) presents with symptoms of debilitating bladder pain and is typically a diagnosis of exclusion. The cystoscopic detection of Hunner's lesions increases the likelihood of detecting tissue inflammation on bladder biopsy and increases the odds of therapeutic success with anti‐inflammatory drugs. However, the identification of this subgroup remains challenging with the current lack of surrogate biomarkers of IC/BPS. On the path towards identifying biomarkers of IC/BPS, we modeled the dynamic evolution of inflammation in an experimental IC/BPS rodent model using computational biological network analysis of inflammatory mediators (cytokines and chemokines) released into urine. The use of biological network analysis allows us to identify urinary proteins that could be drivers of inflammation and could therefore serve as therapeutic targets for the treatment of IC/BPS.

following CYP treatment.This pattern of inflammation may mimic the fluctuating severity of inflammation associated with IC/BPS flares.
Conclusions: Computational analysis of inflammation networks in experimental IC/BPS analysis expands on the previously accepted inflammatory signatures of IC by adding IL-5, IL-18, and MCP-1 to the prior studies implicating IL-6 and GRO as IC/BPS biomarkers.This analysis supports a complex evolution of inflammatory networks suggestive of the rise and fall of inflammation characteristic of IC/BPS flares.
biological network analysis, interstitial cystitis and bladder pain syndrome, inflammation

| INTRODUCTION
Interstitial cystitis/bladder pain syndrome (IC/BPS) is a debilitating chronic inflammatory disorder of the bladder that is believed to affect as many as 1 in 4 or 5 women in the United States. 1,2This serious disorder typically presents with nonspecific symptoms such as either increased urgency, frequency, or pain. 3IC/BPS is also co-morbid with multiple chronic inflammatory diseases including but not limited to endometriosis, irritable bowel syndrome, and fibromyalgia. 1,4C/BPS is a challenging diagnosis of exclusion due to the heterogeneity of clinical presentation and the absence of easily accessible and noninvasive confirmatory testing.While the role of inflammation in IC/BPS is not at the forefront of the IC/BPS field, the cystoscopic confirmation of Hunner's lesions dramatically raises the probability of detecting inflammation on bladder biopsy and the dependent therapeutic success of anti-inflammatory drugs in IC/BPS patients. 5However, the invasive nature of cystoscopy and biopsy motivates the search for noninvasive surrogates such as inflammatory markers (cytokines and chemokines), that can be detected in the urine.It is plausible that the heterogeneity in symptom presentation and severity of IC/BPS may be better captured by a signature panel of inflammatory mediators instead of a single inflammatory mediator.
So far, attempts to identify urinary or plasma biomarkers in IC/BPS patients have been limited to largely cross-sectional studies, [6][7][8][9][10] which reported elevations of nerve growth factor, IL-6, VEGF, MCP-1, and MIP-1 in the urine. 6,7,10][8][9][10] We posit that a fresh perspective examining the evolution of the inflammatory cascade characteristic of IC/BPS is needed to identify a panel of inflammatory mediators for diagnostic use. 11ological network analysis encompasses a set of algorithms that are useful for studying the evolution of inflammation across tissues over time. 12We hypothesized that biological network analysis, specifically time-interval principal component analysis (TI-PCA) and dynamic network analysis (DyNA), can be used to characterize the evolution of the bladder-centric inflammatory cascade characteristic of IC/BPS through urine analysis. 12We tested this hypothesis by studying the dynamic trajectories of 13 different inflammatory mediators (cytokines and chemokines) in rats with bladder centric phenotype of IC/BPS, induced by cyclophosphamide (CYP), a well-accepted model for eliciting the inflammation characteristic of bladdercentric IC through irritation of the urothelium. 13,14Here, we used TI-PCA and DyNA to identify urinary proteins that could be causative drivers of, and not just associated with, inflammation for fulfilling the attributes of a valid diagnostic and prognostic biomarker as well as a potential therapeutic target for IC/BPS. 11

| MATERIALS AND METHODS
All of the primary data on experimental design, animal use, and inflammatory mediators presented in this manuscript were reported previously. 15The development and advancement in biological network analysis provided a unique opportunity to revisit the previously reported data set and to analyze inflammatory biomarkers of IC/BPS with a fresh perspective.

| Animals
CYP was used to induce acute inflammation characteristic of inflammation associated with bladder-centric IC/ BPS in female Sprague-Dawley rats. 16Throughout the text, we refer to IC/BPS in rodent models in the context of acute inflammation induced by CYP characteristic of IC/BPS flares.Rats (276−292 g) received 150 mg/kg of CYP delivered intraperitoneally (n = 8).Baseline urine samples were collected from experimental and control (vehicle-treated, n = 8) rats 24 h before CYP injection.Following CYP injection, urine specimens from each hourly void were collected immediately, frozen in liquid nitrogen, and stored at −80°C as described previously. 15xperimental procedures were performed as approved by the University of Pittsburgh Institutional Review Board and University of Pittsburgh Institutional Animal Care and Use Committee (animal protocol number: 0605931).

| Statistical analyses
Analysis of Variance (ANOVA): A one-way ANOVA and Dunnett's correction for multiple comparison test for significant differences was conducted using Matlab® to identify differences in mean inflammatory mediator concentration between baseline and time points following CYP treatment (data was binned so that time intervals were 2 h long). 15ime Interval Principal Component Analysis (TI-PCA): TI-PCA is a modified PCA that highlights inflammatory mediators that contribute the most to variance in treatment response across consecutive time points. 12The raw inflammatory data were normalized for each inflammatory mediator so that all mediators were on the same scale (0−1).To identify mediators of inflammation that varied the most over a dynamic time interval (two consecutive time points) following CYP treatment, the principal components for each mediator were calculated using Matlab®.Principal components that captured at least 95% of the variance in the data were considered for further analysis.The product of the coefficient of each mediator (weight) and the mediator's eigenvalue for each corresponding principal component was calculated for all principal components.The product of the coefficient for each mediator and the eigenvalue for each principal component represents the amount of variance in the data that the principal component captures.The sum of all variances across principal components for each inflammatory mediator was calculated and the resultant TI-PCA bar graph is organized from greatest sum to least sum of variances.
DyNA: DyNA is a network inference method that identifies correlations among inflammatory mediators over two consecutive time points. 12Nodes are visually depicted as each of the 14 inflammatory mediators quantified by the Luminex™ assay.Mediators with an absolute correlation coefficient > 0.95 are connected by an edge.The network complexity of each graph structure is calculated as number of nodes number of edges maximum possible edges between all nodes × .Increases in network complexity indicates greater coordination amongst mediators in response to the treatment.

| RESULTS
Prior data analysis on this data set has been reported and has identified significant elevations in urine levels of IL-1α, IL-1β, IL-5, IL-6, IL-10, IL-18, and GM-CSF (p < 0.05) at 4 h following CYP-injection compared to baseline. 15No rats were excluded from the study as there was no significant difference in urine cytokine levels between CYP-treated rats at baseline and CYP-vehicle-treated rats. 15While IL-18 demonstrated a significant, 10-fold increase from baseline at 4 h, MCP-1 levels exhibited an 8-fold decrease 2 h after CYP-treatment. 15All data was collected from rat urine in a bladder-centric model of IC/BPS as induced by CYP.CYP is known to induce acute bladder inflammation characteristic of IC/BPS with Hunner's lesions. 16o define which of these mediators might be primary characteristics of this multifaceted inflammatory response, TI-PCA was next used to identify inflammatory mediators that contribute the most to variance in the inflammatory response over a dynamic time interval.For this analysis, we defined dynamic time intervals to be two consecutive time points.When analyzed holistically, the early phase of the inflammatory response (150-330 min) is characterized by inflammation mediated by GRO-KC, MCP-1, IL-18, and IL-5 (Figure 1A TI-PCA indicated that key inflammatory signatures of the CYP-induced inflammation in the urine at very early time points include GRO-KC, MCP-1, IL-18, and IL-5.Analysis of the TI-PCA graph in the first time interval (150-210 min) showed that, GRO-KC, IL-18, and IL-5 accounted for the greatest amount of variance amongst the inflammatory mediators in that time interval (Figure 1A).MCP-1 emerged as a key contributor to variance in the data set in the interval between 210 and 270 min (Figure 1B).
During the intermediate experimental interval between 330 and 450 min, IL-6 accounted for the most amount of variance across the entire interval.During the latter part of this interval, between 450 and 510 min, IL-6 was dominant and accounted for 27% of the variance in As an alternative means for inferring inflammation programs, DyNA was next used to identify correlations among inflammatory mediators across dynamic time intervals.While TI-PCA is useful for identifying mediators whose response is highly time dependent, DyNA is useful for understanding both positive and negative relationships among inflammatory mediators over narrow windows of time.Therefore, this analysis can be used for defining the process of propagation versus resolution of inflammation (Figure 2A−I).During the intervals between 150 and 270 min, IL-1β and IL-4 were correlated positively with each other (Figure 2A).Specifically, between 210 and 270 min, IL-1β was correlated positively with both IL-4 and TNF-α (Figure 2B).Network complexity rose and ultimately reached the first peak between 390 and 450 min (Figure 3).In this interval, GM-CSF was the most highly connected node and was correlated positively with IL-18, IL-10, and IL-2.Notably, IL-6, identified as a potent driver of inflammation by TI-PCA was not correlated with any other inflammatory mediators between 390 and 450 min (Figure 2E).
Following the first peak in network connectivity indicating a coordinated and propagated inflammatory response, there was a massive drop in network connectivity between 450 and 510 min (Figure 3).During this interval, there were no significant correlations between any two inflammatory mediators (Figure 2F).Network connectivity reached a true peak between 570 and 630 min.IL-1β and IL-18 have the greatest number of connections with IL-10 and GM-CSF having the second greatest number of connections (Figure 2H).Notably, IL-6 and IFNγ were connected with each other, but neither mediator shared connections with any other mediator.By 630-690 min, IL-5, IL-6, and MCP-1 were all connected with each other (Figure 2I).

| DISCUSSION
][8][9][10] Recently, symptoms of IC/BPS patients were combined with their urinary and plasma biomarkers as input into machine learning and diagnostic algorithms, but these algorithms 17 will need to be tested in well characterized IC/BPS patients in the future.The challenges facing the development of biomarkers or diagnostic guidelines for IC/BPS are perhaps linked to an inadequately characterized IC/BPS pathogenesis.
Since painful symptoms of IC/BPS are often relapsing and remitting, akin to autoimmune chronic inflammatory diseases such as inflammatory bowel syndrome, it is logical to expect time dependent fluctuations in urine levels of inflammatory mediators. 11Here, we tested the premise that the inflammatory cascade characteristic of IC/BPS flare can be modeled by a careful analysis of the dynamic acute inflammatory response to CYP in animal models because severity of inflammation after CYP mimics the severity of inflammation in flare up of IC/ BPS patients.It is worth noting that urine levels of paracrine messengers (cytokines/chemokines) known to orchestrate inflammation in the bladder are dependent on inflammation severity, 10,18,19 and urine levels of cytokines/chemokines only begin to be elevated from baseline levels after the cognate receptor binding sites in bladder tissue become saturated.Urine levels of inflammatory mediators are also dependent on voiding frequency, bladder distention, and the volume of each void which can cause the concentration of cytokines/ chemokines spilled over into urine from bladder to become diluted. 20While acute flares of IC/BPS in humans last 3−14 days on average, prior works on this data set shows that the 12 h timespan following acute CYP challenge adequately capture the onset and resolution of acute inflammation in a rodent model analogous to an IC/BPS flare. 2,15he etiology of bladder-centric and extra-bladdercentric phenotypes of IC/BPS is multi-factorial.The use of a single-dose intraperitoneal injection of CYP in rodents is the most robust model for studying bladdercentric IC/BPS, as this experimental paradigm induces hallmark Hunner's lesions. 16TI-PCA highlights important variance in inflammatory mediator concentration over time that was not identified by a rudimentary ANOVA.At the earliest time interval in the TI-PCA, GRO-KC/CXCL-1, accounts for the most amount of variance in the inflammatory response to CYP injection between 150 and 210 min.GRO-KC/CXCL-1, a potent neutrophil chemotactic, is found to be significantly elevated in IC/BPS with and without Hunner's lesions compared to patients with overactive bladder. 6,21RO-KC accounts for a high degree of variance in the inflammatory response during 150-210 min and 450-510 min during which DyNA indicates relatively low to no network complexity.Additionally, DyNA suggests that GRO-KC is not correlated with any other inflammatory mediators.That is, GRO-KC seems to account for a large amount of variance in the inflammatory response in the relative absence of inflammation mediated by other inflammatory mediators.GRO-KC elevation is consistent with the infiltration of neutrophils being a hallmark of acute inflammation and not chronic inflammation which typically defines IC/BPS.However, acute inflammation is critical in the flare up of symptoms in IC/BPS patients and therefore elevation of urinary GRO/KC may foreshadow the advent of acute inflammation that serves as a target for early intervention.
MCP-1, a potent activator of mast cells, accounts for a large degree of variance in the inflammatory response at early time points, particularly during the intervals 210-270 min and 390-450 min.While mast cells, and subsequent mast cell degranulation which results in the release of MCP-1, is not typically seen immediately following CYP injection in the bladders of rodents with induced IC/BPS, 22,23 mast cells in bladder tissue are characteristic of chronic IC/BPS. 15,23Interestingly, our DyNA model shows that MCP-1 is highly correlated with IL-5 and IL-6 during the interval between 630-690 min, and yet only appears in the TI-PCA at the very early time points.Late correlations between MCP-1 and IL-5 and IL-6 suggest that the rise in MCP-1 precedes the infiltration of mast cells, and the migration of granulocytes chemoattracted by IL-5.Since mast cell infiltration of the bladder tissue is a hallmark of severe IC/BPS, MCP-1 levels could aid in phenotyping the heterogenous presentation of IC/BPS subsets.
In our dynamic network models, IL-6 shows temporally varying elevations and correlations with other inflammatory mediators.TI-PCA demonstrates that IL-6 either accounts for the most or the second-most amount of variance in the inflammatory response in the intervals between 330 min and 690 min.While the IL-6 response accounts for a large amount of variance beginning at 330 min, this cytokine only emerges as a node of coordinated inflammation at 550 min, and thereafter is correlated with IFNγ, IL-5, and/or MCP-1.Given the highly interconnected network profile of IL-6 at later time points in this experimental model, we hypothesize that IL-6 is a central driver of chronic inflammation in IC/BPS, which could be tested by correlating the urine levels of IL-6 with severity of symptoms.
A significant contribution of this study to the IC/BPS literature would be the consideration of IL-5, IFNγ, and IL-18 as inflammatory mediators characteristic of IC/ BPS.IL-5, IFNγ, and IL-18 account for large amounts of variance in the inflammatory response at early and intermediate time intervals according to TI-PCA.Increased variance in IL-18 and IFNγ at early time points suggests inflammasome activation in the urothelium in response to damage associated molecular pattern molecules generated as a consequence of pathologic tissue injury akin to a flare up of IC/BPS in patients. 24,25nterleukin-18 emerges as one of the most highly connected nodes in DyNA during the interval 570-630 min.The implication of IL-18's role in IC/BPS and its potentiation of downstream inflammation is supported by studies in patients with IC/BPS, in which this cytokine was elevated relative to controls. 26nterleukin-18, a product of the NLRP3 inflammasome, 22 is a key innate immune messenger.The emergence of IL-18 as a key inflammatory node, the mediators with which it is correlated, and the temporal sensitivity of IL-18 in the IC/BPS pathogenesis would not have been possible without biological network analysis.
Interestingly, much of the literature does not focus on the role of IFNγ or IL-5 in IC/BPS although these two nodes of inflammation exhibit a great deal of variance and are highly correlated with IL-6, a known hallmark of IC/BPS. 6,8,9The clinical significance of the correlation between IFNγ and IL-5 with IL-6 is indicative of a Th1 and Th2 response.Assuming that a Th1 response is characteristic of the inflammatory patterns identified at early time points by TI-PCA and DyNA, we might infer that the elevations in IL-18 are associated with inflammasome assembly and activation in the urothelium.Inflammasome production results in the secretion of IL-18 and IL-1β by the urothelium. 25Interleukin-18 accounts for a high degree of variability at early time points according to our TI-PCA model and shows correlation with IL-1a at early time points as well according to DyNA.IL-1β does not account for a high degree of variance across any time interval, but, from the very beginning, shows a significant correlation with IL-4.This significant correlation with IL-4 foreshadows the development of a Th2 inflammatory response. 27Later increases in the amount of variance accounted for by IL-4, IL-5, and IL-6 further implicate a Th2 response.Late peaking cross-correlations between IFNγ and IL-6 endorses their status as downstream indicators of the interplay between a Th1 and Th2 inflammatory phenotype.In line with research that explores either the innate or adaptive immune system in the context of IC/BPS, we suggest a complex interplay between both the innate immune system, which likely drives inflammation associated with IC/BPS flares, and the adaptive immune system, which likely drives baseline chronic inflammation through dynamic temporal modeling. 28hile the computational analysis of inflammatory mediators characteristic of IC/BPS in a rodent model is both novel and insightful, there exist several opportunities for improvement and expansion.Future studies should model a larger set of inflammatory mediators to assess for more nuanced dynamic inflammatory changes correlated with irritation from acrolein generated by CYP or urinary tract bacterial infection (UTI) response evoked by lipopolysaccharide.Additionally, a longitudinal study of UTI and IC/BPS patients with analogous computational modeling to that presented in this manuscript will be necessary to validate a panel of urine biomarkers for IC/BPS.Human studies will be critical to developing accurate inflammatory networks definitive of IC/BPS flares as the IC/BPS field is limited by the lack of an adequate and sufficient animal model.While CYP induced cystitis is not a direct and perfect model of IC/ BPS, it is one of the widely used animal models to study IC/BPS pathophysiology. 29CYP has been used to study inflammation in the skin as well.In contrast to the skin, bladder mucosa is highly vulnerable to free radicals generated by CYP and therefore, the expression of same inflammatory mediators as expressed locally in skin in response to CYP (IL-1, IL6, IL-18, GRO-a, MCP-1, and TNF-a) are higher because the high metabolic rate of bladder mucosa can sustain the high protein expression rate that cannot be matched by inflammatory foci on skin. 30Thus, CYP serves as a more potent inflammatory stimulus in the bladder compared to studies of CYP in the skin.Past studies on bladder centric BPS/IC using traditional staining and ultrastructure studies do not support permeability and inflammation to be independent entities and certainly do not rule out a bidirectional relationship between inflammation and permeability. 31,32pecifically, oxidative stress evoked by free-radical damage of acrolein can activate inflammasome to injure urothelium as well as endothelium of urothelial capillaries. 33][36] −C).The intermediate phase of the inflammatory response (330-510 min) is characterized by IL-6, GRO-KC, IL-5, and IFNγ (Figure 1D−F).The late phase of the inflammatory response (510 -690 min) is characterized by IL-5, IL-6, and IFNγ (Figure 1G−I).The terms early, intermediate, and late phase of inflammatory response are used to describe the three consecutive time periods of the acute inflammatory response captured during the length of this experiment and do not bear any semblance to the clinical use of similar terms for staging IC/BPS patients.

1
Time-Interval PCA indicates time sensitive inflammatory markers of IC/BPS.Graphs of the sum of principal components indicating the extent to which each inflammatory mediator assayed perpetuated inflammation across the specified time interval (PC = principal component).(A−C) During the early time interval between 150 and 330 min following CYP injection, GRO-KC, MCP-1, IL-18, and IL-5 account for the most amount of variance in the acute inflammatory response characteristic of IC/BPS.(D-F) In the intermediate time interval between 330 and 510 min following CYP injection, GRO-KC and IL-5 continue to account for a high degree of variance in the inflammatory response.IL-6 and IFNγ also account emerge as inflammatory mediators with a high degree of variance.(G−I) The late phase of the inflammatory response between 510 and 690 min is characterized by inflammation mediated by IL-5, IL-6, and IFNγ.CYP, cyclophosphamide; IC/BPS, interstitial cystitis and bladder pain syndrome; PCA, principal component analysis.theinflammatory response.During the same interval, IL-6 and GRO-KC together accounted for 47% of the variance in response to treatment amongst the data (Figure1F).Compared to the first time interval, 150-210 min, IL-6 and GRO-KC together only accounted for 16% of the variance in the inflammatory response.While GRO-KC continued to be a key inflammatory signature of IC during the intermediate interval, IL-6 emerged as a key mediator whose response differed greatly between the early and intermediate portions of the acute inflammatory response.During this intermediate interval, IFNγ also emerged as a mediator that varies greatly with time, and IL-5 and GRO-KC remained key contributors to inflammation as well.In contrast to the early inflammatory response, IL-18 did not account for a large amount of variance between 330 and 450 min (Figure1D−F).During the last time interval between 510 and 690 min, IL-5, IL-6, and IFNγ emerged as key mediators of inflammation.Over the interval 510-570 min, IL-6 accounted for 25% of the variance in the inflammatory response.During the last sampled interval, 630-690 min, IL-5, IL-6, and IFNγ account for 37% of variance in the inflammatory response (Figure 1G−I).

F I G U R E 2
Dynamic network analysis depicts correlations between inflammatory mediators that evolve with time.(A-C) Network connectivity remains relatively low.IL-1β is correlated with both IL-4 and TNF-α between 210 and 270 min.(D−F) Network complexity increases and then resolves by 510 min.GM-CSF is a key node of inflammation during peak network complexity.(G−I) Network complexity peaks between 570 and 630 min, during which IL-18 and IL-1β have the greatest number of connections with inflammatory mediators.By the end of the experiments, IL-5, IL-6, and MCP-1 are all connected with each other between 630 and 690 min.

F I G U R E 3
Dynamic network analysis network complexity.Network complexity at any dynamic time interval is calculated as follows: number of nodes number of edges maximum possible edges between all nodes × .DyNA indicates two peaks in network complexity at 390-450 min and 570-630 min respectively.Between the two peaks in network complexity, which indicates a rise in coordinated inflammatory responses, is a trough at which network complexity is 0. The fluctuations in the rise and fall of network complexity is indicative of the temporally sensitive evolution of the inflammatory response characteristic of IC/BPS.CYP, cyclophosphamide; DyNA, dynamic network analysis; IC/BPS, interstitial cystitis and bladder pain syndrome.