Precision Medicine: How multi-’omics will take shape the future of IBD?

Abstract

Purpose of review:

In this article, we provide an overview of studies examining multi-omic profiling in various clinical scenarios in the management of inflammatory bowel diseases (IBD)

Recent Findings:

IBD arises as a result of an interplay between genetic, environmental, microbial, and immunologic perturbations. The access to high throughput technology as well as the decrease in costs associated with such studies has led to a growing wealth of literature examining the utility of single or multi-omic profiles in the management of IBD. Such studies have commonly examined the genome (and less frequently the epigenome), transcriptome, metabolome, proteome, and the gut microbial metagenome in the context of overall IBD status or specific clinical scenarios including the disease progression or response to treatment. The findings have provided important insight into how each of these compartments reflect underlying disease pathophysiologic processes and, in turn, can influence stratification of patients for clinical management.

Summary:

Multi-omic profiling in IBD has the potential to advance the field of personalized precision medicine in the management of inflammatory bowel diseases.

Introduction

Inflammatory bowel diseases (IBD) are multifactorial, immune-mediated chronic diseases that occur due to genetic, environmental, microbial, and immunologic dysregulation. They affect an estimated 6 million individuals throughout the world and continue to increase in incidence in most regions. An important characteristic of IBD, comprising Crohn’s disease (CD) and ulcerative colitis (UC) is their progression from subclinical inflammation to irreversible bowel damage, resulting in hospitalization, surgery, and disability. Yet a striking feature of both diseases highlighted in multiple observational cohorts is the heterogeneity in age of onset, disease features at presentation and rate of progression to permanent bowel damage.

Advances in high-throughput sequencing technologies have made it possible to capture detailed information on biologic processes at multiple levels including genomic, transcriptomic, metagenomic, metabolomic and proteomic variations that contribute to disease states. In the first wave of omic studies, research focused on a single omic profile, most common tissue gene expression or host genetic variation. While such studies have provided important insights into disease pathogenesis, they have only been modestly informative in understanding the biologic basis of the heterogeneity of disease. A combination of multiple levels of omic profiles, termed multi-omics, may perform better in providing an understanding of the underlying biological processes and their interactions and identify molecular phenotypes. This may, in turn, improve the care of each patient by allowing a precision medicine approach, giving targeted treatment base on disease subtypes, severity and rate of progression, and likelihood of response to specific therapeutic classes. Further spurring advances in this field is the growing affordability of omics assays that have expanded the frontier of omics studies. In this review, we will focus on the definition of multi-omics research and their potential clinical impact in IBD.

Definition of multi-omics

Each type of omics analysis aims to understand a particular level of biologic process in human disease. One of the most common is the study of genomics. Genetic Genomics assays study whether genetic variants are linked with disease, therapy response or disease prognosis. It studies the structure, function and inheritance of the entire genome. In these studies, a large cohort of individuals are genotyped for a wide array of genetic markers to define their impact on disease phenotype or trajectory.

Another related type of omics analysis is the study of the transcriptome (transcriptomics). It studies the set of RNA transcripts that has been generated by the genome, presenting a snapshot of all the transcripts present in a specific cell or tissue. By comparing the transcriptomes it is possible to identify differentially expressed genes in certain tissue or even individual cell populations in the context of a given clinical scenario. Recent advances in transcriptomic profiling have moved it from the realm of bulk transcriptomics (studying entire tissues) to single-cell profiling where the expression of pathways in individual cell types residing in a tissue can be studied to provide a more comprehensive view of disease pathogenesis.

One step closer to the functional impact of such variation is proteomics. Proteomics is the large-scale analysis of proteins in specific compartments (for example, blood or stool). The proteome is a set of proteins produced by a certain cell, organism, tissue or body fluid under specific conditions. The proteome is dynamic and varies over time due to a variety of contributing factors making it a unique tool to understand how these proteins interact with each other and their environment and to better understand the biologic system. Such measured proteins can also act as biomarkers for disease states and be followed dynamically over time relatively non-invasively.

Another powerful tool to better understand the biological processes that take place in IBD is metabolomics profiling. The metabolome comprises metabolites that are produced either by the host or microbiome and can be measured in the serum or stool in addition to other compartments.

One of the most intriguing areas of study where there has been an acceleration of research specifically in IBD has been in defining the role of the gut microbiome through either 16s rRNA or metagenomic sequencing. 16s sequencing uses polymerase chain reaction (PCR) methods to profile the hypervariable V1-V9 regions of the 16s ribosomal RNA of bacteria. This allows profiling commonly at the level of the genus and less commonly definition of species level variation. In contrast, shotgun metagenomics utilizes the entire complement of microbial genomic DNA extracted from a sample such as feces. Sequencing the entire metagenomic DNA provides insight not just at the genus level but also regarding individual species and strains. In addition, metagenomics allows for inferences regarding metabolic function, an insight not commonly available using 16S sequencing methods.

Application of multi-omics in specific clinical scenarios in IBD

Predicting disease complications

IBD exhibits heterogeneous disease trajectories with differing rates of progression towards complicated disease. This heterogeneity is particularly prominent in CD where the disabling phenotypes of stricturing, penetrating or perianal involvement are challenging to predict and associated with significant disability. A few studies have applied proteomic or fecal metagenomic analysis to attempt to predict likelihood of disease progression in CD. A study from Townsend et al. performed a proteomic analysis to distinguish between CD patients with stricturing disease, CD patients with non-stricturing disease and UC patients. The analysis was able to differentiate between the three populations with 70% accuracy by the peptides, and 80% accuracy by the proteins¹. The proteins that distinguished the stricturing phenotype were known to be implicated in complement activation, fibrinolytic pathways and lymphocyte adhesion. A collagen protein that has been linked to development of strictures in adult CD patients is COL3A1. In a cohort of 161 subjects, median baseline concentration of COL3A1 was higher in patients with inflammatory CD who subsequently developed strictures compared to the group who did not develop this complication (2294 vs 1763 pg/mL; p < 0.01)². A multicenter study of newly diagnosed patients with pediatric Crohn’s disease with an inflammatory phenotype (B1) were followed over three years an compared those subjects who later converted to a stricturing phenotype (B2) with those who remained B1³. They identified extracellular matrix protein 1 (ECM1) in the serum to be associated with future fibrostenotic disease (B2 phenotype). ECM1 is a glycoprotein that is also known to be associated with fibrostenotic skin diseases⁴.

Studies have also looked at the ability of metagenomics to predict disease trajectory in Crohn’s disease. In the pediatric RISK cohort of newly diagnosed CD patients who had metagenomics performed on baseline stool samples, abundance of Ruminococcus was associated with the development of stricturing disease while patients who went on to develop penetrating complications had higher abundance for Veillonella⁵. A study that combined data from both Chinese and Western cohorts of patients with IBD examined whether any microbials species are associated with stricturing or penetrating disease⁶. They identified high abundance of Enterobacteriaceae and Pseudomonadaceae in stricturing disease while Aeormonadaceae in the Proteobacteria phylum were enriched in CD patients with penetrating disease. Also, the Pseudomonadaceae and Enterococcaceae were shown higher abundance in CD patients experiencing fistulizing disease.

Between 10-35% of the UC patients is refractory to medical therapy and eventually need surgery⁷. A known complication after colectomy with ileal pouch-anal anastomosis (IPAA) is pouchitis. A small prospective study, using metagenomic techniques, reported that presence of R. gnavus (p<0.001), B. vulgatus (p=0.043), C. perifringens (p=0.011) and absence of Blautia (p=0.04) and Roseburia (p=0.008) was associated with a higher risk of developing pouchitis after IPAA⁸. Another study in UC patients showed that abundance of Ruminococcaceae (OR 1.43, 1.02-2.00; p=0.04), and Sutterella (OR 0.81, 0.65-1.00; p=0.05) at baseline were predictive for corticosteroid-free remission at week 52⁹.

Predicting disease relapse

Another purpose of multi-’omics analysis in IBD studies is to predict relapse of disease either during maintenance treatment or following intentional therapeutic de-escalation. Piere et al. used proteomics to identify biomarkers predicting disease recurrence among patients with Crohn’s disease who had infliximab discontinued while in clinical remission¹⁰. They were able to identify two subsets of protein biomarkers, one linked with the risk of short-term recurrence (median 3.6 months (IQR 2.8-4.1)) and a second subset of certain protein biomarkers with mid/long-term recurrence (median 9.8 months (IQR 6.7-12.5)) of disease, both were linked with various pathophysiology. The proteins linked with short-term relapse were characterized by an innate immune response with mainly proteins produced by the liver and factors of the complement system while proteins associated with long-term relapse were more distinct and heterogenous. A prospective study followed patients with quiescent IBD for 2 years and used baseline proteomic and metabolomic profiling in addition to fecal metagenomics to predict relapse¹¹. Three protein biomarkers (interleukin-10, glial cell line-derived neurotrophic factor, and T-cell surface glycoprotein CD8 alpha chain) and 4 metabolomic markers (propionyl-L-carnitine, carnitine, sarcosine, and sorbitol) were linked with relapse and used to develop a risk score. These proteomic (OR = 9.11; 95% confidence interval, 1.90-43.61) and metabolomic (OR = 5.79; 95% confidence interval, 1.24-27.11) risk scores were able to predict a higher risk of disease relapse.

Other studies have reported a role of short-chain fatty acid (SCFA) (especially butyrate) producing microbes such Bifidobacterium, Eubacterium rectale and Faecalibacterium prausnitzii in maintaining health¹². Their abundance is reduced in patients diagnosed with IBD. In addition, their depletion has been associated with therapy non-response and with early recurrence of disease activity in CD patients after anti-TNF withdrawal. The multicenter STORI study followed patients with CD after infliximab withdrawal and compared their microbiome with healthy controls¹³. Decreased abundance of Firmicutes were noted in patients with relapse of disease compared with non-relapse patients. Additionally, low abundances of F. prausnitzii (p=0.014) and Bacteroides (p=0.030) were able to predict relapse independently from inflammatory blood levels such a C-reactive protein.

Predicting therapy response

The IBD treatment armamentarium has been expanded over the past two decades with the availability of multiple different therapeutic mechanisms. Nevertheless, a considerable number of patients have either no response (primary non-response) or experience loss of response to treatment after an initial benefit (secondary loss of response). Several studies have attempted to use single or multi-omic profiling to predict responders. An initial study in pediatric patients with IBD used fecal calprotectin levels to quantify response to anti-TNF therapy at week 6 and related this to gut microbial composition at baseline¹⁴. Microbial diversity at baseline was significantly higher in the responders compared to the non-responders (calprotectin levels >200 ug/g) at week 6 and continued to show strong associations with calprotectin levels at 3 months. Additionally, higher abundance at baseline for Bifidobacterium, Clostridium colinum, Eubacterium rectale, uncultured Clostridiales and Vibrio and a lower abundance of Streptococcus mitis predicted response to anti-TNF therapy. A larger prospective cohort of patients with moderate-to-severe IBD initiating vedolizumab therapy identified enriched levels of Roseburia inulinivorans and Burkholderiales at baseline to be associated with clinical remission at 14 weeks¹⁵. However, more striking than compositional differences were changes in functional pathways between responders and non-responders. Thirteen pathways were identified to be enriched in Crohn’s disease patients achieving clinical remission, involving the branched chain amino acids (BCAA), suggesting also a functional element in addition to metagenomic differences. Recent work from our group applied high-throughput multi-omic profiling (stool metagenomics, serum metabolomic and proteomic assays) to a prospective cohort of patients that initiating anti-cytokine (anti-TNF or ustekinumab) or anti-integrin (vedolizumab) therapy to treat moderate-to-severe IBD¹⁶. Using metagenomic data, two distinct community clusters could be discerned. Intriguingly, this cluster membership informed likelihood of response to the two distinct therapeutic mechanisms. For patients who belong to community cluster type 1, the likelihood of response at week 52 was significantly greater with anti-cytokine when compared to anti-integrin treatment (66.7% vs. 36.4%, p<0.01). This difference was validated in an independent treatment cohort. Serum bile acid profile was also important in separating responders from non-responders. This serum profiles, in turn, was explained in part by fecal microbial composition Additionally, clinical data alone had a modest ability to predict response to anti-TNF therapy at week 14 (AUC 0.624) but integration of microbial metagenome (AUC 0.849), metabolome (AUC 0.773) or serum proteome (AUC 0.806) improved predictive ability. Additionally, combining all three ‘omics profiles and clinical data showed even greater predictive ability (AUC 0.96). Similar increase in predictive value was noted among vedolizumab users where a combination of clinical data and metagenomics (AUC 0.738) was superior to clinical data alone (AUC 0.619).

Another elegant study combining western and Chinese cohorts showed greater gut microbial diversity and high levels of restoration of Clostridiales during clinical remission were linked with response to anti-TNF treatment⁶. Similar to the above cohorts, microbial profile performed better in predicting anti-TNF response to an accuracy of 86.5% in comparison with single disease activity score (CDAI) (58.7%) or fecal calprotectin (62.5%). Combining microbiota and clinical data improved prediction accuracy further to 93.8%.

Proteomic profiling has also shown promise in predicting therapeutic response to IBD though the literature is sparser than for metagenomics. Seventeen proteins regulated by acetylation were identified to be predictive of response to anti-TNF therapy in biologic-naïve CD patients while 4 proteins were linked with loss of response¹⁷. An elegant study measured 13 proteins (ANG1, ANG2, CRP, SAA1, IL-7, EMMPRIN, MMP1, MMP2, MMP3, MMP9, TGFA, CEACAM1, and VCAM1) in blood to identify quiescent disease in CD patients on biological therapy¹⁸. They assessed this test in 2 validation cohorts and concluded that the protein test had a comparable accuracy to identify remission as FC levels and even better accuracy than C-reactive protein.

Another potential biomarker for assessing biological therapy response is Oncostatin M (OSM), a cytokine belonging to the interleukin (IL-6) family. A cross-sectional study in newly diagnosed IBD patients initiating biological therapy reported that high serum levels OSM was highly predictive of non-response to anti-TNF and vedolizumab therapy¹⁹. The same research group assessed several gene expressions in whole blood samples of IBD patients initiating biological treatment²⁰. Non-responders to anti-TNF therapy had low levels of whole blood TREM1 at baseline (FC 0.67, p=0.001), with an area under the curve of 0.78 (p=0.001).

In addition to studies that focused on predicting response to biological therapy, a few studies used similar methods to predict response to corticosteroid therapy. A pediatric cohort of treatment-naïve UC patients reported higher Actinomyces and lower Clostridium abundances in stool in those patients that achieved remission with corticosteroid therapy after 4 weeks in comparison with those with ongoing disease activity²¹. A follow up study (30935734) of the same cohort and previously mentioned showed that abundance of Ruminococcaceae (OR 1.43, 1.02-2.00; p=0.04), and Sutterella (OR 0.81, 0.65-1.00; p=0.05) at baseline were linked with corticosteroid-free remission at week 52.

Gaps in literature and future directions

While multi-omics holds significant promise in advancing care of patients towards the goal of precision medicine, there remain many gaps in this emerging field (Table 1). Most of the studies have a relatively small sample size which consequently impact statistical power and generalizability. Many have lacked external validation cohorts. In addition, the studies thus far are heterogeneous in included participants, definition of study outcomes, timing and processing or samples, as well as profiling technology. This has markedly limited the ability to pool together results across studies and is a key impediment for the advancement of the field. There is important need for a systematic consensus on study procedures and definitions, and where possible, multicenter collaboration with homogeneous study execution to provide robustly actionably information. Second, many of the findings thus far have been either single or at best two-omic studies and have identified associative findings rather than confirm causal mechanisms. The latter is important to understand disease pathogenesis and develop new therapeutic options. Third, many studies are cross-sectional with profiling biospecimens at a single point in time. These studies do not account for the variability of the omics profiles over time. However, examining time-dependent changes in omic profiles require large cohorts with systematic periodic gathering of samples. Critically important collaborative efforts to generate and analyze multi-omic data to shed insight into disease pathogenesis such as the Human Microbiome Project (HMP and HMP2/IBDMBD), The Dutch 1000 IBD initiative, and the Pediatric RISK and PROTECT cohorts are critically important steps to advance the field^22,23.

Table 1:

Gaps in the current multi-omic studies and suggestions for future studies

Gap	Proposed solution(s)
Small sample size	Multi-center studies and collaboration; minimizing regulatory hurdles for such collaboration
Heterogeneous population	Strict definition at inclusion resulting in a homogenous cohort; similar time points for sampling
Variability in study outcomes	Use of common and validation definitions for study outcomes (for example, endoscopic healing vs. physician global impression of treatment response)
Limited causal inference	Prospective studies (rather than cross-sectional alone), bio-informatic methods for causal inference
Heterogeneous and incomplete covariate capture	Standardizing minimum covariate adjustment required for each type of omic study and similar tools for data capture
Suboptimal sequencing and analytic reproducibility	Standardized description of methods including reproducibility; deposition of data in public databases for independent assessment of reproducibility; inclusion of validation cohorts
Single-omic studies	Collection and banking of specimens from different compartments (blood, stool, tissue) to facilitate multi-omic profiling
Limited racial ethnic diversity	Expand diversity in included study population, expand geographic catchment area for studies

Conclusion

The treatment landscape of IBD has evolved but breaking the ceiling of response rates to current medications will likely require a precision-medicine approach that allows for targeting the right treatment to the right patient (Figure 1). With the decreasing cost and increasing access of multi-omic platforms, such results can help to personalize treatment regimens and be incorporated into the clinic. The popularity of lifestyle tracking devices also affords the opportunity for integration of such assessments of environmental determinants into predictive models. Advancing the field requires standardization of study designs and profiling across research groups to achieve meaningfully large sample sizes to provide robust validated results that can then be incorporated into patient care.

Figure 1:

The proposed algorithm for a personalized precision medicine approach for the management of inflammatory bowel diseases

Summary.

• Inflammatory bowel diseases are heterogeneous disorders with varying rates of progression to disease complications and likelihood of therapy response.

• Multi-omic profiling has been applied to prospective cohorts to predict response to medical treatments, and likelihood of Crohn’s disease (CD) or ulcerative colitis (UC) related complications.

• Abundance of butyrate producing phyla in the stool of patients with Crohn’s disease may predict less aggressive and more therapy responsive course of Crohn’s disease.