The gut-brain axis in depression: Are multi-omics showing the way?

Abstract

It is time for a paradigm shift in psychiatry. The need for biologically based models to understand clinical heterogeneity is gaining momentum. Integrating the microbiome into biomarker discovery provides an accessible, biological approach to generate clinically relevant biomarkers that consider the host and the environment in a comprehensive way.

It is time for a paradigm shift in psychiatry. The need for biologically based models to understand clinical heterogeneity is gaining momentum. Integrating the microbiome into biomarker discovery provides an accessible, biological approach to generate clinically relevant biomarkers that consider the host and the environment in a comprehensive way.

Main text

Advances in human microbiome research have revealed the importance of the microbiome to mental health and disease. In the past decade, our understanding of how the microbiome contributes specifically to depression has advanced. Links between gut bacteria and clinical symptoms in depression, including anxiety, sleep quality, severity of disease, and anhedonia, have been reported,¹ and yet, these studies fail to identify the key biological mechanisms that underly these observations. A better biological understanding of how microbiome-host signaling contributes to depression is needed. With the novel consideration of the microbiome in addition to host systems, a biological signature of the gut-brain axis has the potential to provide a surrogate biomarker to distinguish subgroups. Such knowledge will then provide the basis for short-term and long-term clinical trials to validate biomarkers, demonstrate the predictive value of biomarkers for treatment response, and provide guidelines for the development of accessible and effective measurement-based approaches that draw from the biological understanding at an individual person level.

What is a multi-omics approach to biomarker discovery?

Omics approaches are high-throughput tools for the measurement of biological molecules that generate large, comprehensive datasets to better understand the molecular mechanisms of health and disease. While the use of “omics” analyses of host systems is an established approach to biomarker discovery, integration of microbiome omics advances a richer and holistic approach that may more accurately represent the biology of the individual (Figure 1). Omics approaches in host systems include genetics, epigenetics, transcriptomics, proteomics, metabolomics, and, with the inclusion of imaging and electroencephalogram (EEG), the connectome. While considering each of these approaches on its own can advance our understanding of how molecular and functional information underlie a particular trait or disease, multi-omics profiling leverages more than one omics domain in an integrated manner to generate a more comprehensive view of the key biological processes involved.

Figure 1.

Omics approaches in biomarker discovery

(A) Several omics approaches, including genetics, epigenetics, transcriptomics, proteomics, and metabolomics, generate global datasets that represent biological systems in host tissues. Neuroimaging and electroencephalogram (EEG) can reveal brain connectivity patterns. Human microbiome research utilizes parallel omics tools shown in the box on the right to map the diversity, composition, and functional readouts of the microbiome. Integration of microbe-host multi-omics analysis has the potential to generate biomarkers that can be applied clinically at the individual level to improve outcomes in depression.

(B) Integration of microbiome, host, and clinical datasets can identify mechanistic microbe-host interactions. On the bottom left, a Circos plot shows an example that was generated using DIABLO. Data-driven machine- learning approaches that combine mixed features can assess predictive accuracy of a multi-omics biosignature compared to those with individual features (bottom middle graph). Visualization tools can integrate microbiome and host datasets to show taxonomic and phylogenetic microbial community structure in the context of clinical metadata (bottom right).

All surfaces of your body are covered with microbes that include bacteria, protozoa, viruses, fungi, and parasites. Microbiome omic analyses of the related bacteriome, archaeome, virome, mycobiome, and parisitome using amplicon sequencing, metagenomics, metatranscriptomics, metaproteomics, and metabolomics can provide diversity, composition, community structure, and functional readouts of the microbial community. To date, research into depression has focused on the bacteriome, predominantly the gut bacteriome, with a small selection of studies considering the oral bacteriome. Innovative human microbiome studies in depression that include other microbial niches have the potential to advance our understanding of these ecosystems and reveal novel microbe-host signaling pathways that influence depression.²

How can we harness a multi-omics approach in depression?

In the past decade, depression has moved from the third leading cause to the leading cause of disability world-wide, affecting more than 350 million people globally.^3,4,5 Biomarker discovery in depression lags behind other chronic medical conditions because of the reliance on subjective symptoms in clinical diagnosis, the lack of recognition for the contribution of peripheral biological factors to mental health, and a lack of knowledge related to how differences in these peripheral factors contribute to clinical heterogeneity in depression. An individual’s current mood state, their biology, their exposure to early and proximal stressors, and their lifestyle are represented in their microbiome. Therefore, a clear advantage to integrating the microbiome into the omics generation of a clinically relevant biosignature is that, on its own, it provides a holistic combined outcome measure that is driven by host, environmental, lifestyle, and life history factors. In combination with host-based omics datasets, there is a great potential to identify robust biosignatures that can be utilized to (1) stratify individuals into biologically defined subgroups, (2) predict treatment response, (3) identify novel molecular targets for drug development, and (4) better match microbiome-targeted treatment approaches such as diet, exercise, and psychobiotics to individuals with depression.

Innovation in human microbiome research has been driven by advanced molecular and sequencing tools as well as the continual development of state-of-the-art analytical approaches. The development and open sharing of omics integration tools provide opportunities for researchers to generate multi-omics biosignatures that represent host and microbiome as well as host-microbiome interactions. Integration of multiple high-dimensional datasets has the potential to generate multi-omics biosignatures that may explain biological differences contributing to clinical heterogeneity in depression. For example, data integration analysis for biomarker discovery using latent components (DIABLO) implements a modified sparse partial least squares (sPLS) discriminant analysis algorithm to construct multi-omics signatures based on within- and between-dataset correlations that predict group membership.⁶ DIABLO can be applied on large datasets from different domains, with or without prior feature selection.

To demonstrate the utility of DIABLO in the context of the gut-brain axis, we describe two recent studies that used this tool. In the first study, integration of measures of neurophysiological development and maturation with clinical features, immune phenotype (cells and cytokines), and microbiome datasets using DIABLO revealed a microbiota-immune-brain biosignature that predicted brain injury in premature infants.⁷ Notably, this study showed several differences between premature infants with and without brain injury in each of the domains examined; however, the use of DIABLO provided a mechanistic view of microbiota-immune interactions that may influence brain development in this cohort.⁷ A recent case-control study used DIABLO to examine the microbiota-metabolite-immune brain axis in a cohort that included major depressive disorder (MDD) and healthy individuals.⁸ Changes in gut microbial taxa, plasma metabolites, and immune cell subsets were identified, several of which were associated with depression severity. A multi-omics signature that distinguishes individuals with MDD from healthy individuals was generated.⁸ While the sample size was small in this study, the use of DIABLO provided insight into the microbiota-metabolite-immune signaling cascades that may contribute to the development of depression.

In this commentary, DIABLO serves as an example; however, there are numerous bioinformatic and machine-learning tools available. The recent studies described above integrated microbiome and host datasets to examine how the interactions between these systems contribute to clinical presentation. At this juncture, multi-omics approaches that integrate microbiome and host datasets are readily available for use in depression research. On the microbiome side, several datasets can be considered for integration: (1) 16S rRNA and metagenomic sequencing provide diversity, compositional, differential abundance, and community structure datasets; (2) deep shotgun metagenomic sequencing provides bacterial community composition and functional datasets but also opens the door to the exploration of the virome and the mycobiome; (3) metabolomics provides metabolites and their related metabolic pathways; and (4) metatranscriptomics provides RNA expression datasets. As shown in Figure 1B, integrating microbiome datasets with clinical data and selected host omics datasets can provide novel insights. Mechanistic host-microbe interactions can be identified using DIABLO and visualized in a Circos plot (Figure 1B, bottom left). Data-driven machine-learning approaches that combine mixed features can assess predictive accuracy of a multi-omics biosignature compared with those with individual features (Figure 1B, bottom middle). Visualization tools, such as Graphical Phylogenetic Analysis (GraPhlAn),⁹ can combine microbiome and host datasets to show taxonomic and phylogenetic microbial community structure in the context of clinical metadata (Figure 1B, bottom right).

Challenges to advancing multi-omics approaches in depression

Currently, there are several interventions for treating MDD, including lifestyle interventions, psychotherapy, pharmacotherapy, psychological treatment, complementary and alternative medicine, digital health interventions, and neuromodulation treatments.¹⁰ Evidence-based clinical guidelines provide a framework for acute and sequential treatment of individuals with MDD, in which clinical decisions are based on quality of life, risk factors, clinical symptoms, symptom severity, number of past episodes, and previous treatment response.¹⁰ In spite of the number of treatments available, a high proportion of individuals do not receive adequate treatment and do not achieve symptom remission, often failing to respond to several interventions.¹¹ The lack of understanding of the biological mechanisms that underlying clinical heterogeneity and treatment response has limited the ability of clinicians to match individuals to available treatments. In addition, some emerging biological mechanisms do not seem to be specific to any single diagnostic entity but are common to multiple diagnoses, challenging the specificity of our diagnostic constructs. As such, we urgently need biomarkers and specific phenotypes to stratify patients into more homogeneous subgroups. This stratification will allow for more uniform and potentially more effective treatment responses based on an individual’s biology.

A key challenge to translating research findings to the clinical is the need for biomarkers that can be applied at the individual patient level. Standard clinical trial methods and design may identify prognostic and predictive biomarkers, but the results may not generalize to other populations or to a real-world clinical practice. Considering the long journey from biomarker discovery to clinical impact, it would be optimal for researchers to consider outputs that can be applied to an individual. To this end, a recent study that integrated 16S rRNA sequencing data with clinical data utilized weighted correlation network analysis, borrowed from the transcriptomic field, to examine stable gut bacterial communities in a naturalistic cohort of depressed individuals. A key feature of this analytical approach was that it provided a summary statistic, referred to as an “Eigentaxa,” for each of three stable microbial communities at the individual level.¹² Moreover, the association of microbiota community structure with clinical features, using a model that controlled for age, body mass index, and sex, identified a bacterial network associated with anxiety.¹² As such, since each individual’s microbiome is their own, microbiome-based biomarkers are well positioned to provide biomarkers that can be used at an individual level.

Another barrier to clinical translation is that candidate biomarkers are often generated in relatively small, well-defined, diagnostic-specific clinical cohorts, and the findings are not generalizable to broader clinical settings. The rationale for this research design is that it controls for clinical heterogeneity; however, it fails to recognize the “real-life” phenotype of most depressed patients. Attention to the potential stratification role for biotypes (imaging, EEG, peripheral immune, metabolic) in research designs, as well as the consideration of symptom heterogeneity and co-morbid conditions in the analytical design, for clinical trial investigations is the best approach to address this challenge. In the short term, it is necessary to identify and validate multi-omics brain and peripheral biomarkers that are proxies of biological subgroups linked to clinical phenotype or treatment outcome. In the long term, the complexity of multi-omics analyses will need to be simplified to be effectively used in the clinic. Proof of concept of this approach was recently demonstrated in cancer immunotherapy, where a metagenomic-based predictive biomarker was translated into a rapid qPCR test.¹³

Concluding remarks

Advancing our understanding of the complexity of how genetic, environmental, and lifestyle factors influence mental health will require collaborative approaches between clinicians and researchers that understand the analytical approaches is essential to interpret the results and move beyond common methodological pitfalls that limit the reproducibility of candidate biomarkers. Tackling existing limitations through innovative research study design is needed. While not an exhaustive list, suggestions include (1) ensuring that there is statistical power necessary to generate findings at an individual level, (2) adoption of standard outcome measures that can be easily applied in the clinical setting, (3) longitudinal assessment of clinical trajectories and biomarkers, and (4) employing strategies to increase diversity and participation of underrepresented communities in depression studies. While much of the extant research that demonstrates a role for the microbiome in depression has employed standard microbiome analytical approaches, clinical studies are progressively utilizing multi-omics designs that include host and microbiome systems. To advance biomarker discovery in depression, innovative study design in clinical psychiatry that generates high-quality multi-omics datasets will be required. Once identified, translating a multi-omics biomarker for clinical use will be challenging, as there is a need to select key features that can be tested in the clinic and that are accurate and reliable at the level of the individual. This will rely on innovation and collaboration of scientists, industry partners, and other stakeholders as well as dissemination and educational efforts for maximal clinical impact.

Acknowledgments

This research program was funded by the National Science & Engineering Research Council of Canada (RGPIN-2018-06834, J.A.F. PI), the Ontario Brain Institute (POND, J.A.F. PI), the Hersh Foundation (M.H.T. PI), and the Rose Foundation (M.H.T. PI). The content is solely the responsibility of the authors and does not necessarily represent the official views of the various funding organizations.

Declaration of interests

J.A.F. has served on the scientific advisory board for MRM Health NL and has received consulting/speaker fees from Klaire Labs, Takeda Canada, WebMD and Rothmans, Benson, & Hedges Inc.

M.H.T. has provided consulting services to Acadia Pharmaceuticals, Alkermes Inc., Alto Neuroscience Inc., Axsome Therapeutics, BasePoint Health Management LLC, Biogen MA Inc., Cerebral Inc., Circular Genomics Inc., Compass Pathfinder Limited, Daiichi Sankyo Inc., GH Research, GreenLight VitalSign6 Inc., Heading Health, Janssen Pharmaceuticals, Legion Health, Merck Sharp & Dohme Corp., Mind Medicine Inc., Myriad Neuroscience, Naki Health Ltd., Neurocrine Biosciences Inc., Noema Pharma AG, Orexo US Inc., Otsuka America Pharmaceutical Inc., Otsuka Europe Ltd., Otsuka Pharmaceutical Development & Commercialization Inc., Praxis Precision Medicines Inc., PureTech LYT Inc., Relmada Therapeutics Inc., SAGE Therapeutics, Signant Health, Sparian Biosciences, Titan Pharmaceuticals, Takeda Pharmaceuticals Inc., and WebMD. Additionally, he has received editorial compensation from Elsevier and Oxford University Press.