Gut virome dysbiosis following focal cerebral ischemia in mice

Abstract

Stroke leads to gut bacterial dysbiosis that impacts the post-stroke outcome. The gut microbiome also contains a high abundance of viruses which might play a crucial role in disease progression and recovery by modulating the metabolism of both host and host’s gut bacteria. We presently analyzed the virome composition (viruses and phages) by shotgun metagenomics in the fecal samples obtained at 1 day of reperfusion following transient focal ischemia in adult mice. Viral genomes, viral auxiliary metabolic genes, and viral protein networks were compared between stroke and sham conditions (stroke vs sham, exclusive to sham and exclusive to stroke). Following focal ischemia, abundances of 2 viral taxa decreased, and 5 viral taxa increased compared with the sham. Furthermore, the abundance of Clostridia-like phages and Erysipelatoclostridiaceae-like phages were altered in the stroke compared with the sham cohorts. This is the first report to show that the gut virome responds acutely to stroke.

Introduction

Stroke induces significant gut microbial dysbiosis that influences the post-stroke pathophysiology. ¹ As brain damage evolves rapidly during the first day after a stroke, the changes during the acute period are critical for therapeutic interventions. ² Patients with gut hemorrhage, dysbiosis, translocation of gut bacteria, and gut bacteria-induced sepsis exhibit delayed recovery, worse neurological outcomes, and a higher rate of mortality after stroke. ³ Modulation of gut bacterial composition or bacterial metabolites was shown to promote neuroprotection and recovery after stroke. ⁴

Most, if not all, studies to date that evaluated the role of the gut microbiome in post-stroke pathology have focused on bacteria. However, viruses and bacteriophages are major components of the gut microbiome with significant physiological roles. The majority of viruses in the gut are phages that can infect diverse gut bacteria such as those belonging to the phyla Actinobacteria, Bacteroidetes, Firmicutes, and Proteobacteria, and modulate host-bacterial interactions in health and disease. ⁵ Alterations in the gut phage composition have been shown in several diseases such as inflammatory bowel disease, Parkinson’s disease, and diabetes. ⁶ Further, phages regulate microbial colonization and act as an interface in preventing infections by adhering to mucosal membranes and preserving non-host immunity. ⁷ Both lytic and lysogenic phages regulate microbial composition, diversity, and resilience. In addition, phages interact with host innate immunity and cytokine synthesis, and importantly short-chain fatty acids promote phage production in bacteria. ⁸ When high doses of phages were administered to humans/rodents, susceptible bacterial host cells were lysed, triggering the release of endotoxin/cytokines that compromise the intestinal barrier. ⁹ Phage therapies were also shown to be effective in treating drug-resistant infections in humans. ¹⁰ We presently evaluated if transient focal ischemia alters the gut virome and its protein network.

Methods

Transient focal ischemia

All animal procedures were approved by the University of Wisconsin Research Animal Resources and Care Committee. Animals were cared for in compliance with the Guide for the Care and Use of Laboratory Animals [U.S. Department of Health and Human Services publication no. 86–23 (revised)]. All procedures were conducted in compliance with the “Animal Research: Reporting of In Vivo Experiments” guidelines. Mice were randomly assigned to experimental groups. Adult C57BL/6 male mice (12 weeks old; Jackson Laboratories USA) were subjected to 1 h of transient middle cerebral artery occlusion (MCAO) under isoflurane anesthesia followed by 1 day of reperfusion as described earlier. ¹¹ Sham-operated animals served as the control. Regional cerebral blood flow (rCBF) and physiological parameters (pH, PaO₂, PaCO₂, and hemoglobin) were monitored, and rectal temperature was maintained at 37 ± 0.5°C during the surgery and recovery from anesthesia. Post-surgical care and monitoring were provided as specified by IMPROVE guidelines. At 1 day of reperfusion, infarction was confirmed by T2-weighted MRI. Mice with no noticeable neurological deficits or no sign of infarction in MRI scans on day 1 of reperfusion and those that showed a hemorrhage upon euthanasia were excluded from the analysis.

Shotgun metagenomics and virus/phage analysis

From each mouse, fecal pellets were collected at 1 day of reperfusion and the genomic DNA was extracted using the GNOME® DNA Isolation Kit (MP Biomedicals, Santa Ana, CA, USA). After a quality check by the Qubit® dsDNA HS Assay Kit (Life Technologies, Grand Island, NY), samples were prepared according to the Celero PCR Workflow with Enzymatic Fragmentation (NuGEN). The quality and quantity of the completed libraries were assessed using an Agilent bioanalyzer and Qubit® dsDNA HS Assay Kit, respectively. Libraries were standardized to 2 nM. DNA sequencing (paired end, 150 bp) was performed using the Illumina NovaSeq6000. Data were processed using the Illumina Pipeline, version 1.8.2. Reads were quality checked and trimmed using Trim-galore ¹² and assembled using MetaSPAdes v3.12.0 (settings: meta, kmer values of 21, 33, 55, 77, 99, and 127 for assembly). ¹²

Metagenomic virus/phage analysis was conducted as described earlier.^13,14 Viruses were identified and auxiliary metabolic genes were classified from metagenomes using VIBRANT (v1.2.1). ¹³ Taxonomic classification of virus/phage sequences was performed as previously described. ¹⁴ Briefly, a reference phage database of non-redundant prokaryotic virus proteins derived from NCBI databases was used to query unknown viruses using DIAMOND Blastp (v0.9.14.115), and taxonomy was selected based on the hierarchical classification of hits. ¹⁵ VConTACT2 (v0.9.5, default parameters) was used to construct the network diagram and only linkages with edge weights above 5 were retained. ¹⁶ The network diagram was visualized using Cytoscape (v3.7.2). ¹⁷ Virus sequences were dereplicated using mash (v2.0) and MUMmer (v3.1) at 90% identity and 70% coverage and reads to dereplicated sequences were aligned using Bowtie2 (v2.3.4.1).^18–20 Following that, stroke-associated viruses were identified according to read coverage dissimilarity between sham vs. stroke groups. Coverage was extracted using Samtools (v1.11), a differential abundance of viruses across samples was calculated using DESeq2 (v1.28.1) and diversity calculations were made using Python packages scikit-bio and scikit-learn.^21,22 To reconstruct metagenome-assembled genomes (MAGs) from the assemblies, the MetaWRAP binning module was used to bin using metabat2, maxbin2, and metabat1. ²³ To consolidate bins, DAS_Tool was used with the score threshold of 0.4. ²⁴ To predict taxonomy, the gtdbtk (v0.1.3) classify_wf module was used. ²⁵ CRISPR spacers were identified in MAGs using CRISPR Recognition Tool (v1.2). ²⁶ Blastn was used to search spacers against viral genomes and matches of at least 97% identity were retained as host predictions. Sequencing reads were also analyzed for changes in the abundance at strain level.

Results

Post-stroke changes in fecal virome

The relative abundance of virome composition at the family level and the total number of operational taxonomic units in the feces were similar between stroke and sham cohorts (Figure 1(a)). However, Simpson’s evenness and Shannon/Simpson’s indices were significantly decreased in stroke compared with sham (p < 0.05; n = 6/group) (Figure 1(b)). Furthermore, Bray-Curtis dissimilarity depicts lower (cooler/blue colors) dissimilarity between similar conditions (stroke to stroke, sham to sham), and higher (warmer/red colors) between opposite conditions (stroke to sham) (Figure 1(c)). This not only suggests that diversity among stroke samples (by abundance per virus) is different than among sham samples, but also indicates the similarity between the individual samples within a group (either sham or stroke). The pairwise dissimilarity analysis between samples of each condition displayed the lowest dissimilarity between the individual samples of the stroke group. This suggests that following stroke, the gut virome in our study responds more similarly to stroke in comparison to sham. Moreover, stroke-to-sham comparisons display the highest dissimilarity, further suggesting a differential response to stroke.

Figure 1.

In adult mice, gut viral composition at the level of the family was unaltered after focal ischemia compared with sham (a). The α-diversity at family level was similar between sham and stroke cohorts, but Simpson’s evenness, Simpson’s and Shannon’s indices were lower in stroke compared to sham (b). Values are mean ± SD (n = 6/group; *p < 0.05 compared with sham by Student’s t-test). Heatmap depicts pairwise dissimilarity between stroke and sham cohorts calculated based on the relative abundance of the viruses/sample; values on the axis indicate sample numbers (c). Lower dissimilarity indicates a higher relationship. Relative coverages were normalized between stroke and sham.

Stroke-induced changes in virome composition at strain level, metabolic pathways, and protein network clustering

At 1 day of reperfusion following transient MCAO, significant changes in fecal virome taxa at the strain level were observed in comparison with sham (p < 0.05; n = 6/group) (Figure 2(a)). Of the 10 viral taxa identified reliably in both stroke and sham cohorts, 2 decreased and 5 increased in abundance in stroke compared with the sham (Figure 2(b)). Auxiliary metabolic genes and metabolic pathways impacted by viruses in the gut microbiome were not significantly altered between the stroke and sham cohorts (Figure 2(c)). Protein network clustering depicted different clusters along with host taxonomy (a, b; Clostridia-like phages, and c; Erysipelatoclostridiaceae phages) between sham and stroke cohorts (Figure 2(d)). Protein network clusters (a/b) depict that Clostridia-like phages are shared between sham and stroke, but the clusters (c) of Erysipelatoclostridiaceae-like phages appear to be associated exclusively with sham. The Protein network suggests the presence of more unique genomes in the stroke-associated virome. In contrast, as seen by the distinct clustering of stroke- and sham-associated phages distal to the reference phage clusters, we identified phages unique from those present in the reference database. In addition, the MDS plot (supplementary figure 1) showed distinct differences in sham and stroke. Overall, this indicates that the composition of the virome and its protein network in sham and stroke cohorts was different.

Figure 2.

At the strain level, viral taxa altered significantly in the stroke cohort compared with the sham cohort (a & b). AMLV, Abelson murine leukemia virus_u_t; Lp Lj771, Lactobacillus prophage Lj771_u_t; MCR, Murine type C retrovirus; MFAG, Mongoose feces-associated gemycircularvirus c; MMSV, Moloney murine sarcoma virus_u_t; MMTV, Mouse mammary tumor virus_u_t; MOV, Murine osteosarcoma virus_u_t; SFFV, Spleen focus-forming virus_u_t; SV, Siphoviridae_u_t. Values are mean ± SD (n = 6/group); *p < 0.05 compared with sham (Mann-Whitney U test). Auxillary metabolic gene pathways represented by at least 5 viruses in sham and stroke showed similar patterns between the cohorts (n = 6/group) (c). The phage protein networks were dissimilar between stroke and sham cohorts and Protein network clustering is presented along with host taxonomy based on CRISPR match host predictions (a and b are Clostridia-like phages, and c is Erysipelatoclostridiaceae-like phages) with each dot representing an individual viral genome (n = 6/group) (d).

Discussion

The stability of the gut virome is crucial for gut bacterial fitness and overall gut microbiome homeostasis. Stroke leads to time-dependent dynamic changes in microbial diversity and evenness that predict dysbiosis. ²⁷ The α-diversity was not significantly different in ischemic stroke patients than in healthy controls. ²⁸ However, similar to our current findings, a pig model of ischemic stroke showed a significant reduction in the Shannon and evenness indices at the acute phase of stroke, but not at the delayed phase of stroke, which suggests the microbial diversity depends on post-stroke progression. ²⁷ Although viral diversity after stroke has not been explored yet, studies with a high-fat diet in mice showed a reduction in both α- and β-diversity of the bacterial community, but only β-diversity in the viral community. ²⁹

Bacteriophages use commensal bacteria as a vehicle to interact with the host immune system. ³⁰ Human immune cells maintain a dynamic balance between gut viruses and immunity and are capable of detecting pathogen-associated molecular patterns that result from a disease or an infection. Recent studies identified a positive correlation between gut bacterial and viral alpha diversity. ³¹ The gut virome infects and regulates the dominant commensal bacteria which use their phages to kill competing bacteria in the intestine. ³²

Furthermore, the gut virome also directly affects the human immune system by maintaining continuous low-level asymptomatic immune responses. ³⁰ Furthermore, AMGs of viruses are known to be transcribed during infection and involved in viral replication, fitness, and host interactions by direct involvement in the host metabolism. ³³ Identifying AMGs involved in metabolic pathways that can influence inflammatory responses, such as sulfur metabolism and sulfide production, may provide insights into the impacts of virome changes on stroke outcomes. ³³ The proportion of virulent phages increases in patients with digestive or respiratory system diseases compared to healthy individuals. ³⁴ While disease-specific phages were identified in sick compared to healthy individuals, no studies to date evaluated the effect of an acute insult to CNS like stroke on the gut virome. ³⁴

Stroke disrupts the gut barrier, which might lead to phage translocation to different organs. When healthy rats were challenged with a bacteriophage cocktail, gut permeability and inflammatory responses increased and fecal bacterial composition changed significantly. ³⁵ Furthermore, bacteriophage cocktail treatment decreased the abundance of beneficial bacteria such as Lactobacillus in the gut, probably due to impaired gut permeability and leakage. ⁹ In humans, Lactobacillus is an abundant gut bacteria that harbors numerous phages and modulates post-stroke infections. While Lactobacillus ruminis abundance is high and positively correlates with interleukin-6, Lactobacillus sakei abundance is low and deleterious to intestinal mucosal defense in stroke patients. ³⁶ This suggests that understanding the phages that regulate Lactobacillus might be crucial for further comprehending their role in ischemic pathogenesis. In addition, a recent study showed that phage supplementation with Lactococcal 936 bacteriophage increases memory in flies, and Caudovirales enriched fecal microbiota transplantation in mice increased memory. ³⁷ Phages of Siphoviridae are positively associated with cognition. ³⁷ Reduction in phages SV and Lp Lj771 (Figure 2(b)) suggest the possible association with post-stroke cognitive dysfunction.

Prophage integration into bacterial genomes modifies the lysogenic conversion. While positive conversion increases a bacteria’s fitness, the negative conversion inactivates host/bacterial genes. ³⁸ Phage diversity follows bacterial host diversity, and the phages replicate through lytic or lysogenic replication cycles that depend on the gut mucosal membrane. ³⁹ Stroke disrupts the gut mucosal layer altering the lytic replication of phages and thereby bacterial composition and pathological outcome. Fecal filtrates containing phages (without bacteria) are used as an alternative approach for fecal phage transplantation to treat Clostridium difficile infection. ⁴⁰ Several bacteriophages are considered as “Generally Recognized as Safe” by the U.S. food and drug administration for human consumption. Compounding that, a recent study on healthy humans demonstrated that oral ingestion of a bacteriophage cocktail as a dietary supplement is safe and tolerable. ⁴¹ Specifically, when a cocktail of Escherichia coli-targeting bacteriophages was orally administered to healthy humans, there was a selective increase in butyrate-producing genera Eubacterium and a reduction in circulating interleukin-4. ⁴¹ Overall, our study is the first to show that acute ischemic stroke alters the gut virome. This study provides a foundation to conduct larger mechanistic studies to understand the role of bacteriophages in the stroke-associated microbiome and to develop phage-based stroke therapies.