Distinctive Gut Microbiota Alteration Is Associated with Poststroke Functional Recovery: Results from a Prospective Cohort Study

Objectives Functional prognosis is potentially correlated with gut microbiota alterations following the dysregulation of the gut-microbiota-brain axis after stroke. This study was designed to explore the poststroke alterations of gut microbiota and potential correlations between gut microbiota and global functions. Methods A total of thirty-eight patients with stroke and thirty-five healthy demographics-matched controls were recruited. Their fecal DNAs were extracted, and the V3-V4 regions of the conserved bacterial 16S RNA were amplified and sequenced on the Illumina MiSeq platform. Microbial composition, diversity indices, and species cooccurrence were compared between groups. Random forest and receiver operating characteristic analysis were used to identify potential diagnostic biomarkers. Relationships between discriminant bacteria and poststroke functional outcomes were estimated. Results Higher alpha diversity of gut microbiota was observed in poststroke patients as compared to the healthy controls (p < 0.05). Beta diversity showed that microbiota composition in the poststroke group was significantly different from that in the control group. Relative abundance of nine genera increased significantly in poststroke patients, while 82 genera significantly decreased (p < 0.05). The accuracy, specificity, and susceptibility of the optimal model consisted of the top 10 discriminant species were 93%, 100%, and 86%, respectively. Subgroup analysis showed that bacterial taxa abundant between subacute and chronic stroke patients were overall different (p < 0.05). The modified Rankin scale (mRS) (r = −0.370, p < 0.05), Fugl-Meyer assessment (FMA) score (r = 0.364, p < 0.05), water swallow test (WST) (r = 0.340, p < 0.05), and Barthel index (BI) (r = 0.349, p < 0.05) were significantly associated with alterations of distinctive gut microbiota. Conclusions The gut microbiota in patients with stroke was significantly changed in terms of richness and composition. Significant associations were detected between alterations of distinctive gut microbiota and global functional prognosis. It would facilitate novel treatment target selection in the context of stroke while the causal relationships between distinctive gut microbiota alterations and functional variations need to be further verified with well-designed studies.


Introduction
Stroke has been reported to be the major global health issue with an annual incidence of 258 per 100,000 person-years worldwide [1]. Patients suffering from stroke impose a heavy medical and economic burden on both society and the family [2]. With the medical advances in interventions for stroke (e.g., intravenous thrombolysis and endovascular treatment), the survival rate has been significantly improved while the disability rate increases [3]. Although medical treatment for stroke is essential for competing with the life-threatening condition in the acute phase, sequels left by stroke (e.g., motor dysfunction, cognitive dysfunction, or swallowing dysfunction) may subsequently impact the patients' health-related quality of life and increase the burden on families and society [4,5]. This indicates that attention should be, to some extent, moved forward to foresee the long-term functional prognosis; therefore, early modification of corresponding interventions can be provided. According to the literature review, several factors which may impact the poststroke functional recovery have been reported including age, gender, and admission National Institutes of Health Stroke Scale (NIHSS) score [6][7][8]. Nonetheless, these consolidated factors revealed only the individualized properties which could ever be changed or influenced [9][10][11]. Upon this condition, exploring novel 2 Neural Plasticity functional prognosis-associated factors, which may potentially serve as treatment targets, becomes the most warranted task in front of clinicians. Among the several newly proposed theories, investigation into the role of the "gut-microbiota-brain axis" in regulating neurological function has rapidly increased over the past decades [12,13]. Dysregulation of the gut-microbiotabrain axis has been increasingly linked to the pathophysiology of stroke [14,15]. Interactions across gut microbiota and poststroke functional outcomes were mainly observed in animal models [16,17]. Nonetheless, the role of human gut microbiota is indeed somewhat different from animals   Neural Plasticity [18]. The following concerns were the limited knowledge regarding the role human gut microbiota played with neural plasticity and the following prognostic changes. A few crosssectional studies have explored the composition of gut microbiota in patients with stroke compared with healthy demographics-matched controls. For instance, dysbiosis of short-chain fatty acid-(SCFA-) producing bacteria and SCFAs in patients with acute ischemic stroke was previously observed [19]. Alterations in trimethylamine-producing gut bacteria were proved to be associated with stroke [20]. However, these studies provided single-time point observations (e.g., data collected only at the acute phase) or enrolled only subtypes of stroke (e.g., middle cerebral artery occlusion). In addition, clinical trials remain limited in their ability to show the potential links between the gut microbiome, systematic and neural responses, and global functional changes in the context of stroke. Based on the above concerns and rationales, we conducted this study to answer the following two clinical questions: (1) should there be any discrepancies of gut microbiota in terms of either richness or composition between poststroke patients and healthy controls and (2) should there be certain potential correlations between gut microbiota alteration and global functions including general disability, physical function, swallowing function, and activity of daily living (ADL)? With the elaboration of the above two questions, our understanding regarding the role the "gut-microbiota-brain axis" played in the development of stroke would be improved followed by promoting novel treatment target selection in the context of stroke.

Study Design and Patient
Enrollment. The current cohort study was conducted at the First Affiliated Hospital of Nanjing Medical University from 04 July 2020 to 29 January 2021. It was approved by the Committee of Institutional Ethics (Institutional Review Board, 2018-SR-339), and all participants provided written informed consent prior to participation.

Functional Assessment and Sample
Collection. Demographic information including age, gender, BMI, blood pressure, smoking history, alcohol intake, physical activities, stroke subtype, medical history, and family history were collected according to a face-to-face interview or from electronic medical records. Functional assessments were performed by a research assistant with validated and reliable   Neural Plasticity scales. Specifically, stroke severity was assessed with NIHSS score; a higher score indicates greater stroke severity [26]. Modified Rankin scale (mRS) was used to measure the degree of disability and dependence in daily activities. mRS score ranges from 0 (no symptom) to 6 (death) with an unfavorable outcome scored 3-6 and a favorable outcome scored 0-2 [27,28]. ADL was evaluated with Barthel index (BI) ranging 0-100, with a lower score indicating higher dependence [29]. Upper or lower extremity motor function was assessed with Fugl-Meyer assessment (FMA) score such that a lower score indicates worse motor function [30,31]. Swallowing function was assessed with the water swallow test (WST); a higher grade indicates a higher risk of aspiration [32,33]. Fecal samples were collected from both groups and immediately immersed in a solution and stored at -80°C for subsequent DNA extraction.

DNA Extraction and Illumina
Sequencing. DNAs extracted from the fecal samples were used to amplify the V3-V4 region of the 16S rRNA gene targeted with primer set 341 F/806R. This action was performed to determine the gut bacterial community structure. The amplified products were further subjected to the library preparation and sequenced on the Illumina MiSeq platform according to the manufacturer instructions (Illumina technologies, USA).

Bioinformatics and Statistical
Analysis. The raw fastq files obtained from the Illumina sequencing machine were quality filtered with trimmomatic, vsearch, etc. Highquality sequence was used for community structure analysis through the QIIME pipeline. Operational taxonomic unit (OTU) picking was carried out with UCLUST closed reference method, and the representative OTUs were assigned taxonomy using UCLUST classifier by using the SILVA database (Version 132) as a reference dataset. Alpha and beta diversity indicates the gut microbial diversity of the different patients in the poststroke or control group assessed within (alpha diversity) and across (beta diversity) samples. Alpha diversity estimation was computed using the ace and Shannon indexes. Beta diversity was estimated with principal coordinates analysis (PCA), Bray-Curtis, and partial least squares discrimination analysis (PLS-DA). The Wilcoxon test was used to identify significantly differential OTUs (p < 0:05) for further analysis. Significant differences in the relative abundance of associated taxa between groups were further determined with linear discriminant analysis integrated with effect size (LEfSe). Random forest models (R3.4.1, randomForest 4.6-12 package) were performed to develop a predictive model. Receiver operating characteristic curve (ROC) and area under the curve (AUC) were used to evaluate the accuracy of models (R3.3.0, pROC package).
Continuous variables were presented as means and standard deviations. Categorical variables were demonstrated as numbers and percentages. Continuous nonnormal distribution variables were compared between groups with the Wilcoxon rank sum test. Fisher's exact test was used to compare intergroup differences for categorical variables. Correlations between discriminant bacterial and functional assessments were estimated with the Spearman correlation analysis.
Results of Bray-Curtis and PCA were also demonstrated in Figures 1(d) and 1(e). In addition, PLS-DA showed that microbiota composition in the poststroke group significantly differed from that in the control group (Figure 1(f)).
To explore potential biomarkers for the prediction of poststroke functional variation, a random forest model was   11 Neural Plasticity constructed based on the differentiated species (relative abundance > 0). Tenfold cross-validation analysis showed the AUCs of 95% (Figure 3(c)). As per the above analysis, the accuracy, specificity, and susceptibility of the optimal model consisted of the top 10 species (Figure 3(d)) were 93%, 100%, and 86%, respectively.

Discussion
In the current study, we firstly observed the higher alpha diversity and beta diversity of gut microbiota in poststroke patients as compared to those in the healthy controls, indicating the poststroke community richness and composition of gut microbiota differed from healthy controls. Afterward, a panel of microbiota was identified as biomarkers (e.g., Aggregatibacter segnis and Neisseria mucosa) to distinguish disease status. Furthermore, sensitivity analysis was performed to explore the gut microbiota alteration according to the length of stroke from onset (e.g., subacute and chronic) [34]. Significantly variated gut microbiota composition was observed along with the progress of stroke. Finally, we also demonstrated that general disability level, motor function either for upper limb or lower limb, swallowing function, and ADL were significantly associated with alterations of distinctive gut microbiota. Taken together, our results further provided evidence to support the point of view that gut microbiota varies following stroke. The hypothesized linkage between gut microbiota alterations and functional prognosis was preliminary observed while the causal relationships underlying these observations need to be further verified with well-designed animal and clinical studies.
Upon the great complexity of gut microbiota and huge heterogeneity of individual properties, studies into the role of gut microbiota on stroke were started with animal experiments to avoid the interference of confounding factors on outcome observation. By using two distinct mouse models of stroke, the overgrowth of microbiota (e.g., Firmicutes, Bacteroidetes, and Actinobacteria) was previously identified as biomarkers of poststroke microbiota dysbiosis, indicating that microbiota dysbiosis could serve as instruments to guide diagnosis and prognosis prediction of stroke [35,36]. Due to the discrepancy of gut microbiota between humans and animals, several studies on humans were initiated to better understand the poststroke gut microbiota alterations. For instance, a novel parameter termed as Stroke Dysbiosis Index (SDI) was proposed to discriminate stroke patients from healthy controls with AUCs ranging 74.9%-84.3% [37]. This is the first compound biomarker proposed to account for the poststroke gut microbiota alterations. However, the accuracy of the discriminative ability was not as high as in the current study. Here, we used the random forest model with which a panel of 10 significantly variated gut microbiota was detected with AUCs ranging 93%-95%. The accuracy of the discriminative ability was further improved with our model, and it provided reliable results for the following analysis.

Neural Plasticity
The poststroke gut microbiota alterations can be explained by several proposed theories. For instance, central stress responses induced reduction of gastrointestinal motility that may lead to bacterial overgrowth [38]. The autonomic nervous system has also been implicated in mediating the effects of stroke on dysbiosis [39]. Specifically, poststroke stress responses may lead to imbalanced activities of the autonomic nervous system which in turn increase intestinal permeability via releasing corticotropin and glucocorticoid hormones and consequently lead to gut bacterial translocation [40,41]. In addition, mouse model transplanted with poststroke fecal microbiota showed higher expression of the inflammatory T cells Th1 and Th17 indicating that systemic metabolic, immunologic, and inflammatory responses may play roles in the poststroke gut microbiota alterations [35,42]. To elaborate on the underlying mechanisms of the above responses, our results preliminarily demonstrated a decreased abundance of SCFAproducing bacteria (e.g., Blautia, Fusicatenibacter, and Ruminococcus) in patients with stroke. Liu and colleagues also found that a deficiency of SCFA-producing bacteria was significantly associated with poststroke cognitive impairment [43]. Transplantation of fecal microbiota rich in SCFAs was found to be effective in the improvement of stroke [44]. This protective effect was related to the enhancement of gut barrier integrity and attenuating systemic inflammation, which may improve function of the bloodbrain barrier, decrease cerebral edema, and attenuate brain injury [45]. This further verified the role of SCFAs in the development and prognosis of stroke; it may serve as a novel treatment target for stroke. With the clarification of specific mechanisms in animal steps, this flow also indicated that it needs to be moved forward to clinical circumstances for better facilitating the application of achievements from human trials.
In addition, poststroke variation of specific bacteria may have subsequent influences on regional organ responses. According to our results, an increased abundance of opportunistic pathogens was observed in patients with stroke. The potential role of opportunistic pathogens, such as Enterobacteriaceae, was reported to be associated with intestinal epithelial dysfunction [46]. Overgrowth of the Enterobacteriaceae may lead to inflammation exacerbation or exogenous pathogen invasion while internal homeostasis was disrupted [47]. These biological communications between the human body and gut microbiota may highlight the novel targets of stroke management. We also observed decreased potential probiotics. Based on the literature review, probiotics or a combination of probiotics and prebiotics could alter the composition of the gut microbiome followed by neuroinflammatory changes and cytokine releases. On the other hand, probiotics was demonstrated to increase brain-derived neurotrophic factor (BDNF) and inhibit apoptosis [48]. Taken together, the specific bacteria may influence the prognosis of stroke. Identification of this distinctive gut microbiota may contribute to the development of innovative treatment targets.
Although we observed the above promising variations of gut microbiota followed by the explanation of the potential internal mechanisms of subsequent consequences, the next question is that "should there be certain potential correlations between gut microbiota alterations and global functions" based on the newly discovered variations of gut microbiota across healthy and stroke status. The abundance of Christensenellaceae and Ruminococcaceae has been reported to be positively correlated with NIHSS score and mRS while Enterobacter was negatively correlated [49]. Our results preliminarily demonstrated that global functions, including general disability level, motor function, swallowing function, and ADL, were significantly associated with alterations of distinctive gut microbiota. There has been published a number of studies on exploring the relationship between microbiome and prognosis after stroke. Li et al. reported a negative correlation of Enterobacter with mRS score at one month [49]. Similar results have been also observed in our study. Apart from that, Prevotella was found to be correlated with motor function and ADL. The potential explanation would be that Prevotella was considered to be associated with mucosal inflammation, which may impact poststroke functional recovery due to the immune response [50,51]. However, the next challenge in front of us is to answer the question "whether the alterations of distinctive gut microbiota are the causes or consequences of poststroke neural plasticity." This would further improve our understanding of the role the "gut-microbiota-brain axis" played in the development of stroke and facilitate the treatment target selection.
The current study has several strengths and limitations. Firstly, we applied the random forest model to estimate the distinctive gut microbiota across stroke patients and healthy controls. This action further improved the accuracy of discriminative ability in detecting distinctive gut microbiota as compared to the previous studies. Nonetheless, the results should be further verified and validated in larger samples so that several subgroup analyses in terms of microbiota discrepancy could be performed. In addition, the current study preliminarily reported the potential correlations between gut microbiota alterations and global functions while the underlying mechanisms and the causal relationships between the gut microbiota, systematic and neural responses, and global functional changes need to be further investigated to build up the whole picture of the role the "gut-microbiota-brain axis" played in the context of stroke.

Conclusion
To sum up, poststroke gut microbiota was significantly modified as compared to healthy controls. The main characteristics of the stroke-induced shift in composition were the decreased abundance of SCFA-producing bacteria and the increased abundance of opportunistic pathogens. Significant associations were detected between alterations of distinctive gut microbiota and poststroke functional prognosis. A better understanding of the precise interactions between gut microbiota, systematic and neural responses, and global functional changes may be helpful in the identification of novel therapeutic targets to improve poststroke functional recovery.