The changes of gut microbiota in drug-naïve rst-episode MDD patients under treatment by SSRIs

Background: Recently, several studies reported that transplanting feces from depressed patients could induce depression-like behaviors in mice. In addition, antidepressants not only have antidepressant effects, but also have modulation of the gut microbiota in those animals. Therefore, this study rstly investigated on the changes of gut microbiota in depressed patients under effective antidepressant treatment. Methods: We recruited 30 patients with drug-naive rst-episode MDD (Patients group) and 30 healthy controls (Control group), and collected their fecal samples to complete 16S rRNA sequencing. Next, the Patients group received individualised treatment with escitalopram with a maximum dose of 20mg/d. After depressive symptoms improved to a HAMD scale score >50%, a second fecal sample was collected. This was classied as the follow-up group. We then investigated into the differences of gut microbiota between patients (Patients and Follow-up groups) and controls (Control group), the characteristics of gut microbiota under treatment, and the potential differences in metabolic functions. Results: A signicant difference in gut microbiota abundance was found after escitalopram treatment. The Firmicutes/Bacteroides ratio signicantly decreased in the Follow-up group. After treatment, the species diversity of gut microbiota tended to be back to normal state in Follow-up group. The mainly difference of metabolic function were found as follows: Transport and catabolism, Nervous system, Glycan biosynthesis and metabolism. Conclusions: Under escitalopram treatment, the gut microbiota diversity of MDD patients tended to back to normal state. However, several structures and metabolic pathways in microbes remained differences between patients and controls.


Background
As a common mental disorder accompanied by high disability and suicide, major depressive disorder MDD has become a worldwide issue [1]. In recent years, more and more evidences was reported to indicate that gut microbes might take part in the course of MDD [2][3][4]. In several paralleled researches, MDD patients showed speci c features of gut microbes differing from normal controls [5][6][7][8]. Several studies have reported that transplantation of feces from MDD patients into rodents could lead to depression-like behaviors, which had be used to establish animal models of MDD [5,9,10]. The similar phenomenon was observed in fecal bacteria transplantation between animal models. Moreover, it also observed that fecal bacteria transplantation could lead to the increase of microglial cell density and expression of IL-1 in the ventral hippocampus [11]. Taken together, the change of gut microbiota could be correlated with the occurrence of MDD.
At present, selective serotonin reuptake inhibitors (SSRIs) are widely used in clinical practice and have therapeutic effects in the treatment of depression [12]. Besides, several SSRI's drugs, including sertraline, uoxetine, paroxetine and escitalopram, could present antibacterial effects directly [13,14]. For example, staphylococcus and enterococcus are especially vulnerable to sertraline, uoxetine and paroxetine [14][15][16]. The SSRIs could target on potential enterotoxigenic bacteria as well. Therefore, SSRIs have demonstrated both of antidepressant and antimicrobial properties [17]. The mechanism underlying these interactive dual effects remains as an ongoing challenge.
Overall, SSRIs could directly or indirectly in uence the changes of gut microbiota which might play the key role in the development of MDD. The purpose of this study was to further exploring the difference of gut microbiota with rst episode MDD and compare the changes of gut microbiota after treatment by SSRIs.

Participants
The 60 local subjects, including 30 depressed patients and 30 normal subjects, were recruited in this study. The inclusion criteria for subjects described as follows: (1) Age was between 18-65 years old; (2) Body mass index (BMI) was between 18-28 Kg / m 2 ; (3) No history of treatment with antipsychotic medication; (4) Duration of symptoms was between 1 to 24 months; (5) No history of treatment with antidepressant medication; (6) Currently in the acute episode with the Hamilton Depression Rating Scale for Depression (HAMD) score ≥ 24; (7) Other mental disorders such as axis І, personality disorder and mental retardation were excluded; (8) Psychotropic drugs were never used; (9) Diagnosis of depression in Patients group was made by two psychiatrists according to the Mini-International Neuropsychiatric Interview (MINI) [18]; (10) In Controls group, a diagnosis of mental disorder was excluded by two psychiatrists according to the MINI, and HAMD-17 score was < 7; (11) Patients group was de ned as Follow-up group after receiving the drug treatment.
In addition, a series of exclusion criteria, based on the previous work, were employed in this study to exclude the factors affecting gut microbiota [18]. Those include: (1) No somatic diseases known to affect the gut microbiota such as in ammatory bowel disease, immune system diseases, diabetes, etc.; (2) Without antibiotics, probiotics or microbiological products used in recent 3 months; (3) Without the history of medical examination or surgery through the gastrointestinal tract in recent 6 months; (4) Without obvious changes in dietary habits or the presence of obvious diarrhea, constipation and other symptoms in recent 1 month.
According to the questionnaire, subjects' life events that may affect the mood, such as examinations, unemployment and bereavement during the last six months and the whole research period, were surveyed and recorded. All of subjects in this study were required to sign an informed consent. According to the Helsinki Declaration, the protocol for sample collection and analysis was approved by the Ethics Committee of Peking University Sixth Hospital and Beijing Hospital of Chinese Traditional and Western Medicine.

Sample collection
Fecal samples were obtained from subjects enrolled. Subjects were instructed by staff to discharge feces into a clean container. After defecation, the staff collected 2 g fecal sample and quickly placed it into a container containing liquid nitrogen. The samples were then frozen at -80 C until analysis.

Treatment
All patients with depression received individualised treatment with escitalopram. The starting dose of escitalopram was 5 mg / d from day 1-day 7 and increased to 10 mg / d from day 8. According to the individual response, the dose of escitalopram could be adjusted, and the maximum dose was 20 mg / d. After 4-6 weeks of treatment, the patients were evaluated by HAMD scale. When the scale reduction rate of HAMD is ≥50% compared with baseline, their fecal sample was collected for the second time and recorded as 'Follow-up group' to be used in the comparison of gut microbiota. cycles for 20 seconds, 58℃ annealing for 30 seconds, 72℃ extension for 30 seconds, and 72℃ nal extension for 5 minutes. The amplicons obtained by PCR were analyzed on 1.5% agarose gel electrophoresis, and a band of a desired size was puri ed using a QIAquick gel extraction kit (QIAGEN, Germany). The product was submitted to the second-generation sequencing laboratory of Beijing institute of bioinformatics for sequencing on Illumina HiSeq 2500 platform.

Bioinformatics analysis
The QIIME software was used to lter and sequence the original sequence to obtain optimized sequences (Tags) [19]. Fragments containing ambiguous characters in the sequence or more than two nucleotide mismatched primers were removed. Usearch software was used to cluster Tags at a similarity level of 97% to obtain OTUs [20]. OTUs were annotated based on the Silva (bacterial) and UNITE (fungi) taxonomy databases. QIIME software was used to generate species richness tables at different taxonomic levels, and R language tools were used to draw community structure maps at each taxonomic level of the sample. The community structure map of each sample was obtained at the level of taxonomy, class, order, family, genus, species.
In order to identify the difference in microbial community richness between the Patients group and Controls group, the Metastats software was used to perform a T test on the species richness data between two groups to obtain the p-value [21]. The q-value was obtained by correcting the p-value. Species were selected based on p-values or q-values that caused differences in the composition of the two groups of samples. The analysis was performed at the level of phylum, class, order, family, genus, species taxonomy to analyze the signi cance between groups.
Mothur (version v.1.30) software was used to evaluate the Alpha Diversity Index of the samples [22]. The species diversity within a single sample was studied by Alpha Diversity Analysis, and the Ace, Chao1, Shannon, and Simpson indices of each sample at the 97% similarity level were counted; Beta diversity analysis was performed using QIIME software. Beta diversity analysis mainly used the binary jaccard algorithm to calculate the distance among samples to obtain the β value between samples. Based on the distance matrix obtained from the Beta diversity analysis, PCoA analysis was performed using R language tools to further demonstrate the differences in species diversity among samples [23]. Hierarchical clustering was performed on samples using unweighted paired average method (UPGMA) to determine the similarity of species composition among samples. According to the species abundance table obtained by clustering, spearman correlation coe cient among samples was calculated by Psych package in R language, and then the heatmap was drawn by Pheatmap package in R language. The closer of the calculated Spearman correlation coe cient was to 1, the redder of the color was in the heat map, thus indicating the stronger correlation of two samples. Then, LEfSe tools were used the Wilcox.test function of the R language STATS package to estimate the impact of the abundance of each component (species) on the effect of the difference between components, so that the comparison of two subgroups can be realized to nd the species marker (Biomaker) with signi cant difference in the abundance.
PICRUSt software was used to compare the species composition information obtained from 16S sequencing data to deduce the functional gene composition in the samples, thereby determining the functional differences between different groups [24]. Using the KEGG orthology database (KOs) in the Kyoto Encyclopedia of Genes and Genomics (KEGG) database, the changes in metabolic pathways of functional genes of microbial communities between different groups were evaluated through differential analysis of KEGG metabolic pathways [25].

Statistics analysis
Statistical analysis was performed using SPSS19.0 software. Participants' gender, tobacco and alcohol consumption were expressed in terms of proportional or percentages. Independent t-tests, Welch t-tests, and White non-parametric t-tests were used for continuous variables. Pearson chi-square test or Fisher's exact test were used for classi cation variables. All signi cance tests were two-sided tests, and p <0.05 or adjusted p <0.05 was considered statistically signi cant.

Clinical data
In this study, thirty patients with drug-naive rst-episode MDD and thirty healthy controls were recruited, respectively. There were no statistically signi cant differences between patients' group and controls group in terms of age, height, weight, and tobacco and alcohol consumption ( Table 1). The Patients group received a individualised treatment and the maximum dose was 20 mg/d. The average dose of escitalopram was 16.33 ± 3.46 mg / d. The mean time for the HAMD score to be decreased by more than 50% following the commencement of escitalopram treatment was 34.53±5.18 days.  Table S2).
In addition, there were also several differences in the gut microbiota between Follow-up group and Controls group. In Follow-up group, the relative abundance of Parabacteroides, Prevotellaceae, Ruminiclostridium_6, Flavonifractor were signi cantly increased (p 0.05), while that of Prevotella_2, Lachnospira, Collinsella, and Clostridium_sensu_stricto_1 were signi cantly decreased (p 0.05). Moreover, the abundance of Faecalibacterium and Lachnoclostridium in both the Follow-up group and Patients group were signi cantly lower than the Controls group (p 0.05) (Supplementary Table S2).

Diversity analysis
Alpha diversity mainly re ected the richness and diversity of the species in samples. As shown in Figure  3, the Chaos1, Ace, and Shannon indices of Patients group were signi cantly higher than those of Followup group and Controls group, and the Simpson index was signi cantly lower in Patients group than others. This showed that the number and the diversity of gut microbiota in Patients group were signi cantly higher than those of Follow-up group and Controls group, and there were statistically differences. Four indices value of Follow-up group was between the values of other two groups, which was signi cantly different from that of Patients group but not statistically different from that of Controls group. This meant that the Alpha diversity of gut microbiota in patients returned to the normal level. The statistics of Alpha diversity index values of each groups were showed in Table 2. Beta diversity was used to compare the similarity of species diversity among different groups. The binary jaccard algorithm was used to calculate beta diversity, and there were statistical differences among three groups (R = 0.273, p = 0.001) (Figure 4). The gut microbiota of Patients group was signi cantly different from Controls group, and the gut microbiota within Patients group was more similar.
In addition, the gut microbiota pro les of some patients treated with escitalopram were more similar to those of the control group, but the others' pro les remained closer to those of patients. The other analysis methods employed in this study also produced similar results. The unweighted paired average method (UPGMA) was used in the R language tool to perform hierarchical clustering of each groups. It found that the gut microbiota of Patients group was signi cantly different from that of Controls group, and the gut microbiota of Follow-up group was more like that of Controls group (Supplementary Figure S3).
Spearman correlation coe cient between samples was calculated to draw the heatmap. The closer of the calculated Spearman correlation coe cient was to 1, the redder of the color was in the heat map, thus indicating the stronger correlation between two samples. As shown in Figure 4, the follow-up group could be divided into two subgroups. The gut microbiota pro les of some of the treated patients in the followup group remained similar to those in the patients group, while others' pro les were more like those of Controls group. These results suggest that antidepressant drugs may transform gut mirobiota of some patients into that of the control group. We used LEfSe for the quantitative analysis of biomarkers with in two subgroups (LDA>4). We found that several microorganisms could be selected as biomarkers in the two subgroups. Gut microbiota of follow-up group 2, that the subgroup was associated with Patients group, was differently enriched with p_Bacteroidetes, o_Bacteroidales, c_Bacteroidia, g_Prevotella_9, etc. While gut microbiota of follow-up group 1, that the subgroup was associated with Controls group, was differently enriched with p_Firmicutes, p_Actinobacteria, f_Lachnospiraceae, f_Bi dobacteriaceae, o_Bi dobacteriales, c_Actinobacteria, g_Bi dobacterium, etc (Supplementary Figure S4).

Functional properties predicted by PICRUSt
The study considered that the pro les of gut microbiota in Follow-up group could not completely return to the normal state. The PICRUSt software was used to compare the species composition information obtained from 16S sequencing data to infer the functional gene composition between patients and controls. Through the annotation of the KEGG metabolic pathway, it found that there were differences in the metabolic pathways of Transport and catabolism, Nervous system, Glycan biosynthesis and metabolism, Cell motility and Membrane transport between Follow-up group and Controls group (p 0.05) ( Figure 5). In Patients group, the pathways above also had certain differences with those of Control group but not signi cantly differences with those of Follow-up group. It suggested that the gut microbiota of Follow-up group might still involve in the occurrence of depression (Supplementary Table S5).

Discussion
This study demonstrated that the gut microbiota from patients of drug-naive rst-episode MDD was signi cantly different from that of controls, and the composition and structure of the gut microbiota within patients was more similar. This suggested that the occurrence and development of MDD might be associated with a special and similar group of gut microbiotas. The composition of gut microbiota in patients with depression have been found not consistent in several studies. This may be due to factors such as different inclusion criteria, ethnic differences and dietary habits, which in uence the nal sequencing results [26]. However, some studies have shown that transplanting fecal samples from depressed patients can induce depressive symptoms in mice [27,28]. This suggests that a disturbed gut microbiota may be one of the causes of depression.
Our study have found that the metabolic function of gut microbiota in the follow-up group was still different from that in the control group. The function of gut microbiota in depressed patients is abnormal even if the species abundance changes in a certain stage. In 2011, Manimozhiyan Arumugam et al.
proposed the concept of enterotypes [29]. In the sequencing results, they reported that the human gut microbiome could be divided into three robust clusters, including Prevotella-enterotype (enterotype P), Bacteroides-enterotype (enterotype B) and Ruminococcus-enterotype (enterotype R). Different enterotypes have characteristic advantages of respective functional states [30]. For example, enterotype P has stronger fermentation ability and can generate more short-chain fatty acids, but enterotype B has more speci c enzymes su ciently to degrade sugars and proteins [31]. Therefore, further research on the enterotypes and corresponding functional changes of gut microbiota would be helpful to understand the impact of gut microbiota on the development of MDD.
In addition, the gut microbiota tended to "normal" gut microbiota structure under SSRIs treatment, thus indicating a positive effect of SSRI antidepressants on the change of gut microbiota. However, the gut microbiota was still different from healthy controls in Follow-up group. On one hand, fecal samples from the follow-up group were not collected after the patients' symptoms had fully returned to normal. On the other hand, this results suggested that the recovery of the gut microbiota might take longer time than the improvement of depressive symptoms. The recovered gut microbiota might still partially maintain the pathological state, which would be a trigger for depression recurrence. Currently, there are no effective biomarkers to indicate recurrence in depressed patients. The dynamic changes of gut microbiota under antidepressant treatment provided the potential biomarkers for further investigation [32].
The outcomes of current study were limited to the relatively small sample size and short follow-up time.
In this study, we were unable to further observe the correlation between the effects of different doses of antidepressants on gut microbiota. In this study, the in uence of dietary factors on gut microbiota was also not fully considered. Different diets and microbial combinations have different effects on the physiological function and substance metabolism of the intestinal tract [33]. In future studies, the sample size can be further increased and collect fecal samples from patients at different doses of antidepressant. On the basis of exploring the changes of gut microbiota, we can further observe the correlation between gut microbiota and clinical phenotypes. Moreover, metabolomics and proteomics omics technologies are strongly suggested to be employed to elucidate the relationship between gut microbiota and depression.

Conclusion
Page 10/18 In conclusion, this study found that the gut microbiota of patients with rst-episode depression was similar and less different, and it was signi cantly different from healthy controls. After escitalopram treatment, gut microbiota diversity of depressive patients tended to transform to the normal state. However, there were still several structures and metabolic pathways difference in the gut microbiota between follow-up patients and healthy controls, which might be relate to the relapse of depression. Availability of data and material The datasets used and/or analyzed during this study are available from the corresponding author on reasonable request.
Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This study was supported by Beijing Association for science and technology, the Seed fund of Jinqiao project (NO. JQ18005). The role of funding body was in the design of study.
Author contributions YS, XY, JYG and YL designed the study and wrote the protocol. XXX and XY, JYG managed the literature searches and analyses. YS, ZYL and YCH undertook the statistical analysis, and YS, XY and GFL wrote the rst draft of the manuscript. All authors contributed to and have approved the nal manuscript.

Figure 1
Histogram of species distribution. QIIME software was used to generate species abundance tables at different taxonomic levels, and R language tool was used to draw community structure charts at different taxonomic levels. (A) relative proportions of species distribution at the phylum level; (B) relative proportions of species distribution at the genus level. PCoA analysis shows the distribution coordinate diagram of samples: In the graph, the distance between the dots represents the similarity of the samples. Samples with high similarity tend to cluster together.
The yellow dots represent the patient group, the blue dots represent the follow-up group, and the red dots represent the control group. The results showed a statistically signi cant difference among the three groups (R=0.273, p=0.001).

Figure 4
Spearman correlation coe cient heat map between samples. The closer the calculated Spearman correlation coe cient is to 1, the redder the color in the heat map, indicating the stronger correlation between the two samples. Follow-up group 1 had high correlation with Controls group. Follow-up group 2 had high correlation with Patients group.

Figure 5
KEGG metabolic pathway analysis. The left side of the gure shows the abundance ratio between the two groups. The middle section shows the proportional variation in functional abundance within the 95% con dence interval. The p-value is on the right.