Enteric phageome alternations in Type 2 diabetes disease

Background Type 2 diabetes (T2D) is a complex metabolic disease and has been proved to involve in the alternation of the gut microbiota. The previous studies primarily focused on the changes in bacteriome while ignoring the phage community composition. The extracellular phages could lyse the host bacteria, and thus in�uence the microbiota through the positive or negative interactions with bacteria. We investigated the change of extracellular phageome and explored its role in T2D pathogenesis. Results We used a sequencing-based approach to identify the bacteriophage after isolation of VLPs from the fecal samples. We identi�ed 330 phages according to the predicted host bacteria. The phageome characteristics were highly diverse among individuals. In the T2D group, the intestinal phage population is altered and the abundance of 7 identi�ed phages speci�c to Enterobacteriaceae hosts were found increased markedly. Additionally, the abundance of Enterobacteriaceae bacteria in gut was signi�cantly increased and the systemic LPS elevation was observed in T2D group. Several phage consortia were found to have signi�cant correlations with T2D disease indicators. Conclusions

homeostasis, regulating satiety, and producing energy [6,16].For example, low vitamin D production in gut has been associated with an increased risk of T2D [17][18][19] and butyrate produced by gut microbiota inversely associated with the degree of insulin resistance [20,21].Moreover, acetate has been proved to regulate the functions of islet β-cells via mediating a feed-back loop related to metabolic syndrome in microbiota-dependent way [22].Evidences from both human studies and animal trials have illuminated that T2D favors lipopolysaccharide (LPS) released from gram-negative bacteria translocating across the attenuated gut barrier, and results in a moderate increase in serum [23][24][25].LPS induced chronic low-grade in ammation and is associated with leptin and insulin resistance, which contributes to the establishment of T2D [26].
It is well known that bacteriophages, the dominant constituent of the virome, account for a high abundance in intestinal microbiota with phage-bacterial ratios of ~1:1 [27].Since phages are bacterial viruses which can lyse the speci c bacterial hosts after nishing their progeny replications, correlation between temporal population dynamics of phages and their hosts can be expected.The biological characteristics of phages endow their capability of regulating the abundance of their hosts, thereby affecting the structure of microbiota through the cascade reactions of both positive and negative interactions among the bacteriota.Afterwards, phages could in uence intestinal metabolome features through the changed bacteriota [28].On the other hand, gram-negative phage hosts could result in an increase in LPS.From the above, phage community deserves more attention in research of T2D, both for its contribution in regulating microbiota and in circulating LPS.
Since extracellular phages can infect and lyse bacterial hosts and directly in uence the composition of microbiota, we adopt the VLP isolation-dependent method and attempt to explore the changes in the gut extracellular phage community in a cohort with T2D in Shanghai, China.Besides that, we would like to explore whether these changes in the phageome have relationship with the changes in the bacteriome, then discuss the potential relationship between the alteration of phageome and the increased serum LPS.The design diagram is shown in Fig. 1.

Fecal phageome composition based on metagenome sequencing
A total of 46 sample including 17 T2D and 29 control were collected.On average, 11.7G (±2.14, SD) bases were sequenced for each sample.From those sequences, we obtained 3,136 phage scaffolds, with an average length of 7,257bp, resulting in a total of 330 species of phage were defined (Additional file 1).89.5% of the detected genomes of the phages were in form of double-stranded DNA (dsDNA), and less than 0.3% exist as single-stranded DNA (ssDNA).
The fecal phage composition is highly variable between individuals both in the T2D group and Ctrl group.There is a higher percentage of ssDNA in D17 (24.3%) than that in remaining samples (Fig. 2a).Caudovirales accounts for the most abundant phage community at the order level (89.8% on average).Especially in D03 sample, almost all of the phages detected were defined as Caudovirales(>99.9%).While in H18 sample, Caudovirales only comprises 64.1% of the phage population (Fig. 2b).Detailed information was shown in Additional file 2.
At the family level, 5 phage families including Myoviridae, Podoviridae, Microviridae, Siphoviridae, and Tectiviridae were detected.In most samples, the percentage of Siphoviridae was the highest with an average of 58.7%, followed by Myoviridae with 25.7%.Sample D03 and D15 contain almost only the family of Podoviridae and Siphoviridae, respectively.Moreover, the abundance of Microviridae detected in sample D17 and D08 was higher than that in other samples (Fig. 2c).
Furthermore, these detected phageome were aligned to 27 viral genera.Similar to the above, in the genus level, compositions of the virome in samples were highly diverse.Certain viral genus is dominant in several samples, which result in low diverse indexes.Such as the relative abundance of T1-like virus in D02, T4-like virus in D17, H18, and H24 samples, phiKMV-like in H12, and Lambda-like viruse in D07 and D09 are particularly high (Fig. 2d).
2. VLP identification revealed alterations of phage communities in type 2 diabetes patients α-diversity was assessed with 4 different diversity analysis methods.Both Ace and Chao index were significantly reduced in the T2D group, while Shannon and Simpson index showed no obvious changes between the two groups.
(Fig. 3a).For the results of β-diversity analysis, significant different phageome characteristics were observed between T2D patients and nondiabetic population based on principal components analysis (Fig. 3b).
The phages detected were classified according to their putative bacterial hosts.We defined a total of 330 species of phage with 51 putative bacterial hosts, among which 313 in T2D group and 324 in Ctrl group, 307 species were found in both groups (Fig. 3c).The abundances of these phages in the T2D and Ctrl groups were clustered to obtain a heatmap, as shown in Fig. 3d.

Phageome characteristics in T2D patients
Of the 330 species of bacteriophage identified sorting by their bacterial hosts, the abundance of 58 species of phage were significantly different between T2D patients and nondiabetic controls.There are 52 species with a cutoff of FDR<0.25, and four bacteriophages involving Brochothrix_phage_NF5, Enterococcu_phage_phiFL2A, Streptococcus_phage_PH10, and Streptococcus_phage_7201, with FDR<0.05 between the two groups (Additional file 1).The 18 phages with the most significant differences were showed in Fig. 4a.
After being clustered according to the level of the family, shifts of the abundance of Myoviridae, Podoviridae, Microviridae, Siphoviridae, and Tectiviridae in the T2D group versus Ctrl group were compared, no statistically significance was observed (Fig. 4b).
Host bacterial assignments for curated phage contigs were compared at genus level between the two groups, 7 host bacterial genera were remarkable changed (p ≤ 0.05), including Brochothrix, Klebsiella, Enterococcus, Bordetella, Shigella, Clostridium, and Tetrasphaer (Fig. 4c).The phages infecting Brochothrix and Klebsiella had the FDR <0.05 between the T2D group and the Ctrl group.

Fecal bacteriome features and the alteration in T2D patients
The rarefaction plots reached plateau for bacterial species, indicating the sample size is sufficient to reveal the bacterial population (Fig. 5a).The alpha-diversity presented by Chao1 index and PD whole tree index of the T2D group showed no significant changes (Fig. 5b).Unweighted Unifrac analysis of similarities (ANOSIM) displayed a significant difference in β-diversity between the two groups.Unweighted Unifrac principal coordinate analysis (PCoA) (left) and NMDS (middle) illustrated that the cluster of fecal bacteriome in T2D group is clearly distinct from that of the nondiabetic control.(Fig. 5c) The bacteria identified in this study was involved in 11 phyla.The detail information was shown in Additional file 3.
LEfSe is used for identifying the most varied abundant bacterial taxa at different levels, and for data analyzing and visualizing key species which are identified as differentiating factors between diabetes patients and health controls (Fig. 5d).With a log LDA score above 2.0, we found an enriched abundance of OTUs contributed by Phascolarctobacterium, Paraprevotella, Odoribacter, Clostridium XIVb, Butyricimonas, Shinella, Anaerofilum, Methanobrevibacter, and Streptophyta among health controls, while the T2D patients had increased abundance of Erysipelotrichaceae_incetae_sedis, Allisnonella, Lactobacillus, Dialister, and Megasphaera.Moreover, Enterobacteriaceae bacteria is enriched in the T2D group (Fig. 5e).129 bacterial genera were identified by 16S rDNA sequencing.Due to the technological limitations of 16S rDNA sequencing, among the 129 bacterial genera, 21 phylogenetic types bacteria failed to be identified to the exact genera.
Among the 105 genera, the abundance of 14 genera was significantly different between the two groups.Excluding Methanobrevibacter, Anaerofilum, and Shinella, of which the abundance was too low to be visualized in figures, we calculated the differences in abundance of the other 11 bacterial genera between the T2D group and Ctrl group.And the results were presented in the boxplot in Fig. 5f.

Correlations between Type 2 Diabetes-Associated Changes in the Phageome and Bacterial Microbiome.
In this study, the relationship between the phageome and the bacteriome was also investigated.Using 16S rDNA sequencing, 105 bacterial genera were identified.At the same time, the phageome we identified corresponded to 51 bacterial genera hosts.
Spearman correlation analysis was used to assess the correlations between the most abundant bacterial genera and bacteriophages in both T2D and control group.Moderately increased correlations of bacteria and phages were found in T2D group.(Fig. 6a) The bacteria detected and the hosts of bacteriophages could not completely overlap, only 16 putative host genera were identified in the detected bacterial group.These bacteria genera contained Klebsiella, Enterococcus, Lactococcus, Aggregatibacter, Pseudomonas, Pseudoalteromonas, Rhodococcus, Bacteroides, Bacillus, Corynebacterium, Staphylococcus, Streptococcus, Lactobacillus, Acinetobacter, Actinomyces, and Haemophilus.
Pearson correlation analysis was used to evaluate possible linear relation between bacteriophages and their host bacteria.The results indicated neither significant positive nor negative linear correlation between them (Fig. 6b).

Circulating LPS elevated in T2D patients
The LPS concentration in serum samples from T2D patients (n=13) and nondiabetic subjects (n=13) were assessed using tachypleus amebocyte lysate (TAL) based method.The results showed a significant increase in the LPS concentration of the T2D patients (Fig. 7c), as demonstrated in many studies [30,31].We hypothesized that the elevated level of serum LPS in T2D patients may result from the intensified lysis of the intestinal gram-negative bacteria under the action of phages.Based on this hypothesis, we analyzed the difference in the abundance of phageome mapped to gram-negative bacterial phages and gram-positive bacterial phages (Additional file 1).The results indicated that the relative abundance of gram-negative bacterial phages in the T2D group slightly higher than that in Ctrl group (Fig. 7a).While the gram-positive bacteriophages changed in the contrary direction.In addition, the most abundant 6 bacterial phyla Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, Proteobacteria, and Verrucomicrobia were analyzed to explore the changes of gram-positive bacteria and gram-negative bacteria abundances in the T2D group and the Ctrl group, respectively.However, no significant differences of the mentioned bacteria were observed in T2D group or Ctrl group (Fig. 7b).Enterobacteriaceae bacteria are typical gram-negative bacteria, and the abundance of them increased in the T2D group.In view of that, the alteration of the relative abundance of Enterobacteriaceae specific phages were further evaluated and showed an increase in the T2D group with statistically significant (Fig. 7d).

Increasing correlations between fecal phage community and T2D clinical indexes.
Correlations between the phageome and T2D disease indicators were evaluated.The results that comprehensively presented in table 1 indicated that there were some significant correlations between several specific phage consortia and 6 T2D indexes, which referred to fasting blood glucose (blood glucose0) , fasting insulin (insulin0), 0.5hrs insulin after meal (insulin30), 2hrs insulin after meal (insulin120), high sensitive-C reactive protein, and free thyroxine.

Discussion
As a typical disease of metabolic syndrome, T2D is known to involve complex cellular and molecular mechanisms, leading to dysregulated glucose homeostasis in the body, while microbiota has been proved to be one of the fundamental factors in uencing T2D development.Phage community are an important component of microbiota, containing both bacteriophages and the genetic elements of phages integrated in bacterial genome, representing the phage particles and prophages respectively, with the ability to modify the microbiota or participate in the horizontal transfer of host genes.When it comes to the technical routes for investigating phageome, a metagenome-dependent method and a VLP isolation-dependent method are in use [32][33][34][35].A previous study using the metagenomic data analysis has investigated the comprehensive DNA phageome containing both the genomes of phage particles and the genetic elements of prophages in patients with T2D for the rst time [36].In present study, we used a method based on virus-like particles isolation and focused on the extracellular phages, which could directly regulate the microbiota, thus to affect the physiology of the human host.Clinical studies have shown that phage ltrate can achieve good clinical results in patients with severe diarrhea caused by Clostridioides di cile [37].It also con rmed the regulation of extracellular phage on micro-ecological systems.
Viruses are divided into 152 families based on the latest (2019) report by the International Committee for the Taxonomy of Viruses (ICTV, Virus Metadata Repository: version September 9, 2019; MSL34).The ICTV report divided the RNA phages into only two families; Cystoviridae with genus Cystovirus, and Leviviridae, with 2 genera Levivirus and Allolevivirus [38].Despite the genome of phages could be DNA or RNA, we focused on the abundant DNA phageome in this study.The study above based on metagenomic analysis revealed an increase in the number of gut phages and an elevation of the relative pOTU number in the T2D group [36].Using the VLP isolation-based method in our study, no changes in abundance of the 5 phage families were found between the T2D group and Ctrl group (Fig. 4b).. Regardless of the differences in the structures of phageome between the two study cohorts, such divergence illustrated that the study results depend on the chosen research methods.Nevertheless, alternations of phage community were discovered when the phages were clustered according to the bacterial hosts (Fig. 4a, 4c)..
Remarkably, phages host in Klebsiella bacteria and Shigella bacteria that are the most concerned Enterobacteriaceae bacteria, increased in the T2D group.Then the changes in the relative abundance of all the detected Enterobacteriaceae speci c phages between the two groups were assessed, and showed a statistically signi cant increase in the T2D group (Fig. 7d)..Meanwhile, the bacteria belonging to family Enterobacteriaceae signi cantly increased in T2D patients compared with nondiabetic controls (Fig. 5e).. Family Enterobacteriaceae is a typical cluster of gram-negative bacteria and is known as a bacterial family enriched in opportunistic pathogens including Escherichia, Shigella, Klebsiella and Salmonella.It's well known that the LPS could be released from gram-negative bacteria, leading to the systematic subclinical in ammatory and, affecting the insulin sensitivity [39].Given a fact of that, we speculated that the alteration of gram-negative bacteria and their corresponding phages might lead to a circulating LPS elevation.
To con rm our hypothesis above, the relative abundance of phages whose putative host are gram-negative bacteria in the T2D group and Ctrl group were compared, and an increasing trend were found in the T2D group (Fig. 7a)..At the same time, the relative abundance of gram-negative bacteria in the T2D group were also evaluated, and showed no obvious changes (Fig. 7b).. Based on the experimental evidence above, we postulated a novel pathway via which the phage component in microbiota in uence the pathogenesis of T2D (Fig. 8).That is, an enhanced level of Enterobacteriaceae bacteria, and their speci c phages provided basis for the further intensi ed lysis of Enterobacteriaceae bacteria under pathological status of T2D, leading to an increase in serum LPS and then the development or aggravation of T2D.The structures of LPS in different bacteria differ a lot [40], so their pathological consequences also be various.
Since Enterobacteriaceae received more attention in the traditional infection disease, the possible pathogenic mechanisms linking Enterobacteriaceae bacteria-phages-LPS in T2D are supposed to be deeply explored.
The correlation analysis of the most abundant phages and bacterial genera showed a slightly strengthened correlation in disease status, and most of the correlations were positive correlations (Fig. 6a), which is inverse to a previous study about the phages in gut mucosal of ulcerative colitis [29].But the mechanism is not clear.The relative abundance of phages and their putative host bacteria in the fecal samples of neither T2D patients nor nondiabetic individuals showed no linear correlation (Fig. 6b)..That could be explained from the perspective of the dynamic relationship of bacteria-phage.Because phages not only live in bacteria, but also lyse their host bacteria, which have been described as 'predator-prey' dynamic model.Additionally, there are also 'kill-the-winner' and 'arms-race' dynamics at the same time [46].
In the last part of our study, we presented plentiful phages correlated with T2D disease indicators including fasting blood glucose, fasting insulin, 0.5hrs insulin after meal, 2hrs insulin after meal, hs-CRP, and free thyroxine (Table 1).Meanwhile, only a few bacterial genera have signi cant correlation with each T2D indicator (Additional le 5).Changes in disease-associated phageome might be more sensitive than those in the bacteriome.
It is urgent requirement that microbiome related researches should be closely integrated with clinical study, the indicator of bacteriome changes in disease status have been con rmed to be valuable auxiliary measures in clinical diagnosis [47], and expected to serve the precision medicine.It is undeniable that with the continuous improvement of the phageome researches and someday the completing of the standard operating procedure, phagesome variation may occupy a place in clinical diagnosis.

Conclusions
In this study, we investigated the alternations of intestinal extracellular phaegeome in patients with T2D.The data present in this study revealed the similar variation trend in enteric bacteriome and the correlated bacteriophages, the increase of bacteriophages predicted to infect Enterobacteriaceae in the gut was speculated as a novel origin of systemic LPS, which may contribute to the pathogenesis of the disease.Furthermore, we found some phage consortium related to disease indicators of T2D, which is likely to shed considerable light on overall understanding the interactions between microbiome and metabolic diseases.Methods Hypersensitive C-reaction protein (mg/L) 0.90±1.20

Diagnosis of diabetes
We used 2010 American Diabetes Association (ADA)'s criteria [48] for the diagnosis of diabetes.Diabetes was defined as fasting glucose ≥ 7.0 mmol/L, and/or 2-h glucose ≥ 11.1 mmol/L and/or HbA1c ≥ 6.5%.Those using antidiabetic drugs in the examination were also included into the type 2 diabetes cases.

Samples treatment for viral-like particles (VLP) isolation, and the DNA extraction
One aliquot of the stool samples was weighed to about 1.0 g and placed in sterile SM buffer (100 mM NaCl, 8 mM MgSO 4 •7H 2 O, 50 mM Tris-Cl (pH7.5),0.01% gelatin (w/v)).Then the fecal samples were vortexed until they were thoroughly mixed with SM buffer.The sample treatment process is based on previous studies [32,49,50].In brief, centrifuged the sample homogenates (3,500×g for 30min at 4 ºC).Followed by another centrifugation (10,000×g for 20min at 4 ºC) of the supernatants, the resulting supernatants was collected and filtered through a 0.45 μm and 0.22 μm Millipore filter sequentially.After collecting the filtrate, the samples were centrifuged (33,000rpm for 4h at 4ºC) in a SW41Ti swinging bucket rotor (Beckman).The 1.5 g ml -1 layer was recovered for the enriched bacteriophages [49].
This procedure was illustrated in Fig 1.
The 1.5 g ml -1 layer was collected from the step gradient and used for the source of the phageome.With the method described previously [32], the DNA was extracted from each sample, followed by the amplifying of bacterial 16S rRNA and eukaryotic 18S rRNA in one aliquot of the DNA to assess the absence of detectable non-viral DNA.
Before the DNA sequencing, the total DNA was amplified using GenomiPhi V2 kit (GE Healthcare) to generate enough genes for library construction.

16S rDNA gene sequencing
Total DNA of one aliquot of the fecal samples were extracted using QIAamp Fast DNA Stool Mini Kit.The 16S rDNA high-throughput sequencing was performed using the Illumina HiSeq PE250.Variable regions V3-V4 on 16S rDNA genes of bacterial genome were amplified with forward primer F341 5'-ACTCCTACGGGRSGCAGCAG-3' and reverse primer R806 5'-GGACTACVVGGGTATCTAATC-3'.The raw paired end reads were assembled by pandaseq with overlap nucleotides.Then, the reads were quality-filtered.The raw data were then subjected to a quality control procedure using UPARSE.The qualified reads were clustered to generate operational taxonomic units (OTUs) at the 97% similarity level using Usearch.Principal components analysis (PCA), heatmap analysis, Bray-Curtis similarity cluster, and species abundance analysis were performed using R program.Raw reads for each sample were preprocessed using the Trimmomatic [51].This included trimming adapter and removing reads with low-quality and insufficient length via the following parameters: SLIDINGWINDOW:4:5, MINLEN:50.The obtained clean reads were first mapped to human (hg19) reference genomes by bowtie2 [52].The resulting unmapped reads were further assembled into scaffolds by velvet software (version, 1.2.10, https://www.ebi.ac.uk/~zerbino/velvet/).The VLP scaffolds were performed using NCBI blast with the POGs database (1e-10).
. Circulating endotoxin levels measurement Serum endotoxin was assayed using a chromogenic limulus amebocyte lysate (LAL) test, which is a quantitative test for gram-negative bacterial endotoxin (BIOENDO, Xiamen).Gram-negative bacterial endotoxin catalyzes the activation of a proenzyme in the LAL.The activated enzyme catalyzes the splitting of p-nitroaniline (pNA) from the colorless substrate Ac-lle-Glu-Ala-Arg-pNA.The pNA released was measured photometrically at 405~410 nm following termination of the reaction.The correlation between the absorbance and the endotoxin concentration is linear in the 0.1~1.0EU/ml range.All samples were run in duplicate within the same plate.

Statistics analysis
Schematic

5 .
Metagenomic sequencing of DNA from Phages derived from VLP particles Subject phage DNA was first sheared into ~400bp-long fragments on Covaris S2 (Covaris, US).The resulting DNA fragments were used to construct sequencing library according to manufacturer's instruction (NEXTflex TM DNA Sequencing Kit compatible with the Biomek ® FXp(Bio Scientific, US).DNA library were sequenced on Illumina ® X-ten platform with a read length of 150 bp.
diagram of this study.The light-blue part is a sketch map of the preliminary exploratory herein, including exploring the relationship between phageome and putative host bacteriome, correlation between phageome and T2D disease indicators, discussion of the potential connection between the increase in serum LPS and the changes of bacteriophages.

Figure 7 Changes
Figure 7

Table 1
Phage communities with significant correlation with T2D disease indicators* Among the disease indicators in Additional file 4, only those with statistically significant correlations (FDR<0.05)were presented in this figure. *

Table 2
Human subjectsPatients with type 2 diabetes involved in this study were recruited from Shanghai Jiao Tong University Affiliated Sixth People's Hospital.Fecal samples and serum samples were collected, divided into two aliquots respectively, and stored in -80 ºC refrigerator.Healthy volunteers aged from 20 to 50 years old in control group were recruited from Shanghai Jiao Tong University Affiliated Sixth People's Hospital.All subjects in the nondiabetic control group were confirmed without diabetes by oral glucose tolerance test (OGTT), and excluded other metabolic diseases.Clinical characteristics of the T2D patients were presented in table 2. The detailed test results of nondiabetic controls are not shown in this study, while the relevant clinical features and demographic information recorded for each patient are listed in Additional file 4. Clinical characteristics of the T2D patients involved in this study 1.