A total of 131 nonduplicated GBS isolates were recovered during 2014–2021 at Ruijin Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, China. Isolates were grouped into three different types based on the clinical outcomes and isolation sources: invasive isolates (n=3) defined as those derived from normally sterile body sites; non-invasive isolates (n=65) defined as those derived from non-sterile sites and with clinical infection diagnosis of the isolation site; colonizing isolates (n=63) defined as those derived from vaginal swabs and urine samples from asymptomatic patients (most maternal colonizing strains). Detailed data of the isolates is presented in Supplementary Table 1 . All isolates were identified using MALDI-TOF MS (bioMérieux, Paris, France). All GBS isolates were grown in Todd-Hewitt broth (THB, BD) or Todd-Hewitt agar (THA) with 6% (v/v) sheep blood at 37°C in a CO₂ incubator.

Genomic DNA was extracted using W izard^® Genomic DNA Purification Kit (Promega, Madison, WN, USA) according to the manufacturer’s protocol. MLST was performed as previously described (Jones et al., 2003). Briefly, seven housekeeping genes (adhP, atr, glcK, glnA, pheS, sdhA, and tkt) were amplified and sequenced. The S. agalactiae MLST database (http://pubmlst.org/sagalatiae; accessed on October 2, 2021) was used for assigning allele numbers and obtaining the sequence types (STs). Using the eBURST program, the STs were grouped into clonal complexes (CCs) whose members shared at least five of the seven MLST loci (Francisco et al., 2009); otherwise, the ST was considered a singleton. A multiplex polymerase chain reaction (PCR) assay was performed to identify the capsular type as previously described (Imperi et al., 2010). Briefly, a multiplex PCR was performed using a mix of 19 primers and each serotype was identified by the analysis of the unique two or three bands pattern on agarose gel.

A total of 128 complete S. agalactiae genomes were retrieved from the GenBank database at NCBI (https://www.ncbi.nlm.nih.gov/genome/browse#!/prokaryotes/streptococcus%20agalactiae; with a filter of assembly level as complete; accessed on October 6, 2021) to analyze the presence of the T7SS loci orthologs in S. agalactiae ( Supplementary Table 2 ). Representative S. aureus strain COL (accession No.: NC_002951) and Streptococcus. suis strain GZ0565 (accession No.: CP017142) were also obtained and sequence alignment was conducted between the T7SS loci of these genomes by BLASTn and BLASTp. STs of all 128 strains were determined using the S. agalactiae MLST database (http://pubmlst.org/sagalatiae; accessed on October 6, 2021) and confirmed by a separate Web-based method on the basis of WGS data (www.cbs.dtu.dk/services/MLST) (Larsen et al., 2012). WGS-based identification of serotypes was conducted by employing the query DNA sequences and percent identity requirements as previously described (Metcalf et al., 2017) and confirmed using a serotyping database (https://github.com/swainechen/GBS-SBG) (Tiruvayipati et al., 2021).

Comparative analysis and annotations of genome sequences were performed to detect T7SS between two clinical GBS strains Sag37 (accession No.: CP019978.1) and Sag158 (accession No.: CP019979.1) previously described (Zhou et al., 2017). The T7SS sequence in Sag37 (accession No.: CP019978.1; bases from 1,149,232 to 1,161,088) was used to identify the T7SS clusters in S. agalactiae through homology screening using the BLASTp and BLASTn against the NCBI database followed by careful manual inspection. Sequence analysis and comparison of genomes led to the identification of the related copies of T7SS. BLASTp and ClustalW were used to align and calculate the average amino acid sequence identity (Larkin et al., 2007).

Assembled genomes in the public data set were mapped to a reference (Sag37) genome and single nucleotide polymorphisms (SNPs) were called using snippy v3.1 (https://github.com/tseemann/snippy). The phylogenetic tree was constructed using core SNPs alignment and visualized using the Interactive Tree of Life (iTOL) (https://itol.embl.de/) (Letunic and Bork, 2019).

PCR-based screening was performed to distinguish the different T7SS types and estimate their distribution in the clinical strains, as well as to determine the ST-dependent diversity observed in the isolates obtained from the GenBank database. The essC-1, essC-2, essB, esaA, and esxA genes were amplified to elucidate the general structure of the T7SS apparatus in the clinical strains. Different T7SS types were indicated by different PCR-based patterns. Specifically, the amplified sections of all five genes were present in Types A-D. Types E and F showed no amplification of esxA while the other four genes were identified; this was defined as pattern B. Type K was defined by the lack of amplification of all five genes. Types G and H were also distinguished through PCR amplification. Primers used in the present study were designed based on the conserved sequences of T7SS observed in the alignment of the publicly available genomes ( Supplementary Table 3 ). DNA isolation, amplification, and electrophoresis of the PCR products were carried out as reported earlier (Hynes et al., 1992). The PCR products were analyzed by sequencing and BLASTp.

Genomic DNA of 17 representative clinical strains was extracted using Wizard^® Genomic DNA Purification Kit (Promega, Madison, WN, USA) according to the manufacturer’s protocol. Purified genomic DNA was quantified by a TBS-380 fluorometer (Turner Biosystems Inc., Sunnyvale, CA, USA). Next, approximately 1 μg genomic DNA was used from each strain to construct sequencing library. DNA samples were sheared into 400–500 bp fragments using a Covaris M220 Focused Acoustic Shearer (Trendbio, Melbourne, Australia) following the manufacturer’s protocol. Sequencing libraries were prepared from the sheared fragments using the NEXTflex™ Rapid DNA-Seq Kit (Tecan, Canton of Zürich, Switzerland). The prepared libraries were used for paired-end Illumina sequencing (2 × 150 bp) on an Illumina HiSeq X Ten machine (Illumina, San Diego, CA, USA). The original image data were transferred into sequence data via base calling, defined as raw data or raw reads and saved as a FASTQ file. The clean reads were assembled using the SOAPdenovo2 software (Koren et al., 2017). Glimmer (Delcher et al., 2007) was used for coding sequence (CDS) prediction, tRNA-scan-SE was used for tRNA prediction, and Barrnap was used for rRNA prediction. The predicted CDSs were annotated from NR, Swiss-Prot, Pfam, GO, COG and KEGG database using sequence alignment tools such as BLASTP, Diamond and HMMER. Furthermore, each set of query proteins were aligned with the databases, and annotations of best-matched subjects (e-value < 10-5) were obtained for gene annotation.

Statistical comparisons were performed via Fisher’s exact tests, as well as Chi-square and Cramer’s V analyses, considering values of P < 0.05 as statistically significant. Specifically, Fisher’s exact tests were performed to claim significant associations between T7SS subtypes and STs. Correlation between T7SS types and host range was determined by Fisher’s exact tests, as well as Chi-square and Cramer’s V analyses. All tests were performed using SAS software (version 9.4).

The results of the whole-genome shotgun sequencing projects were deposited at DDBJ/ENA/GenBank under the corresponding accession numbers ( Supplementary Table 4 ).

This study was approved by the ethics committee of Ruijin Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, China and the Review Board exempted the requirement for written informed consent because this is a retrospective study only focused on bacteria and did not affect the patients.

The T7SS apparatus was detected in the clinical GBS strain Sag37 isolated from the blood sample of neonate suffering from bacteremia (Zhou et al., 2017). The T7SS cluster was arranged in the order of essC-essB- esaB-essA-esaA-esxA-esxA (duplicated esxA), presenting insertion sequence IS5 in the essC gene ( Figure 1A ). Homology analysis demonstrated that the putative proteins encoded by the T7SS genes had limited similarity with those found in the S. aureus strain COL and S. suis strain GZ0565 ( Figure 1A ). The amino-acid identities of the proteins encoded by the T7SS genes varied between 36&ndash;59% in Sag37 vs. S. suis GZ0565 and between 24&ndash;46% in Sag37 vs. S. aureus COL; however, no significant similarity was observed in essA between the strains. Staphylococcus aureus possessed a more complicated T7SS apparatus than S. species, with eleven type VII system-associated genes (Burts et al., 2005; Anderson et al., 2013). The essD, esxD, esxB, and esxC genes found in the S. aureus strain were absent in S. agalactiae and S. suis.

A total of 128 complete S. agalactiae genomes were retrieved from the GenBank database to build a public data set and investigate the T7SS gene cluster and its genetic environment in S. agalactiae. Thirty STs were included and arranged into six major CCs groups (CC1, CC7, CC10, CC19, CC61, and CC260) except for three non-typeable strains as only six housekeeping genes were identified ( Table 1 ). CC260 was the most predominant group (30.4%), followed by CC1 (13.3%) and CC19 (7.8%) ( Table 2 ). Furthermore, 43 isolates (33.6%) were grouped as singletons. Serotypes of 128 strains included Ia, Ib, II, III, IV, V, VI, and VII. In addition, four non-typeable isolates were identified. Serotype Ib was the most predominant (39.1%), followed by III (23.4%) and V (13.3%), respectively. 62 strains were animal-derived with the majority of ST260/261/283 (mainly fish), while 56 strains were human-derived with the majority of ST1/12/17/19/23/283.

The screening for the T7SS coding regions within S. agalactiae strains demonstrated that T7SS was harbored on chromosomes and not on plasmids. Screening yielded eleven different types of genetic characteristics of T7SS that displayed similarities in gene order but differences in gene content ( Figure 2A ). The clusters with coverage and percent identity greater than 99% were grouped into the same type.

T7SS in Sag37 shared significant similarity with Type A except for the disruption of insertion sequence IS5 in essC. Type B had only one esxA, while Sag37 and type A had two tightly duplicated esxA. Type C was similar to Type B except for an IS1182 family transposase locating downstream of the esxA. Compared with Type C, Type D had an extra hypothetical protein downstream the IS1182 family transposase. Incomplete T7SS clusters with obvious deletions were also observed during comparison. A ~300 bp deletion at the end of the esaA was observed in Type E, with deletion of subsequent esxA as well. Type F had more deletion (~550 bp deletion) up to the flanking gene downstream. Distinct essC variants was detected in Type F and G with a ~650 bp difference in the C-terminus from the aforementioned types. Truncated T7SS was also observed. Incidentally, strains completely lacking T7SS were also discovered (Type K, not shown in Figure 2A ).

Among all the types analyzed by us (Sag37 not included), type A formed the majority (36.2%), followed by Type K with 29.1% ( Figure 1B ). The proportion of both Type B and J exceeded 10%, whereas Type D/E/G/H/I appeared in only one representative strain in our data set.

Putative T7SS effector-encoding genes were detected downstream essC, together with genes encoding DUF4176 domain-containing proteins and LXG domain-containing proteins. Although the proteins encoded in this region contained similar conserved domains/motifs, the downstream flanking nucleotide sequences were of distinct patterns. Types A and E were of the same pattern, while Types B, C, and D were of another. Aside from the major patterns illustrated in Figure 2A , five subpatterns were observed with the lack of genes coding for DUF4176 domain-containing proteins and/or LXG domain-containing proteins ( Figure 2B ). Moreover, a distinct A-5 pattern was observed with no homology to other patterns. Four different T7SS effector-encoding genes were detected in the public dataset. Although no significant similarity was found in the nucleotide sequences between A-5 and A-1, similar conserved domains/motifs were encoded in this region. Percent identities between the T7SS effector among the different subtypes were within 41-75% ( Figures 2C, D ). Similar structures were absent in types with essC variants (Type E and F), and the T7SS effector-encoding gene was absent in Type H. Compared to the diversified downstream flanking genes, upstream flanking genes displayed few differences.

Among 131 clinical GBS strains analyzed in this study, 22 STs were included and grouped into five major CCs (CC1, CC10, CC17, CC19, and CC23) ( Table 1 ). CC10, CC19, and CC23 account for 76.4% of strains ( Table 2 ). Nine isolates (singletons) including ST529, ST862, and ST226 were not related to any of the five CCs. Serotypes of all the 131 strains included Ia, Ib, II, III, IV, and V. Serotypes III was the most predominant (30.5%), followed by Ia (22.1%), and Ib (23.7%).

The PCR screening of the clinical strains demonstrated three patterns of T7SS clusters (see Figure 1C ). Strains with identical STs showed similar results. Specifically, 84 isolates (64.1%) harbored all five genes present in Types A-D (see Figure 1D , Pattern A). Thirty-eight isolates (29.0%) harbored none of these genes (see Figure 1D , Pattern C). Nine isolates (6.9%) displayed pattern B. Patterns indicating Types G and H were not detected through PCR amplification.

The specific number of esxA genes could not be determined due to the limitation of the PCR. Hence, we performed genome sequencing; 17 strains presenting different patterns were selected and sequenced (see Supplementary Table 4 ). Further analysis revealed four T7SS types similar to that of the strains obtained from the GenBank database; strain 189 (ST10), which displayed type A T7SS rather than type C or E found in the GenBank, was the exception.

Characterization of the T7SS gene clusters showed a possible relation to the STs. Strains with identical STs in the public dataset, collected in different study covering a wide geographical range and time span, were found that could be grouped into the same T7SS types. The phylogenetic tree constructed based on the core SNPs alignment further confirmed that genetically closely associated strains had the same type of T7SS system ( Figure 3 ). Exceptions were also found. Therefore, Fisher&rsquo;s exact tests were performed to identify correlations. Some of the STs were represented by a single genome in the database, which could only bear a single T7SS type. Therefore, STs that appeared only once were excluded from comparisons. Significant associations were found between most strains and types displayed in Figure 2A , including Type A with ST1/7/103/283, Type B with ST23/110/335, Type F with ST17, Type J with ST260/927, and Type K with ST19/552. For some of the less represented STs carrying multiple T7SS types, e.g., ST10, ST12, ST22, ST261, no correlations were found. No significant correlations were found between CCs/serotypes and T7SS characterization.

In our public dataset, 56 strains were human-derived while 62 strains were animal-derived. Correlation between T7SS types and host range was identified by Fisher&rsquo;s exact tests. To further evaluate the relationships between T7SS types and host range, Chi-square and Cramer&rsquo;s V analyses were performed between the T7SS subtypes discovered in both origins. T7SS subtypes demonstrated strong positive correlations with the host range (p &lt; 0.05, Cramer&rsquo;s V=0.538). Statistical significance was claimed by adjusted residuals that human-derived GBS was associated with complete T7SS apparatus (Type A and B), whereas animal-derived GBS with an incomplete one (Type K).