This study was performed using a dataset mined from the National Center for Biotechnology Information (NCBI) database (available at https://www.ncbi.nlm.nih.gov/genome/?term=Streptococcus+pneumoniae). The latter contains whole-genome sequences (assembled or otherwise) for more than 8500 pneumococcal genomes (last accessed January 30, 2021). Sequence types (STs) (Enright and Spratt, 1998; Jolley et al., 2018) of the strains included in the dataset were determined on the basis of whole genome sequencing data (Larsen et al., 2012).

To detect PPHs, the dataset was searched, using the BLAST platform (Johnson et al., 2008), for homologs of int- and lytA _PPH-like genes encoding (respectively) Ints and endolysins [with respect to the latter, only N-acetylmuramoyl-L-alanine amidases (NAM-amidases; EC 3.5.1.28) of the Amidase_2 family (Pfam database identifier: PF01510) (Morales et al., 2010) are currently known]. The corresponding genes/proteins of previously characterized PPHs, such as MM1 (Obregón et al., 2003a), фSpn_OXC, фSpn_6, and фSpn_18 (Romero et al., 2009a), and others (Càmara et al., 2018). The endolysins encoded by various virulent pneumococcal phages (see below) were used as query sequences. The original and newly found genes/proteins with a coverage ≥ 90% and ≥ 50% identity were used in searches in an iterative manner.

Sequence comparison and alignments were performed using the BLAST platform and/or Clustal Omega package (Sievers and Higgins, 2021) running at the European Bioinformatics Institute (EMBL-EBI) website (https://www.ebi.ac.uk). Potential tRNAs genes were identified using ARAGORN (Laslett and Canback, 2004) and tRNAscan-SE (Schattner et al., 2005) software. Detections were recorded only when both programs coincided in the results returned. The codon usage of S. pneumoniae D39 was obtained from the corresponding database (Nakamura et al., 2000). Protein domains were preliminarily identified using CDD/SPARCLE (Lu et al., 2020). PHASTER (running at https://phaster.ca/) was also employed to identify phage-homologous regions in bacterial genomes (Arndt et al., 2016). To ascertain equivalent detection of protein datasets, open reading frames were identified and translated from prophage sequences using Prodigal v2.6.3 (Hyatt et al., 2010). Prophage proteomes were then compared all against all through the calculation of the weighted gene repertoire relatedness (wGRR) (Pfeifer et al., 2021). Best bi-directional hits were found from hits detected by the ‘easy-search’ workflow of MMseqs2 (Mirdita et al., 2019) with e-value < 10⁻⁴, identity ≥ 35%, coverage ≥ 50% thresholds. Agglomerative hierarchical clustering from the resulting wGRR matrix was carried out by the linkage tool of the scipy.cluster.hierarchy python library using the ‘ward’ method on euclidean distances.

The dataset produced contained 4003 strains, including 126 whose genomes were sequenced to complete or near-complete (chromosome) assembly level and deposited in the RefSeq database (O'Leary et al., 2016) (last accessed, October 11, 2020) (labeled in green in Table S1 ). The strains included in the dataset represent 447 different STs. Moreover, up to 50% of the strains belonged to one of 35 (out of a possible 43) Pneumococcal Molecular Epidemiology Network (PMEN) clones [including single and double locus variants (McGee et al., 2001)]. PMEN clones are resistant to one or more antibiotics in wide clinical use and dominate the population of antibiotic-resistant pneumococci. Globally susceptible clones known to be important in disease are also included in the PMEN clone dataset (https://www.pneumogen.net/pmen/, last accessed September 15, 2020). Further, 2557 strains in the dataset were assigned to one of the 169 Global Pneumococcal Sequence Clusters (GPSCs) and, among these strains, 1096 could be classified as belonging to one of the 35 dominant GPSCs (Gladstone et al., 2019). From a total of 46 serogroups (numbered 1–48) described to date (numbers 26 and 30 are not in use) (Lund and Henrichsen, 1978), 3573 strains belonged to one of 40 serogroups. In addition, the dataset included 4 non-encapsulated laboratory mutants and 426 non-typeable (NT) isolates ( Table S1 ).

In agreement with data reported in previous studies (Ramirez et al., 1999; Brueggemann et al., 2017), PPHs were seen to be widely distributed across different S. pneumoniae isolates. Indeed, only 434 strains in the dataset (10.8%) (highlighted with a red background in Table S1 ) appeared to lack temperate phages. In addition, among the 126 RefSeq strains with complete (or near complete) genomes (see above), only 26 (20.6%) lacked any discernible prophage. On the basis of their Ints, more than 7000 putative PPHs (including putatively full-length and partial prophage sequences) were found in the dataset. Visual inspection of bacterial genes flanking the prophages revealed up to 26 different integration sites ( Table 1 ). PPHs were clustered into 26 groups [named from PPH005 to PPH130 (in steps of 5)] according to their purported insertion sites, and arranged in order using the genome of the non-lysogenic S. pneumoniae D39 (Acc. No. NC_008533.2) as a reference. It is worth noting that many more complete Ints (7234) than endolysins (1210) could be identified. In addition, the proportion of partial sequences was much higher for the endolysin genes than the Ints genes. Thus, 42% of the endolysins were found incomplete (887 incomplete out of 2097 total proteins), whereas only 7% of the Ints were apparently incomplete or partly deleted (548 out of a total of 7782) ( Table 1 ). This might be because endolysin-coding genes are among those known to be difficult to assemble from short read data due to the presence of choline-binding motif clusters (Croucher et al., 2017).

In sharp contrast to the moderate number of endolysins compared to Ints in the current dataset, the number of different PPH endolysin alleles (362) was similar to that of the Int alleles (293). This strongly suggests that the PPH endolysin-coding gene shows great genetic variability. Nevertheless, 11 PPH groups, namely 005, 020, 025, 030, 045, 055, 060, 070, 095, 110, and 115, appeared to completely lack an endolysin gene and thus were understood to represent incomplete (or defective) prophages. Taking into account the number of full Ints, these incomplete prophages correspond to nearly half (47%) of the total PPHs.

On the basis of amino acid (aa) sequence similarities and catalytic residues, site-specific Ints (recombinases) can be classified into two major families: the tyrosine or serine families (Groth and Calos, 2004). These have different structures and functional mechanisms indicating that they have evolved separately. Thus, serine Ints are usually larger than tyrosine Ints, and the aa residues important for catalysis and structure, as well as their locations, can be very different. For example, the N-terminal domain of tyrosine Ints is involved in binding the arm-type sites of attP and the C-terminal domains involved in catalysis, whereas in serine Ints the opposite is true (Groth and Calos, 2004; Van Duyne and Rutherford, 2013).

Table S2 shows the complete catalog of the Ints found in this study. Excluding the prophages of the PPH090 group, which will be analyzed separately, only three Ints were found in more than one PPH group. These corresponded to the most frequent type of each group, i.e., WP_00876735 (in PPH010 and PPH130), WP_00266847 (in PPH040, PPH075, and PPH080), and WP_000704678 (in PPH055, PPH095, and PPH110). This observation agrees with the common idea that strong (albeit not absolute) specificity exists between prophage Ints and attB/attP sequences. For example, it has long been recognized that Int-mediated recombination occurs between attP of phage λ and the secondary attachment sites in the host genome, although the frequency is low (Shimada et al., 1975). In the present work, sequence identities between Ints were usually >90% within each PPH group. The PPHs contained either tyrosine or serine Ints, although PPHs with tyrosine Ints were 6.5 times more common (6243) than the latter (957) in the dataset (the Ints of PPH090 were not taken into account) ( Table 2 ). Serine Ints (groups PPH085, PPH100, PPH120, and PPH125) were much more similar to each other than were the tyrosine Ints ( Table 3 ). Among the tyrosine Ints, those from groups PPH010/PPH130 and PPH035 diverged greatly from the rest.

Endolysins encoded by virulent (lytic) phages infecting S. pneumoniae have been extensively reviewed (López and García, 2004; García et al., 2005; Galán-Bartual et al., 2015; Maestro and Sanz, 2016; Vázquez et al., 2018). Briefly, three biochemically and structurally different peptidoglycan hydrolases of phage origin have been described so far. Two of them are choline-binding proteins: the NAM-amidase Pal (YP_004306947; 296 aa), with an Amidase_5 domain (PF05382), and the Cpl-1 lysozyme (=muramidase) (NP_044837; 339 aa) with a Glyco_hydro_25 (PF01183) domain. Both enzymes contain a C-terminal cell wall-binding domain composed of six choline-binding repeats (Choline_bind_1; PF01473). Cpl-7 (YP_009623604; 342 aa) is a choline-independent muramidase that differs from Cpl-1 in the C-terminal domain responsible for peptidoglycan binding (CW_7 repeats; PF08230). Pal is encoded by phage Dp-1, whereas Cpl-1 and Cpl-7 are the endolysins of phages Cp-1 and Cp-7, respectively. The endolysin from a recently isolated virulent podovirus very similar to Cp-1 (SOCP) (Ouennane et al., 2015) (NP_044837; 339 aa), is a muramidase 100% identical to Cpl-1. Similarly, the NAM-amidase (AQY55407; 295 aa) of the MS1 pneumococcal siphovirus (Kot et al., 2017; Silva et al., 2020) is 81% identical (89% similar) to Pal.

Taking into account that only a few virulent S. pneumoniae-infecting phages have been currently isolated, the diversity of their endolysins contrasts with the lytic enzymes encoded by PPHs. The latter are 318 aa-long, belong to the Amidase_2 family (PF01510) of NAM-amidases, and contain six Choline_bind_1 repeats in the C-terminal domain (Morales et al., 2010; Morales et al., 2015). Remarkably, in addition to having the same length (957 bp), the known phage genes coding for these endolysins (lytA _PPH) are closely related (85–92% identity) to lytA_Spn , which encodes the major pneumococcal autolysin —also a NAM-amidase of the Amidase_2 family (Morales et al., 2010). LytA _Spn is a well-known virulence factor that plays a role(s) during different steps of infection (Canvin et al., 1995; Ramos-Sevillano et al., 2015; Ramos-Sevillano et al., 2016; Corsini et al., 2021). Taking into account the strong similarities between bacterial and phage NAM-amidases, the latter may also be important in pneumococcal pathogenesis.

The current notion of an exclusive presence for lytA-like genes among PPHs could not be fully confirmed in the present study. As already mentioned, 1210 full-length endolysins were identified in the dataset ( Table 1 and Table S3 ). Among those, 29 proteins (≈2.5%) did not correspond to NAM-amidases of the Amidase_2 family. There were 28 proteins homologous to the Cpl-1 lysozyme, with one containing a cysteine, histidine-dependent amidohydrolase/peptidase (CHAP) domain (PF05257) instead of an Amidase_2 domain at the N-terminal moiety. The lysozymes were endolysins of PPH065 (22 identical proteins; WP_000739159) and PPH105 (6 proteins; 4 alleles: WP_054365577, WP_054368721, WP_054380492, and WP_054392100), which are 314 aa- and 334 aa-long respectively. Notably, all strains harboring PPH065 are members of GPSC19 and have serotype 22F (the only exception is strain 2245STDY6178828 which belongs to serotype 42) (see below).

The main differences between Cpl-1 and the PPH lysozymes detected in the present study are located in the linker region connecting the N- and the C-terminal domains. Thus, in Cpl-1, the linker (189-DDEEDDKPKTA-199) (Hermoso et al., 2003) is longer than (and different to) that of the newly discovered endolysins (188-DDEEAKAK-195). Moreover, WP_000739159 lacks the fourth Choline_bind_1 repeat that forms part of the C-terminal domain of the enzyme, and which is responsible for binding the enzyme to the cell wall. A detailed analysis of other PPHs (156 additional genomes, either complete or not) revealed four additional examples of 334 aa-long Cpl-1 homologs, specifically those encoded by prophages IPP16, IPP25, IPP27, and by a nameless prophage harbored by S. pneumoniae strain R34-3131 ( Table S4 ). The latter prophage (which belongs to the PPH105 family) was already present in the dataset ( Table S1 ).

The CHAP domain-containing endolysin (WP_057595562; 288 aa) found in the genome of S. pneumoniae strain SMRU392 (serotype 35F) is encoded by one of the PPH080 group of phages, and is identical to that of prophage 33888. The sequence of the latter was recently reported in an independent study (van Tonder et al., 2019) ( Table S4 ). Since the host strain of prophage 33888 was the same as in the present study (SMRU392), and the Ints were also identical (WP_ 050256063), both prophages are probably the same. The WP_057595562 PPH endolysin is 63% identical (77% similar) to Skl, a proven NAM-amidase (WP_033686260; 288 aa) of an unnamed temperate phage of Streptococcus mitis SK137 (Llull et al., 2006). The complete genomic sequence of, presumably, the same S. mitis prophage (now designated as Javan331; MK448732) has recently been reported (Rezaei Javan et al., 2019). Remarkably, the existence of putative CHAP-endolysins encoded by those prophages was not mentioned in any study from other laboratories.

Among the 1181 endolysins with a predicted Amidase_2 domain, the presence of one atypical protein was noted. This protein (WP_130892444; PPH120) is a 328 aa-long, putative NAM-amidase with a divergent Amidase_2 domain. It is an endolysin encoded by the 36.7 kb-long prophage inserted into PPH010_4, giving rise to PPH120 (see below). Searches for proteins very similar (≥87% identity) to WP_13089244 revealed several endolysins of the same length encoded by prophages of three members of the Mitis group streptococci, i.e., Streptococcus pseudopneumoniae, S. mitis, and Streptococcus oralis. Specifically, these prophages are 277_SPSE, 289_SPSE, and 380_SPSE (WP_049511163) from S. pseudopneumoniae (Roach et al., 2015), S. mitis strains SK564 (WP_000238871) (Kilian et al., 2014) and DD22 (WP_061864892) (Denapaite et al., 2016), S. oralis U-o11 (Javan367; QBX17588) (Rezaei Javan et al., 2019), and S. oralis subsp. tigurinus 859 (WP_084868230) (Diene et al., 2016). Sequence alignments also revealed the variant domain of WP_130892444 to be remarkably similar to that of the Amidase_2 domain of LysGH15, the endolysin encoded by a myovirus phage (GH15) that infects Staphylococcus aureus (Gu et al., 2011). This lysin possesses a modular structure containing an N-terminal CHAP domain, a central Amidase_2 domain, and a C-terminal SH3_5 (PF08460) bacterial-binding domain. Elucidation of the crystal structure of the Amidase_2 domain of LysGH15 (Gu et al., 2014) revealed the aa residues involved in Zn²⁺ binding (H214, H324, and C332), catalysis (E282, and T330), as well as other important residues (W263, and N75), to be conserved in WP_130892444 at comparable positions ( Figure S1 ). In sharp contrast, some of the equivalent residues in the NAM-amidase of S. pneumoniae TIGR4 (Mellroth et al., 2014; Li et al., 2015)—namely H26 and D149, as zinc ligands, and the catalytic residue H147—differed from those of WP_130892444 (and LysGH15) but were fully conserved among the widespread 318 aa-long NAM-amidases ( Figure S1 ).

Only four att core sequences for PPHs were reported in studies published up until 2009 (Gindreau et al., 2000; Obregón et al., 2003b; Romero et al., 2009a), and additional core sequences have only been seldom reported in more recent papers (Càmara et al., 2018; Garriss and Henriques-Normark, 2020). To the best of our knowledge, only seven different attB core sequences have been previously described ( Table S5 ). Remarkably, genome examination of many PPHs currently deposited in public databases (131 out of 158; last accessed, January 30, 2021) allow for no precise identification of the attP core site because most of the reported phage genomes only include the DNA region running from the first nucleotide of the int gene to the last one of the endolysin-coding gene (even though the definition was that of ‘complete genome’). Further, they lack the corresponding intermediate sequence where attP ought to be located ( Table S4 ).

To gain insight into the sequences and locations of the integration sites of the PPHs described in the present study, a detailed analysis of all the PPHs harbored by 126 strains with near complete (chromosome assembly level) genomes was performed. The strains included in this subset represent 88 different STs, 27 PMEN clones, 66 GPSCs, and 27 different serotypes. In addition, the dataset included one non-encapsulated laboratory mutant and four NT isolates ( Table S6 ). As mentioned above, 100 out of 126 strains (79.4%) were lysogenic and fulfilled the above assembly requirements in the original dataset. Fifty one isolates harbored only one PPH per genome whereas 49 were polylysogenic, and one of them (strain GPSC72) contained up to five different prophages ( Table S6 ). Taking advantage of previously determined locations of int genes, the genomes of these strains were examined in detail and the sequences corresponding to the attL and attR sites of 176 prophages belonging to 20 different PPH groups recorded. The sequences and locations for the remaining six groups (PPH035, PPH065, PPH090, PPH095, PPH105, and PPH110) were estimated using the partial genomes (either at the contig or scaffold assembly level) of nine additional strains included in the dataset. Prophage genomes were defined as the DNA sequence running from the first nucleotide of attL to the last one preceding attR. The attL sequences (obviously identical or near identical to their corresponding attR within each PPH group) were used as queries to map the corresponding attB sites on the S. pneumoniae D39 genome ( Table 4 ). With the exception of PPH130, the core sequence of 24 different attB could be determined. As mentioned above, seven of them had been already reported, namely, PPH005, PPH010, PPH015, PPH035, PPH080, PPH085, and PPH100 (compare the data of Table 4 with those of Table S5 ). With the exception of PPH075, PPH125 and PPH130, in which attB sites appear to be located inside insertion sequences (ISs), the precise chromosomal location of the different attB sites could be established. Although many PPH insertion sites are located in intergenic spaces (something typical among prophages), six core attachment sites mapped within the 3’ region of genes annotated for the D39 genome, and included the termination codon (TAA) of the gene where the prophage was integrated ( Table 4 ). In addition, attB _PPH035 was found to partially overlap (but apparently not interrupt) the 3’ part of ccnB encoding csRNA2, one of the five small non-coding csRNAs (cia-dependent small RNAs) that form part of the two-component regulatory system CiaRH (Sinha et al., 2019). It should be underlined that ccnB is not annotated for the S. pneumoniae D39 genome. On the contrary, seven prophage groups potentially interrupt gene translation: 1) the core integration site of PPH010 viruses (attB _PPH010), previously reported to lie between SPD_RS00115 and SPD_RS00120 (Romero et al., 2009a) but actually located in ccnC coding for csRNA3 (Furi et al., 2019); 2) PPH015, which integrates into SPD_RS00125 encoding the signal recognition particle sRNA involved in membrane protein targeting (Steinberg et al., 2018); 3) PPH040, which integrates into SPD_RS01460, potentially encoding an intramembrane metalloprotease of the CAAX proteases and bacteriocin-processing enzymes (CPBP) family (Pei et al., 2011); 4) PPH050, inserted into SPD_RS01935 encoding a frameshifted choline-binding protein (CbpG) in D39 (Frolet et al., 2010); 5) a gene (SPD_RS09795) encoding UlaR [a transcriptional activator of the ula operon in the presence of ascorbic acid (Afzal et al., 2015)], which may be inactivated by the insertion of bacteriophages of the PPH095 group; 6) PPH100, which integrates into SPD_RS09885 encoding ComGC, the major subunit of the competence pilus (Laurenceau et al., 2015); 7) and bacteriophages of the PPH120 group, which disrupt SPD_RS10405, i.e., the gene encoding for the PcpA choline-binding protein (Sánchez-Beato et al., 1998) (now an important component of several new protein-based, pneumococcal vaccines currently under evaluation) (Masomian et al., 2020).

Table S6 attempts to organize the different PPHs in a manner that takes into account their wide genetic diversity. For this, a concept of ‘equivalent PPH genomes’ was followed, i.e., genomes that overlapped for ≥90% of their total length and showed ≥90% sequence identity were considered ‘equivalent’, and only one of them was used in further comparisons. Figure S2 shows a comparison of the different PPH genomes. PPH095 and PPH110 are not depicted since, although they inserted into two different attBs, these PPHs were equivalent to PPH055 (see below). Due to the evident sequence variability of prophage genomes, pairwise nucleotide alignments were initially performed only among PPHs of the same group. Several main conclusions were then drawn: 1) int genes were present in every prophage found, but this was not the case for endolysin-coding genes; 2) two major PPH groups were recognized, one with genome lengths ranging from ≈10–20 kb on one side (e.g., PPH005 and PPH030 groups), and the other from ≈35–45 kb on the other side (e.g., PPH010 and PPH080 groups); 3) major differences can still be seen among the members of a PPH group (e.g., PPH015, PPH020, PPH045 and PPH090 show large differences); 5) several PPHs with genomes of 51–92 kb appear to result from recombination events between two different prophages, e.g., i) PPH075 is the result of the insertion of an unknown defective prophage (12,227 bp) into a PPH085_5 equivalent. This defective prophage is flanked by a 33 bp-long sequence (5’-CCCTAGACTTGAAATAAAGCGCATTTCTCTATA-3’) located at positions 32,873–32,905 and 45,100–45,132 of the PPH075 genome; a near-identical sequence (1,911,317-TCCTAGACTTGAAATAAAGCGCATTTCTCTATA-1,911,349) may contain part of the promoter region of SPD_RS10265 (malA) in the D39 genome (Nieto et al., 1997); ii) PPH080_11 appears to result from the integration of a PPH010-like prophage into a PPH080_3 equivalent; as expected, the insertion of the PPH010-like phage occurs in the sequence 5’-CTTTTTCATAATAATCTCCCT-3’ at positions 2069–2089 and 32,251–32,271 of the PPH080_11 genome; iii) the insertion and rearrangement of a PPH010_4-equivalent into a previously unknown prophage gives rise to PPH120, possibly with the assistance of two ISs (ISSpn5 and IS1167) and the addition of a gene cluster encoding several tRNAs. Sequence comparison revealed this prophage to be very similar (82–85% query coverage and >95% nucleotide sequence identity) to two Streptococcus anginosus prophages, namely Javan83 (Rezaei Javan et al., 2019) and SA01 (van der Kamp et al., 2020); interestingly, the latter two prophages appear to harbor no tRNA genes. In addition, both encode an endolysin 76% identical (88% similar) to Cpl-7 (see above), whereas, as already discussed, the new component of PPH120 codes for a peculiar 328 aa-long NAM-amidase of the Amidase_2 family of proteins; iv) a tandem insertion of two near identical PPHs (PPH080_1A and PPH080_1B) gives rise to the 92 kb-long hybrid prophage PPH080_1AB.

Among the various prophage identification tools, we used the PHASTER program to analyze the 109 genomes included in Table S6 (186 PPH). Up to 263 putative PPHs were predicted; 40 of them were designated as ‘intact’ by the program and actually corresponded to real prophages ( Table S6 ). The predicted limits of the PPHs did match the real coordinates in only few cases (e.g., PPH030 from strain SP64 was almost correctly located (real: 177,043–189,925; predicted: 177,043–189,926). Unfortunately, in most instances, this was not the case. For example, in strain GPSC47 (NZ_LR216060), PPH010_9 is located between coordinates 3152 and 36,649 whereas PHASTER predicted the range 1771–42,683). Of note, the existence of PPHs of the groups 010, 015, 080, and 110 was well predicted, whereas that of those belonging to groups 020, 045 and 115 were not ( Table S7 ). In our hands, the sensitivity of PHASTER (0.74) was similar to that previously reported (Lopes de Sousa et al., 2018; Reis-Cunha et al., 2019).

Pairwise nucleotide alignments were then performed among genomes belonging to different PPH groups. In addition to equivalent PPHs (see above), a second level of similarity was allowed, i.e., those prophages with an overlap of between 80 and 89% of their genomes and showing nucleotide identities of ≥90%. Figure 1A shows ‘equivalent’ and ‘very similar’ genomes on white or gray backgrounds respectively. It should be underlined that equivalent prophages were distributed between different PPH groups. Thus, in addition to the equivalency between the defective prophages PPH055, PPH095, PPH110 and PPH115_2 partly mentioned above, other equivalent, putatively complete prophages of different groups were seen: 1) PPH010_2, PPH035 and PPH130; 2) PPH015_1, PPH080_8 and PPH090_3; and 3) PPH040 and PPH080_3. As expected, most of the ‘very similar’ category of prophages corresponded to different members of the same group, as exemplified by members of the PPH010 group such as PPH010_7–PPH010_10 and PPH010_12–PPH010_14 ( Figure 1A ). Pairwise comparison of 80 PPH sequences and a dendrogram depicted seven major prophage clusters, and three singletons ( Figure 1B ). Four of the clusters matched those described by Brueggemann et al. (2017) and one more to the defective prophages mentioned by Rezaei Javan et al. (2019) that were recognized as separate entities from full-length PPHs. The remaining PPHs (either clustered or not) correspond to previously undescribed prophages.

Many prophages from both Gram-negative and Gram-positive bacteria integrate into tRNA genes (Williams, 2002) although this is not the case for PPHs (see above). Nevertheless, sequence analysis revealed the presence of potential tRNA genes in some PPH genomes ( Table S8 ). Of note, PPH120 has nine tRNA genes located at the right end of the prophage, whereas other PPHs have these genes located closer to the middle. A search of PPH genomes reported elsewhere revealed that five previously sequenced phages (namely, IP8, IP16, IP25, IP26, and IP27) also harbor a tRNA-Ser gene, a feature not mentioned in previous reports ( Table S8 ).

PblB may be an important virulence determinant in some PPHs. PblB has been described as a phage minor tail protein (a putative anti-receptor) that behaves as an adhesin mediating the galactose-specific adhesion activity of pneumococci to platelets and human lung epithelial cells in vitro. It is also required for nasopharyngeal and lung colonization in a mouse model of infection (Hsieh et al., 2015). The sequence diversity of PblB is notable (Brueggemann et al., 2017). Amino acid sequence alignments in the PPH dataset fully confirmed this (as shown diagramatically in Figure S3 ), even among PPHs of the same group (e.g., in PPH010_7, in which the length of PblB was 2134 aa in strain GPSC97 but 3256 aa in GPSC25). It is also reported that the PblB encoded by the temperate phage SM1 infecting S. mitis (1062 aa) (Bensing et al., 2001) is very different to the protein (P15; 1987 aa) previously analyzed in the S. pneumoniae NTUH-P15 strain (Hsieh et al., 2015). Moreover, these two proteins differ notably from most of the PblB proteins found in the present study, particularly in their middle regions where different combinations of five different (albeit related) sequence repeats were found ( Figure S3 ). PblB-like proteins were encoded by some of the phages belonging to the PPH010, PPH035, PPH040, PPH080, PPH085, and PPH100 groups. Willner et al. showed that the S. mitis prophage SM1 can be induced by commonly ingested substances and that the oral cavity is indeed a reservoir of pblA and pblB genes (Willner et al., 2011).

First described in the Gram-negative bacterium Dichelobacter nodosus (an ovine footrot pathogen), virus-associated protein E (VapE _Dno ) has been proposed involved in virulence (Billington et al., 1996). Proteins similar to VapE have also been reported encoded by prophages of other microorganisms such as S. aureus (Lindsay et al., 1998), Pseudomonas putida (Canchaya et al., 2003), Vibrio parahaemolyticus (Seguritan et al., 2003), Staphylococcus epidermidis (Chen et al., 2013) and Elizabethkingia anophelis (Peng et al., 2020), although direct evidence of the involvement of VapE in virulence was provided in neither of these studies. In S. pneumoniae, Romero et al. (2009a) first noted that several prophages encoded proteins similar (about 30% identity; E value = 10^–34) to VapE _Dno (also designated as virE). More recently, several gene products from defective prophages have been shown similar to a Streptococcus suis protein named VapE _Ssu (Rezaei Javan et al., 2019), a protein apparently unrelated in primary structure to VapE _Dno but which has been shown involved in virulence in a mouse model (Ji et al., 2016). The translated nucleotide dataset of PPHs was compared to both VapE _Dno and VapE _Ssu and using an E value cutoff of 10^–20, two sets of putative virulence-associated proteins were found ( Table S9 ).

It is well known that bacteria have developed defense mechanisms to help them survive phage attacks. These mechanisms may interfere with the adsorption of the phage onto the cell surface, with phage genome injection, or target the foreign nucleic acid after its injection, etc. (for recent reviews see Labrie et al., 2010; Koonin et al., 2017; Rostøl and Marraffini, 2019). Phages, in turn, have evolved mechanisms to overcome restriction-modification (R-M) systems. One of these is based on the presence in phage genomes (both virulent and temperate) of orphan methyltransferase (Mtase) genes (i.e., without concomitant endonuclease-coding genes) (Murphy et al., 2013). There are three functional classes of type II DNA Mtases that function by transferring a methyl group from S-adenosyl-L-methionine to a target nucleotide base, forming either N-6-methyladenine (class I), N-4-methylcytosine (class II), or C-5-methylcytosine (class III). Our laboratory previously reported the existence of putative Mtases-encoding genes in two S. pneumoniae temperate phages, namely MM1 (Obregón et al., 2003a) and VO1 (Obregón et al., 2003b). Taking the Mtases of these two phages as query sequences, the complete set of PPH genomes was screened. Interestingly, more than 30 putative orphan Mtases of either the MM1 or VO1 class were detected. None of them was encoded by defective PPHs ( Table S10 ).

The proposed core sequence of attB _PPH090 (46 bp-long) mapped to the 3’ end of the lytA_Spn gene overlapping its termination codon ( Table 4 ). Consequently, phage integration should not disturb lytA translation. In sharp contrast with other PPHs, the most striking feature of prophages of the PPH090 group is that there is no preference for a particular Int (or group of Ints) ( Table S2 ). Specifically, different PPHs of the PPH090 group encode Ints of the tyrosine (12 entries) or serine (6 entries) families, some of which are identical to those of other prophage groups: 1) tyrosine Ints [PPH010 (WP_000876726, WP_000876733, WP_000876735, and WP_050199954), PPH015 (WP_023396066, WP_044813689, WP_050127433, W_050203801, and WP_050221901), PPH080 (WP_000266839, WP_000266847, and WP_000266854)]; and 2) serine Ints [PPH100 (WP_024478469, WP_044812715, WP_050203932, WP_050226267, WP_050230313, and WP_050974546)]. These data strongly suggest that PPH integration into lytA_Spn does not require a specific interaction between attB, attP, and the Int, and that some other factors may be involved—possibly recombination events between lytA_Spn and lytA _PPH (see below).

The four PPH090 genomes shown in Figure S2 encode, respectively, one of three different Ints of the tyrosine family (PPH090_1–PPH090_3) and one Int of the serine family (PPH090_4). Among them, the genome of the NT strain SMRU257 (master record NZ_CKNY01000000) appears to be unique since, in addition to the PPH090_2 genome (39,182 bp; contig NZ_CKNY01000002), it contains a 39,182 bp-long sequence (contig NZ_CKNY01000010) inserted into the ffs gene (corresponding to SPD_RS00125 in S. pneumoniae D39) ( Table 4 ). Moreover, this prophage harbors all the characteristics of a full member of the PPH015 group, including the near identical flanking attL and attR core sequences (5’-TTGTGTGCTCTTTTTTCGTGC-3’). Remarkably, PPH090_2 and the ‘new’ PPH015 only differ at two nucleotide positions, i.e., 918 (G) and 951 (T) of the 957 bp-long lytA _PPH and lytA _PPH* genes (the asterisk indicates that the phage and bacterial lytA genes were probably chimeric [a hybrid] because they contain sequences of both phage and bacterial origin) ( Figure S4 ).

The hypothesis that lytA _PPH recombines with the lytA_Spn gene and integrates into the pneumococcal genome was tested by making a detailed analysis of both the lytA genes and the DNA regions located between int and lytA_Spn *, as well as downstream of lytA _PPH*, in the integrated state ( Figure 2A ). Figure 2B shows the appropriateness of the hypothesis since, immediately after the termination codon of lytA _PPH*, the sequence was syntenic, i.e. shares the same genetic order, with the DNA region located 3’ of lytA_Spn in a non-lysogenic strain. In fact, previous studies had already provided different kinds of evidence on the existence of recombination between these genes (Romero et al., 1990; Whatmore and Dowson, 1999; Morales et al., 2010). It has also been reported that lytA_Spn and plyA (the gene encoding the main pneumolysin, a cholesterol-dependent cytolysin) form part of a pathogenicity island in S. pneumoniae, characterized by the presence of direct repeats (ISs), phage-related genes, and/or genes potentially encoding virulence factors and mobility proteins (Ints, transposases). These findings also suggest that the plyA–lytA island is a recombination hotspot. At least eight genomic arrangements of this island have been recognized (Morales et al., 2015). During the present study, an additional arrangement that appears to be exclusive to some NT pneumococci was found ( Figure 2B ). It is worth mentioning that a larger protein (3628 aa residues; WP_142367134) of unknown function is encoded within the plyA–lytA island in the pneumococcal strain SMRU257.

The DNA region located immediately 3’ of the corresponding lytA _PPH* gene of each of the four PPH090 was aligned with a 2 kb-long region located immediately downstream of the lytA_Spn gene of the non-lysogenic S. pneumoniae D39 strain (NC_008533.2; complement [1,727,602–1,729,601; the position of the last nucleotide of lytA_Spn is 1,729,602]). In agreement with the hypothesis formulated above, the four PPH090 genomes showed >96.5% nucleotide identity with the D39 genome in that region (data not shown). The PPH090 regions that run from the 3’ end of int to the 3’ of lytA_Spn* after integration, were also aligned ( Figure 3 ). This region was 548 bp-long in PPH090_4 (serine Int family), and shorter (305–306 bp) in prophages with tyrosine Int-coding genes. The DNA regions corresponding to the two putative PPH015 prophages (PPH090_2 and PPH090_3) were very similar (97.7%; 298 identical nucleotides in a 305 nt overlap). The presence of a 51 bp-long sequence (5’-ATAGAAAGGAAACTTTCTAAATTGTTCTTTCACCGCAGGCTTAGGCTTGCG-3’) located immediately after the termination codon of the lytA_Spn * gene and found to be conserved in all the PPHs studied (indicated by orange rectangles in Figure 3 ) should be noted. Among the 126 pneumococcal strains with either complete or near complete genomes, this 51 bp-long signature was absent from the 26 non-lysogenic strains (as expected), but it was found immediately downstream of the 3’ end of the lytA_Spn gene among two lysogenic strains: 4041STDY6836166 (Acc. No. NZ_LS483451; positions 1,865,445–1,865,394) and GPSC72 (Acc. No. NZ_LR216049; positions 1,745,355–1,745,304). In these strains, the 51 bp-long sequence is part of a ≈700 bp-long fragment that is ≥95% identical to another fragment located 3’ of several lytA _PPH genes, particularly those within the PPH010 group (unpublished observations). This finding indicates that those strains were once lysogenic, but that at some point, most of the prophage was lost. Interestingly, a putative transcriptional terminator is located close to the 3’ end of the lytA _PPH (and lytA_Spn ) genes and may function in a manner similar to that of the transcriptional terminator of the lytA_Spn gene previously determined (Díaz and García, 1990) ( Figure S5 ).

PPH090-like prophages appear to be made via recombination of the lytA _PPH and lytA_Spn genes after the prophage is spontaneously excised from the host genome. However, recombination may also take place during the lysogenic (integrated) state of the prophage. In this instance, however, lytA recombination may result not in the emergence of a new PPH090 prophage but in chromosomal rearrangements. To gain insight into the recombination events that appear to occur between lytA_Spn and lytA _PPH, the corresponding genes from the four PPH090 reported here were aligned and the position of nucleotides conserved in lytA_Spn* and lytA _PPH* —but not in both groups— were annotated ( Figure 4A ). A search for polymorphisms at those positions was then performed in all the lytA_Spn (325) and lytA _PPH (356) alleles of the dataset ( Figure 4B ). Few polymorphic positions appeared between positions 1 and 100 (approximately), but beyond there the average polymorphism reached 7 and 14% in the tested positions of lytA_Spn * and lytA _PPH* respectively. This suggests that recombination events (if any) are more frequent in the 3’ than the 5’ moiety of the gene. The possible significance of the notable polymorphism observed at three nucleotide positions (14 and 17 in lytA _PPH, and 282 in lytA_Spn ) was not further analyzed, but the importance of the polymorphism at position 951 (very close to the 3’ end of lytA which is located at position 957) mentioned above was studied in detail. As shown in Figure 4B , searching among the 325 lytA_Spn alleles in the resulting data showed 84 of them (154 strains) to have a ‘T’ instead of an ‘A’ nucleotide at position 951. Table S11 summarizes a systematic analysis of this region among a subset of 51 of those 84 lytA_Spn alleles (111 strains). The results indicate the existence of recombination events that produced the translocation of several DNA fragments from their original position (taking the non-lysogenic S. pneumoniae D39 genome as a reference) to a region located immediately downstream of the lytA_Spn * gene ( Figure 5 ). Five major rearrangement events were recorded and, in agreement with the genes found, the direct participation of PPH010 (SPD_RS00120), PPH015 (SPD_RS00130), PPH080 (SPD_RS07410), PPH085 (SPD_RS09110), or PPH100 (SPD_RS09880) was deduced. By comparing these arrangements with those of the different PPHs described in this study, the bacterial and phage lytA genes were seen to represent inverted repeats in arrangements A and B, and individual direct repeats in arrangements C, D, and E. A recombination event in cases A or B would likely produce a chromosomal inversion ( Table S11 ). In addition, since the region between the lytA genes (the spacer) is asymmetrical around the origin of replication, a chromosomal inversion may induce changes in the relative positions of the origin and terminus of replication, as previously noted (Achaz et al., 2003). In arrangements C, D and E, a recombination event produced between repeat sequences in the same orientation on the same chromosome could result in a translocation (Hughes, 2000). A close examination of the region located immediately downstream of lytA_Spn * (SPD_RS09250; depicted as gray squares in Figure 5 ) in 14 different strains revealed complex organizations of different length that contained prophage fragments or even putative full length prophages. For example, strain SMRU51 harbored a prophage equivalent to PPH035 ( Figure 6 ). Of note, an additional recombination event in strain SMRU1319 (arrangement D2) produced an inversion of the genes located upstream of SPD_RS09115 and downstream of SPD_RS09110 ( Table S12 ).