ABCs IPD cases are defined through isolation of Streptococcus pneumoniae from a normally sterile site in a surveillance area resident. The surveillance sites included entire states or selected counties in 10 states, representing approximately 32.2–35 million individuals during 2016–2022 [1]. Case medical charts were reviewed to attain demographic information, including residence type and clinical information. AEH were defined as adults documented as homeless or residing in a shelter, mission, medical respite, or church community center at the time of positive culture. ABCs data were determined to be nonresearch, public health surveillance, such that review by an institutional review board was not required.

Isolates were assigned serotypes and multilocus sequence types (STs) employing whole-genome sequencing [7, 9]. All ABCs isolates received since January 2015 are subjected to short-read sequencing with genome sequences deposited in the National Center for Biotechnology Information Sequence Read Archive under BioProject number PRJNA284954. A range of 76.6%–95.5% of case isolates were assigned serotypes in the 10 sites during 2016–2022, with the average being 87.2%. Of 19 457 isolates from ABCs IPD cases in 2016–2022, 17 017 (87.5%) were sent to the Centers for Disease Control and Prevention (CDC), of which 16 964 (99.7%) were assigned serotypes. For the incidence data, correction factors were used to estimate serotype 4 rates within each of the 4 western sites as follows: (reported serotype 4 cases) × (all IPD cases ÷ all IPD cases assigned serotypes). Selected deduced serotype 4 isolates were subjected to Quellung testing and immunodiffusion tests against CDC type 4 typing antisera [10].

The PacBio Microbial Multiplexing procedure was implemented (Pacific Biosciences). Genomic DNA extracted for short-read sequencing was also employed for 3 representative genomes (isolate identifiers 20234287 [4/ST10172], 20166336 [12F/ST220], and 20234295 [4/ST18704]). Libraries were generated with the SMRTbell Express Template Prep Kit 3.0. Libraries were size-selected using a 0.45 × Ampure bead cleanup (Pacific Biosciences) to remove small DNA (<5 kb). Final libraries were sequenced for 30 hours after 30 minutes preextension times on a Sequel II (Pacific Biosciences) using Sequel II Binding Kit 3.2. PacBio HiFi reads were utilized in Flye version 2.9 [11, 12] to conduct de novo assembly for the genome size of 2.13 Mb. Assembled circularized contigs were evaluated using BLAST+ version 2.9.0 [13]. Read coverages of 327–681 were obtained for the 3 genomes that were assigned accession numbers CP152418 (20234287), CP154876 (20155336), and CP154877 (20234295).

Prokka version 1.14.5 was used to obtain annotated open reading frames (ORFs) [14]. Recombinant regions of 3 representative strains were aligned to generate figures using Easy Fig version 2.2.3 [15]. Core genomic maximum parsimony genomic trees and pairwise SNP matrices were generated from short-read bacterial genome sequences employing kSNP3.0 with a kmer size of 19 [16]. Resultant core.tre files generated from kSNP3.0 were used to generate core genomic phylogenetic diagrams employing Mega7 [17]. Progressive Mauve [18] was employed to generate consensus 1 944 951 bp core genomes of two 4/ST18704 isolates plus the two most related 12F/ST220 isolates. The core alignment was subjected to Gubbins [19] to identify recombinant regions. Circular maps were generated employing Proksee [20]. BLAST comparisons of ORFs within recombinant regions corresponding to the progeny, donor, and recipient strains were generated employing Proksee.

Fisher exact tests were used for the different comparisons in this study evaluating underlying risk factors (experiencing homelessness, time periods, different serotypes). Differences were considered statistically significant at P < .05 (2-sided P values).

Denominators for IPD incidence calculations for adults ≥18 years of age within each of the 4 western ABCs sites were obtained from US Census Bureau estimates within ABCs counties [21]. For estimates of IPD among AEH, we used US Department of Housing and Urban Development (HUD) point in time data [22] for the 4 corresponding geographic areas only for years 2019 and 2022 because point in time data methodology was modified during 2020–2021.

Overall IPD incidence calculated for ABCs areas during 2016–2019 was stable at between 9.2 and 9.6 cases per 100 000 individuals each year and, presumably due to nonpharmaceutical interventions, plummeted to 5.3–5.4 cases per 100 000 during 2020–2021, before increasing to 8.3 cases per 100 000 during 2022 [1]. There was a trend of disproportionate increases of IPD in AEH relative to ANH in prepandemic (2018–2019 compared to 2016–2017), pandemic (2020–2021 compared to 2018–2019), and postpandemic (2022 compared to 2018–2019) years (Table 1), as shown by the proportions of combined reported cases due to the 6 serotypes causing the most IPD in AEH (the first 6 depicted in Figure 1), combined reported cases due to the remaining serotypes, and combined reported cases due to all serotypes. Twenty-six of the 27 percentage comparisons (5 of the 6 individual serotypes, 6 serotypes combined, remaining serotypes, all serotypes) showed increases of the later time period compared to the earlier time period. All increased rations for the combined cases due to the 6 serotypes causing the most IPD among AEH, as well as for all serotypes combined, were highly significant (P < .001).

Except for 2 isolates, during 2016–2022, serotype 4 IPD was only found in adults 19 years and older (634/636 isolates) and serotype 4 IPD affecting persons experiencing homelessness was completely restricted to adults. Of 1440 total cases recorded from persons experiencing homelessness during 2016–2022 (including 200 IPD cases for which isolates were not available), only 5 were from individuals younger than 18 years (1 lacking isolate, 2 serotype 22F, 1 serotype 12F, and 1 serotype 10A).

During 2016–2022, serotype 4 was the most frequent cause of IPD in AEH compared to any other serotype (Figure 1 and Table 1). During 2021–2022, the emergent serotype 4/ST10172 was the most common serotype 4 lineage (Figure 2). Serotype 12F ranked second as a cause of IPD among AEH (Figure 1), primarily due to the closely related ST220 and ST218 lineages, as seen during 2018–2019 [2, 3]. Also, the highest percentages of IPD cases among AEH relative to all adults of known residence status were caused by serotypes 12F and 4 during the entire period of 2016–2022 (237/631 [37.6%] and 155/614 [25.2%] for serotypes 4 and 12F, respectively, compared to 851/14771 [5.8%] for all remaining serotypes; P < .001 for the individual serotype 4 and serotype 12F proportions compared with the proportion for all remaining serotypes combined). Although IPD incidence in adults was lower during 2020–2022 compared to 2019 [1], reported serotype 4 IPD cases increased among all adults (both AEH and ANH) during 2020–2022, and more than doubled during 2022 relative to 2019 (Figure 1 and Table 1). Of all adult cases assigned serotypes in ABCs, only serotype 4 exhibited overall increases during each of the years 2020–2022 relative to 2019 (Table 1 and Figure 1).

As shown in Table 2, serotype 4 IPD was largely confined to western ABCs sites. Using ABCs, US census, and HUD data to calculate adult incidence for all IPD and serotype 4 IPD within the 4 western states, increases in IPD incidence due to all serotypes among AEH in 2022 relative to 2019 were observed in New Mexico and Oregon, with a statistically significant increase seen in New Mexico (P = .014) and New Mexico (P = .002) (Table 2). Within New Mexico and Oregon serotype 4 IPD significantly increased among both AEH and ANH. Remarkably, there was no reported serotype 4 IPD in Oregon during 2019, but there were 42 reported cases during 2022.

Among isolates recovered through ABCs between October 2022 and August 2023 and received by the CDC Streptococcus Laboratory for analysis between February 2023 and June 2024 there were 30 serotype 4 blood isolates recovered from adults aged 27–71 years, 16 of whom were AEH (22 men and 7 women; unknown residence status for 1 New Mexico case subsequently determined to reside outside of ABCs areas) that shared the new multilocus sequence type ST18704, a single-locus variant of ST220 independently recorded on 21 March 2023 from a serotype 4 isolate recovered from an adult with IPD in the Southwestern United States [6]. These isolates and the CDC serotype 4 reference strain were strongly serotype 4 and immunodiffusion tests indicated antigenically identical capsules (data not shown). The ABCs serotype 4/ST18704 cases occurred in the Oregon (25 isolates), Tennessee (2 isolates), Minnesota (1 isolate), and California (1 isolate) ABCs sites. Twenty-three of these bacteremic cases were diagnosed as pneumonia and 2 as meningitis (1 fatal).

Phylogenetic analysis of available serotype 4 isolates recovered during 2021–2022 (Figure 2) and ST18704 isolates received by the CDC Streptococcus Laboratory between 2022 and June 2023 (Figure 3) revealed that the serotype 4/ST18704 isolates were highly related, with pairwise SNP distance of 0–21 and an average of 6 SNPs (Supplementary Table 2). Twenty-three of the 29 patients (79.3%; 13 AEH and 10 ANH), including 20 of the 25 (80%) cases in Oregon, were persons with substance use disorder involving use of 1 or more of (in order of incidence) methamphetamine, opioids, alcohol, and cocaine (Figure 3). In contrast, 33 of 187 (17.6%) 2022–2023 IPD cases in Oregon caused by nonserotype 4 strains were persons with substance abuse disorder involving, in order of incidence, methamphetamine, opioids or alcohol, and cocaine (P < .001). Only 2 of the 29 serotype 4/ST18704 ABCs case patients had reported history of pneumococcal vaccination and 21 of these individuals smoked tobacco. Although defined epidemiologic links were not identified, all 25 of the Oregon cases were listed from the same county (22 isolates) or the adjacent county (3 isolates). Both cases from Tennessee were from the same county and zip code. There was also extensive overlap between identified hospitals where patients were admitted (data not shown).

Pneumococcal serotype-switch progeny can have multiple genomically unlinked recombinant regions originating from the genetic donor of the cps operon [23]. Strain 4/ST18704 differs from 12F/ST220 strains from CDC bioinformatic pipeline data only in the cps serotype-specific markers (wzy4 vs wci12F) and the multilocus sequence type aroE target (aroE1 vs aroE10). In addition to serotype 4/ST10172 prevalence in western ABCs sites [2–4] (Figure 2), the aroE1 marker made this strain the top serotype 4 candidate strain tracked in ABCs as the genetic donor for the putative 12F/ST220 to 4/ST18704 serotype-switch event. Seven recombinant regions were mapped in the representative 4/ST18704 genome that ranged from 3096 to 70 341 bp in size (Figure 4). Six widely separated recombinant regions in the progeny shared sequence identity or near identity with the strain 4/ST10172 reference genome, including the cps4-containing R2 and the aroE1-containing R4 fragments. One predicted 4652-bp recombinant fragment (fragment R5) differed extensively from both recipient and donor lineage strains. The number of genes containing 1 or more missense mutations relative to the recipient and donor strains is provided for each fragment, as well as providing representative gene identifiers, in Figure 4. A total of 174 ORFs, not counting the 7 ORFs within fragment R5 and 5 nonconserved cps4 genes within fragment R2 (see GenBank cps4 accession CR931635), were included within the 6 recombinant fragments from the putative serotype 4/ST10172 donor strain, 124 of which contained 1 or more missense mutations compared to the recipient 12F/ST220 reference single-contig genome. In contrast, only 2 of the 174 ORFs contained single missense substitutions compared to the reference 4/ST10172 donor sequence. The regions replaced primarily revealed the same genes/orientations shared between the donor/progeny and recipient sequences (Supplementary Figures 1 and 2). There was a precise replacement of the 17 cps12F structural genes with the 15 genes of the cps4 operon (Figure 5), predicted to place the cps4 operon under transcriptional control of the cps12F promoter and associated recipient lineage transcriptional regulators [24] in the progeny. None of these transcriptional regulators, with the exception of malR, mapped within recombinant fragments; the predicted MalR products of the recipient, donor, and progeny (fragment R6 in Figure 4 and Supplementary Figure 2) share amino acid sequence identity (data not shown).

Serotype 4/ST10172 expanded in number during 2021–2022 and first appeared in Oregon during 2022 (Figure 2). This coincided with the appearance of 25 serotype 4/ST18704 cases in Oregon that were recovered during 2022–2023 (Figure 2 and Figure 3). This coincidence conceivably supports a causal effect related to the density of the donor strain 4/ST10172 within Oregon during this period. Phylogenetic analyses of representative serotype 4/ST18704 core genomes together with 76 available core genomic sequences of 12F/ST220 recovered during 2015–2022 (Supplementary Figure 3) did not support this notion, in that it revealed a large genomic distance between 4/ST18704 and its closest 12F/ST220 neighbor (difference of 761 SNPs). This SNP difference was reduced by subtracting SNPS shared between progeny and serotype 4/ST10172 donor strain within the 6 recombinant fragments that were not shared with the closest matching serotype 12/ST220 recipient lineage strains (data not shown). The SNP distance of 308 is still indicative of a large temporal difference between recipient and progeny according to calculations of the generation of about 4 core SNPs per year [25]. Nonetheless, differences of >500 SNPs evident between individual 12F/ST220 strains (Supplementary Figure 3) suggest that the most related recipient strains have potentially not been recovered through ABCs.