Commensal E. coli isolates used in this work were isolated from healthy college students, as described in Stephens et al. (2020). Bacteria were routinely cultured on LB broth or agar. Utilization of sugars as sole carbon and energy sources was tested by growth on M9 minimal salts medium, with the designated sugar at a starting concentration of 10 mM. Plates or liquid media containing D-glucose are referred to as “M9G,” while those containing D-sucrose are referred to as “M9S” herein. Cultures were incubated at 37°C (with shaking for liquid cultures). Fermentation was tested by growth in phenol red broth (Hardy Diagnostics) with the designated sugar present at a starting concentration of 0.5%. Cultures were incubated at 37°C with no shaking.

The genome sequences of commensal E. coli used herein have been described previously (Stephens et al., 2020). Isolates were assigned to MLST groups using the web-based MLST 2.0 algorithm (Larsen et al., 2012). BLAST/MegaBLAST searches (Altschul et al., 1990) of these genomes, as well as basic molecular biological processes such as PCR primer design, were done locally using the Geneious Prime desktop bioinformatics package (Dotmatics). Searches of the RefSeq database (Pruitt et al., 2007) were done with NCBI BLAST (Johnson et al., 2008). scr genes from the closest known related species were used as queries, to compensate for evolutionary drift between gene sequences from more distant relatives. To identify E. coli genes uniquely associated with sucrose utilization, a subset of commensal E. coli genomes of known sucrose phenotypes were submitted to the Genomes Online Database (GOLD; Mukherjee et al., 2023). The Phylogenetic Profiler tool within the Integrated Microbial Genomes and Microbiomes platform (IMG/G; Joint Genome Institute, Walnut Creek, USA) (Chen et al., 2023) was then used for analysis.

Genomic DNA was extracted using the NucleoSpin microbial DNA purification kit (Macherey-Nagel). For amplification of the scr genes by polymerase chain reaction (PCR), primer pairs for the target gene regions were designed using Geneious bioinformatics software (Biomatters LTD), and synthesized by Integrated DNA Technologies (Alameda, CA, USA). The entire 6.6 kb scr locus was amplified using primers scrK43F (TCC CGG CAT ATT CAC GTT TCC AC) and scrR6689R (CCG TTT TAC AGG GGC GAT GCA). The scrY gene alone was amplified using primers scrY1058F (ACC GCC TTA CCC CGA CAA CA) and scrA2818R (TTC AGT AAA AGC CTC ACA TCC GT). Genomic DNA from E. coli strain SCU-147 (Stephens et al., 2020) was used as a template for amplification of the scr genes for cloning. Products of appropriate sizes were verified by agarose gel electrophoresis, then prepared for cloning using the Gel and PCR Clean-up kit (Macherey-Nagel). The PCR Cloning Kit (New England Biolabs) was used to clone amplicons into plasmid pMiniT2.0, and ligated DNA was electroporated into E. coli NEB 10-beta C3019H, with selection on LB agar containing 100 μg/mL ampicillin. Four colonies were selected from each ligation and transformation and used for colony PCR. The same primers were used to assess putative recombinant plasmids. Colonies that produced strong product bands of the target size were grown overnight in liquid LB with 100 μl/mL ampicillin, and plasmid DNA was prepared using a commercial kit (Zymo). Plasmid DNA was digested with restriction enzyme EcoRI to verify expected insert size, and candidate clones were further verified by Sanger sequencing (Sequetech, Mountain View, CA, USA) to confirm the expected insert. Plasmid DNA was subsequently used to transform natural commensal E. coli isolates from Stephens et al. (2020), prepared by calcium treatment (Chang et al., 2017). To remove scrR from the scr locus, pMM007 was digested with BamHI, resulting in a 7.7 kb fragment product (pMM008). A partial internal 500 bp fragment of scrY was removed by digestion with EcoRV. The larger linearized plasmid was re-ligated with T4 DNA ligase using quick T4 DNA ligase (New England Biolabs), then transformed into NEB 10-beta C3019H.

To explore the genetic and physiological basis for sucrose utilization in E. coli, we employed a collection of diverse commensal E. coli strains with sequenced genomes (Stephens et al., 2020). The collection includes more than 100 representatives of all major E. coli phylogroups (A, B1, B2, C, D, E, and F) (Denamur et al., 2021). Table 1 shows results for a limited subset of 32 isolates, each representing a different MLST group, and each with fully assembled, closed genomes available in GenBank. Broth and agar plate-based assays were used to examine growth using sucrose as sole carbon and energy source. Representative growth curves and images of M9G and M9S plates are shown in Figure 2. The “Suc⁺” phenotype (Figures 2B,D) was defined here as being able to utilize D-sucrose as sole carbon and energy source in M9 minimal salts medium at a growth rate similar to glucose; under the conditions applied (small test tube cultures grown in a shaking incubator at 37°C), this corresponded to a growth rate μ of 0.8–1.1 generations/h. Roughly 20% of commensal E. coli strains in our collection were Suc⁺ (6/32 shown in Table 1). A slightly higher fraction (9/32, Table 1) were able to more slowly utilize D-sucrose for growth (Figure 2C and Table 1; isolates with μ = 0.4–0.7 gen h⁻¹ were designated “slow,” isolates with μ = 0.1–0.3 were designated “very slow” growth). The growth rate of colonies on M9S for these isolates was noticeably slower (Figure 2C). The remaining isolates (17/32 in Table 1) were unable to use D-sucrose as the sole carbon and energy source for significant growth (Suc⁻, Figure 2A).

To identify genes associated with the Suc⁺ phenotype, fully assembled and annotated genomes of four Suc⁺ isolates (SCU-115, phylogroup B2, MLST type ST1155; SCU-147, B2/ST1193; SCU-182, B2/ST131; SCU-387, B2/ST14) were compared to the genomes of eight phylogenetically diverse Suc⁻ isolates (SCU-103, A/ST10; SCU-152, B1/ST54; SCU-108 B2/ST95; SCU-112 B2/ST73; SCU-321, D/ST69; SCU-106, E/ST3695; SCU-120 F/ST648, and SCU-301 F/ST967) using the Phylogenetic Profiler tool of the IMG/G platform. Just five genes (scrA, scrB, scrK, scrR, and scrY) were present in the genomes of all four Suc⁺ isolates (minimum 80% identity) and absent in all of the genomes of the eight Suc⁻ isolates. The scr genes are clustered in two adjacent transcriptional units, scrK and scrYABR (Cowan et al., 1991). The 6.7 kb scr locus was highly conserved, with ≥99% identity among the four Suc⁺ isolates.

To test the sufficiency of the scr locus for sucrose utilization, the region was amplified by PCR and cloned into plasmid vector pMiniT2.0 (Figure 3). The recombinant scr⁺ plasmid construct pMM_027 was introduced into the E. coli NEB10-beta strain, a derivative of the standard DH10B cloning strain, which is in turn derived from E. coli MG1655 (Durfee et al., 2008), a phylogroup A/ST10 strain. Phylogroup A E. coli in our collection were universally unable to use sucrose, and lack the scr or csc loci. Neither the parental NEB10-beta strain or the recombinant NEB10-beta/pMM_027 strain were able to grown on M9 minimal medium with glucose or sucrose as sole carbon and energy source, presumably due to auxotrophies resulting from the complex genotype of NEB10b. They were, however, able to ferment glucose in phenol-red broth. NEB10-beta/pMM_027 gained the ability to rapidly ferment sucrose in phenol red broth. When pMM_027 was moved into SCU-113 (B1, ST5974, scr⁻ csc⁻), a prototrophic E. coli commensal isolate, the resulting strain was able to both ferment sucrose and utilize sucrose as sole carbon and energy source on M9 medium, albeit more slowly than a native Suc⁺ B2 strain. These results indicate that the scr locus is sufficient to convey the Suc⁺ phenotype.

Removal of the putative scrR transcriptional regulator from the recombinant construct (pMM_047) accelerated growth and fermentation of sucrose, as did deletion of the porin-encoding scrY gene (pMM_052). The scrY deletion removed the scrYABR operon promoter, but we speculate that it is possible that read-through transcription from scrK in the deletion construct provided sufficient expression of scrA and scrB to allow sucrose fermentation and growth.

The sequenced commensal E. coli genomes in our collection were queried via BLAST for the scr locus (Table 1). The Suc⁺ phenotype and scr are largely associated with a branch of the B2 clade that includes the ST1193 and ST131 MLST groups, known for their extraintestinal virulence (Pitout et al., 2022). Overall, scr was found in nearly half of B2 isolates.

The presence of the scr locus correlated well with the Suc⁺ phenotype, as the majority of scr-containing isolates were Suc⁺ (6/9 in Table 1, and 20/24 in our larger sequenced E. coli collection). Three unusual scr-containing isolates (SCU-175, 311, 312) from outside the B2 clade showed a Suc⁻ phenotype when incubated on M9S plates or liquid media for up to 1 day. However, isolated colonies did appear over time on these plates (SCU-175 is shown as an example in Figure 4). Genome sequence analysis revealed identical 1.9 kb Is3 insertions located 12 bp into the scrY gene (Figure 1). This insertion is expected to knock out transport by ScrY, the sucrose-specific porin of the Scr system (Hardesty et al., 1991; Schmid et al., 1991). The pMM_052 recombinant construct discussed above showed that scrY is not strictly necessary for the Suc⁺ phenotype, but the Is3 insertion in scrY may also have negative polar effects on expression of the downstream scrA, scrB, and scrR genes. GenBank searches identified four other examples of this precise configuration in E. coli F/ST62 isolates from around the world, but the Is3 insertion is not universal in F/ST62 isolates. In fact, the closely-related SCU-114 isolate in our collection has an intact scr locus and is Suc⁺ (data not shown). As shown in Figure 4, the isolated colonies selected from prolonged incubation of scrY::Is3 strains on M9 sucrose media stably gain the ability to utilize sucrose, as would be expected if the Is3 element has been excised.

Some scr⁺ isolates examined were not fully Suc⁺; for example, SCU-479 grew slowly, and SCU-176 grew very slowly on M9 sucrose. The basis for these phenotypes is not yet known, although we also noted that SCU-176 grew very slowly on glucose as well, so the growth phenotype may not be attributable specifically to a sugar utilization pathway.

Among 79 isolate genomes in our collection initially determined to lack scr (using the SCU-147 scr locus as a BLAST query), 54 isolates were Suc⁻, 26 were Suc^slow, and two were Suc⁺: SCU-391 (A, ST216) and SCU-490 (B1, ST6496). Analysis of the SCU-391 and SCU-490 genomes did not reveal hits above 80% identity to the SCU-147 scr locus. However, further inspection showed that these genomes contained a more distantly related scr locus (~67% identity with SCU-147 scr over 6.7 kb). The SCU-391 scr locus closely matches SCU-490 for the scrKYA genes, but they diverge at codon 229 of scrB, continuing into scrR (Figure 1). The scrKYA genes of SCU-391/490 were used as queries to search GenBank, identifying closely related loci (>95% identity, >95% query coverage) in more than 200 E. coli RefSeq genomes (0.8% of the total). Strikingly similar (>99.7% identity, 100% query coverage) scr loci were found in 230/494 (47%) C. freundii RefSeq genomes, and in the genomes of several other non-freundii Citrobacter, Enterobacter, and Klebsiella isolates in GenBank. The divergent SCU-391 scrBR region was found in genomes of distinct C. freundii isolates. It therefore seems likely that C. freundii or a closely related species was the ancestral source of the scr loci in SCU-391 and 490, via independent transfers to A and B1 E. coli. Phenotypic analysis indicated that these scr loci are fully functional in E. coli; indeed, growth curve analysis of SCU-391 found that it grew more rapidly on sucrose than glucose (μ_suc = 1.15 gen h⁻¹ vs. μ_gluc = 0.85 gen h⁻¹) and that it’s growth rate on sucrose is higher than B2 isolates with the native scr locus (μ_suc = 0.8–1.1 gen h⁻¹).

In E. coli SCU-147 and nearly all of the other scr⁺ E. coli B2 and F isolates in our collection with fully closed genomes, the scr locus is positioned on the chromosome adjacent to the queE gene. As noted by Díez-Villaseñor et al. (2010), in scr⁻ E. coli strains this is the location of the CRISPR2.2 array of CRISPR2 repeats, also known as iap repeats (Treviño-Quintanilla et al., 2007). The queE-CRISPR2.2 region is roughly 25 kb from the region encoding the CRISPR-CAS-E system in E. coli with the CAS-E genes. The 6.7 kb scr locus replaces approximately 0.5 kb of non-homologous DNA adjacent to queE. No IS or transposable elements flanking scr were noted that might mechanistically account for the insertion. In SCU-391 the C. freundii-like scr locus was located nearly on the opposite side of the chromosome, adjacent to a prophage genome resembling Shigella SFII. It appears likely that the C. freundii-like scr locus has been horizontally transferred among Gram-negative enterobacterial species. Indeed, GenBank searches suggest that many of these C. freundii-like scr loci in E. coli genomes reside on plasmids (data not shown), the chromosomal location in SCU-391 notwithstanding.

For a more historic view of the scr locus in E. coli, genomes of strains in the ECOR collection (Ochman and Selander, 1984) were examined. This collection was assembled to reflect the genetic diversity of E. coli as understood in the mid- to late 20th century. The scr locus was present in 8/72 of the ECOR collection genomes (11%), lower than the 21% in our contemporary commensal collection, which is significantly richer in B2 isolates. The scr locus is found in 4/16 (25%) of the ECOR B2 strains and 4/6 (67%) of the F strains (all intact scr loci). No other versions of the scr genes, such as the C. freundii-like scr, were detected in the ECOR genomes.

Shigella species have emerged evolutionarily multiple times from within the broader E. coli lineage (The et al., 2016), but clinical diagnostic databases such as API20E indicate that Shigella isolates rarely if ever ferment sucrose. The presence of the scr locus in Shigella genomes was queried in the RefSeq genome database. The scr genes were very rare, being present in less than 1% of sequenced S. sonnei and S. flexneri genomes, and in no S. dysenteriae or S. boydii genomes.

Beyond E. coli and Shigella, sucrose utilization is variable within the family Enterobacteriaceae. Data on sucrose fermentation from the API 20E database was used as a surrogate phenotypic assay for comparison with genome-based surveys of the scr locus in the RefSeq database. Results are shown in Table 2 for the scrA locus. The presence of scrA was correlated with sucrose utilization by species. For example, less than 1% of Salmonella isolates ferment sucrose, and in the Salmonella enterica RefSeq database of over 10,000 genomes, only eight contained the entire scr locus (minimum 80% coverage, 80% nucleotide identity). A rare exception to the exclusive role of the Scr pathway in sucrose utilization was seen in Citrobacter freundii. Isolates of this species near universally ferment sucrose, but only half of the C. freundii genomes in RefSeq contain the scr genes; however, as described below, the alternative csc locus is ubiquitous in C. freundii. Other Citrobacter species rarely or never utilize sucrose, and the scr and csc genes are concomitantly scarce. Conversely, Enterobacter and Klebsiella species almost always ferment sucrose, and nearly all of those genomes in the RefSeq collection contain the scr locus. Proteus vulgaris isolates almost all ferment sucrose and contain the scr locus, but P. mirabilis isolates rarely do either. Moving beyond the Enterobacteriaceae to related families in the order Enterobacterales, the scr locus is near universal in Erwinia amylovora, Serratia marcescens, and Yersinia enterocolitica, but Y. pestis and Y. pseudotuberculosis rarely ferment sucrose or contain scr genes.

Identifying the Scr pathway in a wider range of species offers the opportunity to examine evolutionary conservation of the proteins and regulatory components. On the protein side, the ScrA PTS transporter subunit (456 amino acids) is the most highly conserved across species, ranging from 91% amino acid identity between the E. coli and C. freundii proteins, to 81% between the E. coli and Yersinia enterocolitica proteins. The outer-membrane-spanning ScrY porin, the largest of the gene products at 505 amino acids, retains 64% identity between the E. coli and Y. enterocolitica proteins. The least conserved of the five scr gene products is the ScrB fructose-6-phosphate hydrolase enzyme (467 aa), with only 55% conservation between the E. coli and Y. enterocolitica proteins.

Comparison of non-coding DNA sequences with homologous functions can allow identification of conserved regulatory elements across species. Expression of scrYAB is highly inducible by sucrose in vivo, with regulation dependent on ScrR, a member of the LacI family of transcriptional regulators. Cowan et al. (1991) and Jahreis and Lengeler (1993) characterized the scrY promoter region of pUR400, identifying two 14 bp inverted repeats immediately upstream and downstream of the scrY −35 and −10 promoter elements, and providing mutational evidence that these serve as operators for ScrR binding. Jahreis and Lengeler (1993) further showed that, although sucrose induces expression from the scrY promoter in vivo, ScrR bound to D-fructose or fructose-1-phosphate (products of sucrose degradation) is released from operator DNA to allow RNA polymerase to initiate transcription. Figure 5 shows scrY promoter regions from eight species aligned to determine the extent to which the Scr regulatory sequences are conserved across species. The ScrR operator sites are indeed highly conserved sequences in the promoter region. The putative operators are positioned immediately upstream and downstream of the −35 and −10 regions of the promoter, and the downstream operator overlaps directly with the expected transcription start site.

Jahreis and Lengeler (1993) also identified a CRP operator positioned further upstream of the pUR400 scrY promoter and demonstrated catabolite repression of this promoter. As seen in Figure 5, the putative CRP operator is not well conserved across species. CRP-mediated catabolite repression of scr gene expression therefore may not be conserved in other Enterobacteriaceae, but this has not been tested experimentally.

Most of the literature on sucrose utilization in E. coli focuses on the csc locus. In our collection of sequenced commensal E. coli, seven Suc⁺ isolates contained both the scr and csc loci, and 15 Suc⁺ isolates contained only the scr locus, again demonstrating that it is the scr locus that is critical for rapid sucrose utilization. In the absence of the scr locus, the presence of the csc genes was most strongly associated with the Suc^slow phenotype, often with an extended lag phase (see Figure 2C for an example). No csc⁺ scr⁻ isolates were Suc⁺, but 23 out of 29 were scored as either “slow” or “very slow” for growth on M9 sucrose, including 2 “slow” and 5 “very slow” isolates shown in Table 1. These results are consistent with the literature describing csc-dependent sucrose utilization in E. coli to be “unusually slow” (Jahreis et al., 2002). Five csc⁺ scr⁻ isolates [SCU-171 (B2/ST657), 152 (B1/ST54), 483 (B1/ST10955), 106 (B1/ST3695), and 316 (E/ST57)] were completely unable to utilize sucrose for growth. SCU-106 has a 10 bp insertion in the cscB coding region, with an accompanying frameshift, but the other isolates have no obvious genetic defect in the csc locus to explain the growth deficiencies.

In our contemporary collection, the csc locus was found in 37% of isolates. The csc locus was universal in phylogroup D and E isolates, nearly universal in B1 isolates, uncommon in B2 isolates, and completely absent in A and F isolates. In the historical ECOR collection, the csc locus was present in 25/72 ECOR isolates (35%), similar in frequency to our contemporary collection. The csc genes were present in nearly all ECOR B1, D, and E strains (23/27), but rare in B2 strains (2/16), and absent in A and F strains (29 in aggregate).

csc resides at a recombination hotspot on the E. coli chromosome in close proximity to the argW tRNA gene (Jahreis et al., 2002). Moritz and Welch (2006) examined utilization of D-serine by diarrheagenic and uropathogenic E. coli and found that D-serine and (slow) sucrose utilization were mutually exclusive, consistent with earlier observations by Alaeddinoglu and Charles (1979). The cscRAKB and dsdCXA loci typically reside at roughly the same chromosomal location, though there is more than one possible configuration of each locus, perhaps because this is also a frequent target for integration of lambdoid phage. We observed diverse arrangements in this genomic region in our collection, with dsdC often retained even when the csc locus was present. In a few strains, both the csc and dsd loci were present in their entirety, but more often isolates had one functional locus or the other, as seen by Moritz and Welch (2006).

The csc locus was examined in Shigella sonnei, S. flexneri, S. dysenteriae, and S. boydii genomes, and was found to be defective in nearly all cases. Figure 4 compares the wild-type E. coli csc locus with loci commonly observed in the genomes of Shigella species. The cscA and cdcR genes were often present and intact, but the cscK (fructokinase) and cscB (sucrose permease) genes were usually interrupted with IS1, IS3, or IS4 elements, or missing altogether. An absence of scr loci, combined with disruption of csc loci, may explain why utilization/fermentation of sucrose is rare or absent in Shigella isolates. Shigella genomes rarely contain the dsd locus either; only 4% of nearly 2,200 Shigella genomes in the RefSeq database contained a full-length, uninterrupted dsd locus.

Querying GenBank with the entire E. coli csc locus (Ec_csc hereafter), only four examples were found in Escherichia genomes other than E. coli, four examples in coliphage genomes, four in Citrobacter farmeri genomes, and one each in a Citrobacter telavivensis genome and an Enterobacter hormaechei genome. These may have arisen from horizontal gene transfer from E. coli. Further inspection also revealed that Citrobacter freundii genomes contain a csc locus (Cf_csc hereafter) that falls below the detection threshold with the Ec_csc query in Blastn. Cf_csc is clearly homologous to Ec_csc when the amino acid sequences of the encoded proteins are compared. The arrangement of genes in Cf_csc is altered (Figure 4), with cscA and cscR each independently inverted. Nearly all (95%) of C. freundii RefSeq genomes contain a complete, uninterrupted (and likely functional) csc locus with this arrangement. Querying GenBank with Cf_csc identified several homologous loci, possibly also resulting from horizontal gene transfer, in Citrobacter portucalensis, Citrobacter youngae, and Kluyvera ascorbata genomes.

As with scr, identifying the csc locus in multiple species allowed comparison of the promoter regions to search for potentially conserved regulatory elements. In E. coli, expression of cscA is inducible by sucrose in vivo and dependent on CscR, which like ScrR is a LacI family member. Jahreis et al. (2002) identified two copies of a 12 bp inverted repeat in the intergenic region between cscA and cscK; one of these putative operators was the site of a mutation that upregulated expression of cscA. The promoter regions upstream of the cscA gene in five species were aligned and compared (Figure 6). The inverted repeats identified as potential CscR operators were indeed conserved between E. coli and the other species. Assuming a similar functional organization to the E. coli cscA promoter, these operators flank either side of the −35 and −10 regions of the promoter, with the downstream operator overlapping the transcriptional start site. Jahreis et al. (2002) also identified a putative Crp-cAMP operator, conventionally positioned adjacent to the −35 region of the cscA promoter and demonstrated catabolite repression of the Csc pathway. Extensive conservation of the putative Crp operator is seen between species (Figure 7).