However, despite the importance of zinc, it is toxic at large concentrations, because it competes with other metals for binding to active centers of enzymes. Thus all bacteria tightly regulate zinc transport. In E. coli and B. subtilis, as well as their closest relatives, transcription of the zinc transporter genes is regulated by ZUR repressors belonging to the FUR family (6) or, in the case of zosA of B. subtilis, by the PerR repressor from the same family (7). ZUR proteins from Gram-negative and Gram-positive bacteria are distant homologs, with the average identity of 25%, and we will relate to these regulators as nZUR and pZUR, respectively. In Streptococcus species, zinc transporter genes are regulated by AdcR, a transcription factor from the MarR family (8).

Several ZUR-binding sites of E. coli and B. subtilis were identified in experiments (5, 6). Binding sites of AdcR are unknown. We constructed recognition profiles using known nZUR sites of γ-proteobacteria and pZUR sites from Bacillus spp. and identified de novo signals for nZUR in α-proteobacteria and for AdcR in Streptococcus species. The obtained recognition rules were used to study zinc-dependent regulation in these four groups of genomes.

We described the evolution of the nZUR recognition signal in Gram-negative bacteria. We identified candidate ZUR-binding sites upstream of known and previously uncharacterized zinc transporters; in the latter case the analysis of regulation allowed us to assign zinc specificity to these transport systems. In Streptococcus species, we described a unique class of zinc-regulated proteins from the histidine-triad family that could be involved in adhesion and pathogenesis.

An unexpected finding was that zinc controls transcription of some paralogs of ribosomal protein genes, namely L36, L33, L31, and S14. In ref. 9 it was shown that of two or more copies of L36, L33, L31, and S14 proteins, one usually contains a predicted Zn-ribbon motif, whereas in other copies this motif is disrupted. An intriguing correlation between the presence of a Zn ribbon in a ribosomal protein and gene duplication was observed (9). However, the function of paralogous ribosomal proteins was unresolved.

Here we demonstrate the presence of candidate binding sites for zinc repressors (nZUR, pZUR, and AdcR) upstream of genes encoding paralogs of ribosomal proteins L36, L33, L31, and S14 with disrupted Zn ribbons. We suggest that these proteins act as alternative ribosomal proteins replacing the original proteins in zinc-depleted conditions. This decreases the total zinc requirement of ribosomal proteins and thus frees some zinc for the use by other proteins, improving survival during zinc starvation.

Sequence Data. Complete genome sequences of E. coli (EC), Salmonella typhi (ST), Yersinia pestis (YP), Vibrio cholerae (VC), Agrobacterium tumefaciens (AT), Brucella melitensis (BM), Sinorhizobium meliloti (SM), Mesorhizobium loti (ML), B. subtilis (BS), Bacillus halodurans (HD), S. pneumoniae (PN), Lactococcus lactis (LL), Streptococcus pyogenes (PY), Staphylococcus aureus (SA), Listeria monocytogenes (LM), and Listeria innocua (LI) were downloaded from GenBank (10). Partially sequenced genomes of Bacillus stearothermophilus (BE), Streptococcus mutans (MU), and Enterococcus faecalis (EF) were extracted from the ERGO database (11). Partially sequenced genomes of Rhodobacter capsulatus (RC), Rhodobacter sphaeroides (RS), Bacillus anthracis (BA), and Klebsiella pneumoniae (KP) were obtained from the web sites of the University of Chicago (http://rhodo.img.cas.cz), University of Texas Health Science Center (www.rhodobacter.org), Institute for Genomic Research (www.tigr.org), and the Washington University Consortium (http://genome.wustl.edu), respectively.

Identification of Regulatory Signals. Common palindromic words in the upstream regions of genes likely to be regulated by the same transcription factor were identified by using the signalx program (12). Positional nucleotide weights in the profiles and Z scores of candidate sites were calculated as described in refs. 12 and 13.

All genomes were divided into groups based on the type of Zn-dependent regulator they contain, and genomes from each group were scanned by using the constructed profile for the respective regulator. We set the cutoffs for profiles such that they produced at most 10 candidate sites per genome in the 300-bp regions upstream of the translation start sites. A string of genes transcribed in the same direction was assumed to form a candidate operon if all intergenic spacers were shorter than 80 bp.

Software. Genomic analyses (protein similarity searches using the Smith–Waterman algorithm, analysis of orthology, and identification of candidate sites in genomic sequences) were done by using genomeexplorer (12). Multiple protein alignments were constructed by using clustal (14). Phylogenetic trees were constructed by using phylip (15). Transmembrane segments in proteins were predicted by tmpred (16); only scores >1,000 were considered significant in the tmpred prediction. A protein similarity search was done by using blast (17), whereas functional analysis was done by using the COG (18) and InterPro (19) databases.

Gene Names. By default, genes in unannotated genomes were given the names of their orthologs in annotated species. Thus, in KP, genes were named as in EC; in MU, genes were named as in PN; in BE, BA, and EF, genes were named as in BS, except for the zinc ABC transporter genes adcABC, which had been experimentally studied in PN (2).

Zinc Regulation and Transport in Gram-Negative Bacteria. Orthologs of the nZUR protein from EC were found in the genomes of γ-proteobacteria ST, KP, YP, and VC and α-proteobacteria AT, BM, SM, ML, RC, and RS. The identity of candidate nZUR proteins with nZUR of EC is 35–45% for α-proteobacteria and 50–90% for γ-proteobacteria.

The phylogenic tree of the nZUR proteins from Gram-negative species (Fig. 1) has three major branches: the Agrobacterium group (AT, BM, SM, and ML), the Rhodobacter group (RC and RS), and the γ-proteobacteria group (EC, ST, KP, YP, and VC). We analyzed the nZUR-binding signals in all three groups separately.

Fig. 1. Phylogenetic tree of ZUR repressors from proteobacteria and consensuses of binding sites. Underlines, center of symmetry of inverted palindromes; boldfaced type, positions coinciding among the three signals.

A 23-bp palindrome GAAATGTTATAWTATAACATTTC was known to serve as an nZUR-binding site upstream of the znuA gene in EC (6). We identified similar sites upstream of orthologous znuA genes of ST and KP. A recognition profile was constructed based on these three nZUR-binding sites. The profile is highly selective: only three to five candidate sites per genome score >5.00.

Table 2, which is published as supporting information on the PNAS web site, www.pnas.org, lists candidate nZUR-binding sites identified in the genomes of γ-proteobacteria. First, candidate nZUR sites were found upstream of the znuA genes in YP and VC. Thus, the identified signal is conserved not only within Enterobacteriaceae but also beyond this family in Vibrionaceae. Second, candidate nZUR sites were found upstream of the b1973 gene in EC (new name zinT, see below) and its orthologs in ST (STY1858) and KP. No zinT orthologs were found in other γ-proteobacteria. Next, a candidate nZUR site was observed upstream of the KP operon paralogous to the znuC-znuB operon (we will further refer to it as znuC2-znuB2). The identity between two pairs of proteins is ≈30%, and znuC2 and znuB2 have no orthologs in other genomes. Finally, we identified candidate nZUR sites upstream of the EC gene ykgM and its orthologs in ST (rpmE2), YP (YP03134), and VC (VC0878). This gene encodes a paralog of the ribosomal protein L31 (see below).

To identify candidate nZUR-binding sites in the Rhodobacter species, we selected the upstream regions of the znuA orthologs from RC and RS and the upstream region of the zinT ortholog from RS. No zinT orthologs were found in RC. A common 18-bp palindrome GATATGTTATAACATATC with few mismatches was found in these regions (Table 2). A profile was constructed based on these three sites; however, it did not identify any new genes preceded by candidate sites in either Rhodobacter species.

A similar procedure was used to identify candidate nZUR-binding sites in the Agrobacterium group. The upstream regions of the znuA orthologs were taken from AT, BM, and SM in addition to the upstream region of the zinT ortholog from AT. No zinT orthologs were found in three other genomes, and no znuA orthologs were found in ML. A common 18-bp palindrome GTAATGTNATNACATTAC with few mismatches was found in all analyzed sequences (Table 2). A profile was constructed based on the identified sites. At most two genes per genome have sites scoring >5.00. Two candidate sites were identified by using this profile (Table 2). The first site is located upstream of the SMc03799 gene (with the new name zinL) in SM. This gene is 55% identical to the BMEII0308 gene of BM annotated as “low-affinity zinc transport membrane protein” (10). The second site was found upstream of the mll8315-mll8314-mll8313 operon in ML. The Mll8315 and Mll8314 proteins are orthologs of ZnuC and ZnuB from AT with the identity of 34% and 30%, respectively. Mll8313 is a hypothetical periplasmic protein that is only 24–28% identical to ZnuA proteins from other species. Because a candidate regulatory site was identified in the upstream region of mll8315-mll8314-mll8313 and no znuA orthologs were found in ML, we propose that mll8313 (with the new name zinA) encodes a zinc-transporting periplasmic protein.

pZUR Regulation in the Bacillus, Staphylococcus, Listeria, and Enterococcus Species. Binding sites of pZUR in the BS genome were found upstream of yciC, yciA, and ycdH (4, 5). The pZUR-binding signal is a 23-bp palindrome, TAAATCGTAATNATTACGATTTA. We will further refer to the ycdH gene as adcA, because it is orthologous to the adcA gene of PN (4, 8).

A recognition profile was constructed based on several known pZUR-binding sites from BS and candidate sites from HD, BE, and SA. Orthologs of the BS pZUR protein were found in the genomes of HD, BE, BA, SA, EF, LM, and LI. Accordingly, these genomes were screened for candidate pZUR-binding sites (Table 2). Three to six candidate sites per genome score >5.80.

We identified candidate pZUR-binding sites upstream of the adcA orthologs in HD, BE, LM, LI, SA, and EF.In LM, LI, and EF as well as in BS, the adcA gene, encoding a lipoprotein, forms a candidate operon with genes orthologous to adcC and adcB of PN, the latter two encoding an ATPase and permease components of a zinc transporter, respectively. In HD, BE, BA, and SA, the adcC and adcB genes are located separately from the adcA gene. Instead, they form operons with zur orthologs. Candidate pZUR-binding sites were found upstream of the latter operons. Further, in LM and LI, a candidate pZUR-regulated operon was found, which was composed of the adcC and adcB paralogs and the zur gene. In EF, the zur gene is located separately from other zinc-related genes, and it is preceded by a strong candidate pZUR-binding site. Thus, in HD, BA, SA, and EF, pZUR seems to act as an autorepressor, whereas in BS, LM, and LI it does not.

Candidate pZUR-binding sites were found upstream of yciC orthologs in HD, BA, and SA. We also found a candidate pZUR-binding site upstream of BH0366 of HD, which encodes a paralog of YciC. Finally, we found a candidate pZUR-binding site upstream of the s14p-l33p-yciC candidate operon in EF. The first two genes of the operon encode proteins paralogous to ribosomal proteins S14 and L33. More candidate pZUR-binding sites were found upstream of genes encoding paralogs of ribosomal proteins, namely S14 paralogs in SA, EF, LM, and LI and S14, L31, and L33 paralogs in BS (see below).

Finally, we found candidate pZUR sites upstream of genes encoding hypothetical proteins: the yciA genes of BS (7) and HD and the zinT (old name yrpE) gene of BS. The yciA gene is located within the yciC locus, which thus has two strong candidate pZUR-binding sites. The detailed analysis of the ZinT structure and function is given below.

YciC is a GTPase of the G3E family (COG0523) (20). All but one member of this COG contain a conserved three-cysteine motif CXCC, where X is I, V, L, or M. This motif has mutated into CVSC in the most distant member of this COG, RV0106 from Mycobacterium tuberculosis, whereas it is still conserved in YNR029c from yeast Saccharomyces cerevisiae. The G3E family also contains the UreG/HypB group of proteins (interpro IPR002894) containing accessory proteins for incorporation of nickel into urease (21) and hydrogenase (22). Thus it is tempting to suggest that YciC performs a similar function of zinc incorporation into some protein.

YciA belongs to COG1469, containing proteins of unknown function. All members of this COG seem to contain the pattern CPC... HXH, where X is L, A, P, Q, N, or R, or variants of this pattern CPH... HKH and CPS... HPH, suggestive of metal binding.

AdcR Regulation in the Streptococcus Group and Structural Characteristics of Zinc-Regulated Pneumococcal Histidine Triad (PHT) Proteins. In PN, zinc uptake is regulated by the AdcR protein belonging to the MarR family (8). AdcR seems to regulate transcription of the adcR-adcC-adcB-adcA operon; however, no AdcR-binding sites have yet been identified. We compared the upstream regions of the adcR-adcC-adcB-adcA operon of PN,the adcR-adcC-adcB operons and the adcA genes of PY and MU. A common 12-bp palindrome, TTAACYRGTTAA, was identified; moreover, in the upstream regions of the adcA genes of PY and MU this palindrome occurred twice. The profile constructed by using these sites selected two similar sites upstream of the zitR-zitS-zitQ-zitP operon of LL (the zitR, zitQ, and zitP genes are orthologous to the adcR, adcC, and adcB genes of PN, respectively; the zitS gene is a homolog of the adcA gene with 42% identity). One more adcA homolog of LL, yndG, is 64% identical to adcA, but this gene is not preceded by a similar palindrome. The constructed profile produces three to eight high-scoring (>5.00) sites per genome (Table 3, which is published as supporting information on the PNAS web site). Because the identified palindromes specifically occur upstream of genes encoding zinc-uptake proteins and some of these genes are known to be AdcR-regulated, we propose that these palindromes are AdcR-binding sites.

Two candidate AdcR-binding sites were found upstream of the SP1002-SP1003 operon in PN. SP1002 encodes a putative adhesion lipoprotein that is 67% identical to the laminin adhesion protein Lmb from Streptococcus agalactiae. We will further refer to SP1002 and its orthologs as Lmb proteins. Notably, it was shown recently that the lmb expression in PY decreases at high Zn²⁺ ion concentration (23). The product of the SP1003 gene (also known as phtD) belongs to the PHT family of proteins. Proteins of this family are characterized by multiple histidine-triad (HXXHXH) motifs (24). Three other genes of the PHT family present in the genome of PN are phtA, phtB, and phtE. They are 65–95% identical to phtD. The phtA, phtB, and phtE genes have one, two, and three high-scoring AdcR-binding sites in the upstream regions, respectively. The PhtE protein contains six histidine-triad motifs (HXXHXH), whereas each of three other proteins contains five motifs. For PhtA, PhtB, PhtD, and, to a lesser extent, PhtE, the localization on the bacterial cell surface has been shown by flow cytometry (24).

Two candidate AdcR sites were found upstream of the orthologous lmb-phtD operons in the genomes of PY, S. agalactiae (AG), and Streptococcus equi (EQ). In S. agalactiae, the Lmb protein was shown to mediate attachment of S. agalactiae to human laminin, which is essential for bacterial colonization of damaged epithelium and translocation of bacteria into the bloodstream (25). All three previously uncharacterized PhtD proteins have strong candidate hydrophobic regions at the N termini identified by the tmpred algorithm and signal peptidase II motifs (LXXC in PhtD of EQ and IXXC in PhtD of PY and AG). Therefore, we expect them all to be exposed on the cell surface, similar to the PHT proteins of PN. Transcriptional coregulation of the lmb and phtD genes suggests a functional link between these proteins. No orthologs of either lmb or phtD genes were found in any other available genome.

We also identified an additional AdcR-regulated gene of PY, SPY1361, encoding a protein with multiple HXXHXH motifs and named it phtY. The PhtY protein contains an N-terminal hydrophobic region and a signal peptidase II motif (LXXC), typical for the PHT family. PhtY consists of three domains. The N-terminal domain (≈360 aa) contains four HXXHXH motifs and is 25% identical to the PhtE protein of PN. The second domain (≈390 aa) is 30–35% identical to a family of internalins from LM. The C-terminal domain consists of 42 aa and is abundant with histidine and aspartic and glutamic acids (HDE), which suggests its strong metal-chelating properties. The internalins of LM allow this bacterium to enter eukaryotic cells. All seven members of the listerial internalin family have been shown to share two structural features: an N-terminal leucine-rich-repeat domain (LRR), followed by a conserved interrepeat region (IR) (26). It has been demonstrated that LRR and IR are both necessary and sufficient for the internalin binding to E-cadherin, and this interaction is critical for the internalin-mediated invasion. (27, 28). Both LRR and IR are present in the N terminus of the second domain of PhtY, which implies that PhtY may play a role in the PY invasion.

Candidate AdcR sites were also found upstream of the LL and PY genes encoding paralogs of the ribosomal protein S14 (rpsN2 and rs14, respectively; see below).

zinT Genes and Their Products. In this study, six genomes contained orthologs of zinT, namely BS (a Gram-positive bacterium from the Bacillus group), EC, ST, and KP (γ-proteobacteria), and AT and RS (α-proteobacteria). Six other genomes (Gram-positive bacteria from the Streptococcus group, namely PN, MU, PY, LL, and EF as well as SA) had zinT fused to genes encoding zinc-binding lipoproteins (adcA orthologs). We will further refer to the zinc-transport lipoprotein domains as ADC and to ZinT domains as ZINT. The fused ADC-ZINT protein of PN was shown experimentally to participate in the zinc transport (2). Note that in LL, there are two paralogous genes similar to adcA, zitS, and yndG. The YndG protein consists of the ADC and ZINT domains, whereas ZitS consists of the ADC domain only.

Analysis by tmpred revealed transmembrane helices at the N termini of all ADC domains as well as at the N termini of ZINT domains in isolated ZinT proteins (Fig. 2). Being attached to the cell surface by the N-terminal region, a ADC-ZINT protein is totally exposed in the extracellular space, and its ZINT domain in particular, which strongly suggests that isolated ZINT domains bearing an N-terminal transmembrane helix also are extracellular and function on the cell surface. The only exception is the YndG protein of LL, which has no predicted transmembrane helix in either the ADC or ZINT domains.

Fig. 2. ZinT domains and regulation of zinT genes. Circles, candidate zinc repressor-binding sites; striped arrows, adcA/znuA; hatched arrows, zinT; black rectangles, transmembrane segment; gray rectangles, histidine/aspartate/glutamate-rich segment.

It is well known that histidine is a rare amino acid with a strong propensity to bind metals. HDE-rich regions were observed in both ADC-ZINT proteins and single-domain ZINT proteins (Fig. 2). Notably, both transmembrane helices and HDE-rich regions were observed only in the isolated ZINT domains, whereas in the fused ADC-ZINT proteins as well as in single-domain AdcA proteins, these structural features occur in the ADC domains; the only exception is the YndG protein from LL, which has no HDE-rich region in either domain. This shows that the isolated ZINT proteins have acquired both these structural features and zinc-dependent regulation. Thus, these features seem to be critical for the ZinT function.

Sequence Analysis and Regulation of Ribosomal Proteins. As shown above, a large number of genes encoding paralogs of ribosomal proteins have been identified under candidate zinc regulation in all analyzed groups of species.

In ref. 8 it was shown that genes encoding four ribosomal proteins, L36, L33, L31, and S14, are each duplicated in several bacterial genomes. If a duplication occurs, the original copies of L36, L33, L31, and S14 contain predicted Zn-ribbon motifs that consist of two pairs of conserved cysteines (in some cases, one cysteine can be replaced by histidine), whereas in the paralogs this motif is usually lost (9). Cysteines in S14 of Thermus thermophilus are indeed involved in zinc binding and formation of the Zn-ribbon domain (29).

To this list we add several duplications not considered in ref. 9. The L31 protein is duplicated in ST and KP. S14 is duplicated in SA, LM, and LI. L33 is duplicated in LM and LI; it is triplicated in BS, SA, PN, MU, and PY. Finally, in EF, S14 is present in two copies, whereas L33 is present in four.

These duplications fit the above-stated rule: the original proteins contain predicted Zn ribbons, but the paralogs do not. The triplicated L33 proteins display the same cysteine pattern as described in ref. 9 for the L33 triplication in LL: one copy has four cysteines (the intact Zn ribbon), another one has three, and the last one has none. Of four L33 copies in EF, again one has four cysteines, another one has three, and the remaining two have no cysteine residues.

With the constructed profiles, we identified candidate nZUR-binding sites upstream of the L31 paralogs (L31p) in EC, ST, YP, and VC (Table 2). Moreover, in YP and VC, L31p forms a candidate operon with the L36 paralog (L36p); thus both ribosomal protein paralogs in YP and VC seem to be under the nZUR regulation. Further, we observed candidate pZUR-binding sites upstream of the L31p and L33p genes in BS and upstream of the S14p gene in BS, SA, EF, LM, and LI (Table 2). In EF, S14p forms a candidate operon with the L33p gene and, notably, with the yciC gene encoding a component of the low-affinity zinc transporter. As mentioned above, the latter gene was shown to be regulated by pZUR in BS (4). Finally, we identified candidate AdcR-binding sites upstream of the S14p genes in PY and LL (Table 3).

Table 1 summarizes the known and new data about duplications of the Zn-ribbon ribosomal proteins in the analyzed groups of bacteria. Thus far all duplications of ribosomal proteins in the analyzed genomes except for the L31 duplication in KP and L33 duplication in Listeria fit the following rule: if one copy retains a Zn ribbon and the other one lacks it, then the gene encoding the copy without a Zn-ribbon motif is regulated by a Zn-dependent repressor and therefore is induced under Zn-restricted conditions.

Table 1. Ribosomal proteins L36, L33, L31, and S14 and their paralogs

The triplication of L33 in SA, PN, MU, PY, and LL represent an exception, because none of the copies are regulated. However, one of three copies of L33 in BS seems to be regulated by pZUR. Additionally, in EF, where L33 is present in four copies, one of the no-cysteines copies is located in a zinc-regulated operon.

Four Regulators That Mediate Bacterial Response to Zinc Starvation and Their Regulons in Various Microbial Species. Zinc-dependent regulation involves four transcriptional factors in four different groups of bacterial species: nZUR in γ-proteobacteria; orthologous nZUR proteins in α-proteobacteria; pZUR in the Bacillus group; and AdcR in the Streptococcus group. ZUR-like proteins were observed also in other proteobacteria not considered here, in particular Pseudomonas aeruginosa, Xylella fastidiosa, and Neisseria gonorrhoeae (3). The DNA-binding signal of the nZUR repressor in γ-proteobacteria (GAAATGTTATAWTATAACATTTC) is conserved through the Vibrionaceae group. We identified previously unknown DNA-binding signals for the nZUR repressors in two groups of α-proteobacteria, the Agrobacterium (GTAATGTAATAACATTAC) and Rhodobacter (GATATGTTATAACATATC) groups (Fig. 1). Further, we identified DNA-binding signal for the AdcR regulator in the Streptococcus group (TTAACYRGTTAA). The pZUR-binding signal in the Bacillus group is TAAATCGTAATNATTACGATTTA (5).

The signal of γ-proteobacteria is an inverted repeat of the highly conserved 9-bp box GAAATGTTA, separated by a 5-bp AT-rich spacer. This box is similar to the symmetric boxes of the nZUR signals from α-proteobacteria. In particular, the Agrobacterium signal has two differences (T instead of A in position 2 and A instead of T in position 8), and the Rhodobacter signal has only one (T instead of A in position 3). Thus it is likely that the DNA-binding surface of the nZUR proteins in these genomes has not evolved much, and the main difference, a larger spacer in γ-proteobacteria, is caused by a change in the mutual orientation of the two subunits in the nZUR dimer.

We identified a number of genes encoding various components of zinc transporters and predicted that they are regulated by zinc repressors. Among these transporters are orthologs and paralogs of the high-affinity zinc ABC transporters AdcABC/ZnuABC and the low-affinity zinc transporter YciABC. The specificity of transporters is known to evolve rapidly even in orthologous systems. Thus candidate zinc repressor-binding sites upstream of genes encoding such systems allow us to assign zinc specificity to these systems with higher confidence.

Zinc-dependent regulators act as autorepressors in most of the analyzed species. However, the distribution of autoregulation does not fit the phylogenic pattern. Indeed, of two closely related species, one may have an autoregulatory protein, whereas the other may not (e.g., SM and ML or BS and HD). At the same time, genes encoding nonautoregulatory proteins always lie isolatedly, whereas genes encoding autoregulators are usually belong to zinc transporter operons. Thus, the autoregulation of zinc repressors seems unessential and is readily lost once the regulator gene separates from the transporter genes. No autoregulation of zur was observed in E. coli (3), which is in agreement with the absence of candidate pZUR-binding sites upstream of zur.

Three previously uncharacterized functional classes of proteins regulated by zinc were identified: ZinT proteins, PHT proteins, and paralogs of ribosomal proteins.

ZinT Represents a Previously Uncharacterized Type of Zinc-Binding Outer Membrane Components of ABC Transporters in Gram-Positive and Gram-Negative Bacteria. The zinT gene is present in bacterial genomes either as an isolated ZINT domain or a part of the ADC-ZINT fusion, where ADC is a zinc-binding component of the zinc ABC transporter. Both isolated ZINT domains and the ADC-ZINT fusions appear to be exposed to the cell surface, contain metal-chelating HDE-rich regions, and be regulated by zinc repressors. The only exception is the fusion in LL (the YndG protein), which has lost all three features. Candidate binding sites for zinc repressors upstream of the zinT and adcA genes strongly suggest that HDE-rich regions in the ZINT and ADC domains serve for zinc binding. These observations lead to the conclusion that ZinT is a previously uncharacterized type of zinc-binding protein, which likely functions as an alternative or additional zinc-chelating component of the zinc ABC transporter. ACD-ZINT fusions suggest that ZinT may associate with the same ATP-binding and permease proteins as AdcA.

We also suggest that the yndG gene of LL does not encode a functional protein, because it has lost both the regulatory site and the obligatory structural features. Alternatively, it might have changed the function dramatically and is not involved in the zinc acquisition any more.

The PHT Proteins: A Candidate Family of Adhesins in Streptococci. The PHT protein family is restricted to the genus Streptococcus. All members of this family contain four to six histidine-triad motifs (HXXHXH) and seem to be under zinc control either directly, by candidate AdcR-binding sites in the upstream regions, or as members of the AdcR-regulated lmb-phtD operons. The metal-chelating property of histidine together with the predicted regulation of PHT proteins by zinc repressors suggest that histidine-triad motifs in these proteins are involved in zinc binding. However, these motifs differ from HDE-rich regions that function as zinc-adsorption sites of zinc transporters. The latter have no regularity of histidine positions and are abundant with aspartic and glutamic acids in addition to histidine. On the contrary, positions of histidines within triads in the PHT proteins are strongly conserved, and these histidines are surrounded by aromatic amino acids rather than negatively charged ones. This leads to the conclusion that histidine-triad motifs likely play a structural or functional role in the PHT proteins.

The PHT proteins of PN are located on the cell surface. The phtD gene forms a candidate operon with the lmb gene encoding a putative laminin adhesion protein, which is probably involved in the colonization of the human epithelium by streptococci and their subsequent invasion into the bloodstream. The PY lmb mutant showed reduced laminin binding, adherence, and internalization into epithelial cells (23). The transcriptional association of the phtD and lmb genes may indicate a functional link and point to a possible role of the PHT proteins in Streptococcus adhesion and invasion.

This hypothesis is supported further by analysis of the AdcR-regulated PhtY protein of PY. PhtY consists of three domains: the N-terminal PHT domain, the middle domain that has two characteristic features of internalin proteins of the Listeria species (LRR and IR), and the C-terminal HDE-rich domain. The LRR and IR were shown to be both necessary and sufficient for the internalin-mediated invasion of Listeria species. Thus, it is likely that the PhtY protein plays a role in the invasion of PY. The C-terminal HDE-rich domain of PhtY might be involved in zinc scavenging before zinc incorporation into histidine-triad motifs of the PHT domain.

The first major reservoir for streptococcal infections is the human oral/nasal mucosa. After successful colonization of the nasopharynx mucosa, a small fraction of the streptococci invades the epithelium cells. Invasion is considered a multistage process initiated by adherence. As a result, invasion provides a window for streptococci to reach deeper tissues including the bloodstream (30, 31). Zinc concentration in bronchoalveolar lavages is 5- to 10-fold lower than in the human plasma (32, 33). Thus we suggest the following very speculative scenario. At the initial stage of Streptococcus infection in the human nasopharynx, bacteria face zinc-restricted environment and induce expression of the Lmb and PHT proteins, which are likely involved in the adhesion and invasion processes. It takes time for the PHT proteins to archive a functional conformation, because they have to bind zinc ions by the histidine-triad motifs. Indeed, in accordance with this hypothesis, zinc was shown to stimulate the protein synthesis-independent adhesion of PY (34). The molecular mechanism of this adhesion still remains unknown, but we expect it to be mediated by the PHT proteins. After streptococci have reached the bloodstream, they face zinc-sufficient conditions, and expression of the PHT proteins and zinc transporters is blocked. This may function as a protective mechanism that allows bacteria to avoid adhesion to macrophages and extermination by the human immune system.

Paralogs of Ribosomal Proteins Help Bacteria to Survive During Zinc Starvation. Only four ribosomal proteins, L36, L33, L31, and S14, are duplicated in more than one bacterial genome, and the original copies of these proteins contain the Zn-ribbon motif. This motif was suggested to play a role in the ribosomal stability under high temperatures in thermophilic Archaea and bacteria (9), whereas it is not absolutely indispensable under normal conditions. The present study demonstrated that almost all genes encoding the paralogs of L36, L33, L31, and S14 proteins in the analyzed species are likely to be regulated by zinc (Table 1), and all zinc-regulated copies have lost the Zn ribbons in contrast to the original proteins that retain these motifs. Because there are four different proteins and three different regulators involved, this seems to be a general rule rather than a set of coincidences. We propose the following scenario for the L36p, L33p, L31p, and S14p function in bacteria (Fig. 3). In zinc-rich conditions, the L36p, L33p, L31p, and S14p genes are repressed by the respective regulators, whereas the original L36, L33, L31, and S14 proteins function in ribosomes. In zinc-depleted conditions, however, the original Zn-ribbon-containing L36, L33, L31, and S14 proteins would use all available zinc, leading to zinc starvation for other cellular proteins that require zinc for their function (indeed, the number of ribosomes is orders of magnitude larger than the number of zinc-containing enzymes such as DNA polymerase, primase, etc.). To prevent this situation, non-Zn-ribbon paralogs are expressed under zinc-restricted conditions. They partially replace the original proteins, which results in releasing some zinc into the cytoplasm. This zinc then can be incorporated into other zinc-requiring proteins.

Fig. 3. The predicted regulation of ribosomal proteins by zinc. (Upper) Zn-rich conditions. The ribosomes contain the protein with functional Zn ribbon (black circles), and transcription of the gene encoding the paralog (gray arrow) is inhibited by the zinc (more ...)

This situation resembles the correlation between atomic composition and metabolic function of enzymes (35): enzymes involved in assimilation of sulfur and thus switched on in conditions of sulfur starvation are depleted in sulfur-containing amino acids. Similarly, the carbon composition of carbon-assimilation enzymes is less than average. Similarly, zinc-independent paralog proteins are switched on in the conditions of zinc starvation.

We are grateful to Eugene Koonin, Olga Dontsova, Alexey Kazakov, Dmitry Rodionov, and Alexey Vitreschak for useful discussions. This study was partially supported by Howard Hughes Medical Institute Grant 55000309.