Functional equivalence of HMGA- and histone H1-like domains in a bacterial transcriptional factor

Abstract

Histone H1 and high-mobility group A (HMGA) proteins compete dynamically to modulate chromatin structure and regulate DNA transactions in eukaryotes. In prokaryotes, HMGA-like domains are known only in Myxococcus xanthus CarD and its Stigmatella aurantiaca ortholog. These have an N-terminal module absent in HMGA that interacts with CarG (a zinc-associated factor that does not bind DNA) to form a stable complex essential in regulating multicellular development, light-induced carotenogenesis, and other cellular processes. An analogous pair, CarD_Ad and CarG_Ad, exists in another myxobacterium, Anaeromyxobacter dehalogenans. Intriguingly, the CarD_Ad C terminus lacks the hallmark HMGA DNA-binding AT-hooks and instead resembles the C-terminal region (CTR) of histone H1. We find that CarD_Ad alone could not replace CarD in M. xanthus. By contrast, when introduced with CarG_Ad, CarD_Ad functionally replaced CarD in regulating not just 1 but 3 distinct processes in M. xanthus, despite the lower DNA-binding affinity of CarD_Ad versus CarD in vitro. The ability of the cognate CarD_Ad–CarG_Ad pair to interact, but not the noncognate CarD_Ad–CarG, rationalizes these data. Thus, in chimeras that conserve CarD–CarG interactions, the H1-like CTR of CarD_Ad could replace the CarD HMGA AT-hooks with no loss of function in vivo. More tellingly, even chimeras with the CarD AT-hook region substituted by human histone H1 CTR or full-length H1 functioned in M. xanthus. Our domain-swap analyses showing functional equivalence of HMGA AT-hooks and H1 CTR in prokaryotic transcriptional regulation provide molecular insights into possible modes of action underlying their biological roles.

In eukaryotes, the DNA architectural factors histone H1 and high-mobility group (HMG) proteins remodel chromatin structure and modulate the action of other regulatory factors in various DNA transactions (1). H1 (or linker histone) interacts with linker DNA at or near the nucleosomal dyad axis, stabilizes higher-order chromatin structure, and lowers nucleosomal access to transcriptional factors so as to up- or downregulate their action; HMG proteins compete with H1 for overlapping binding sites in chromatin to decrease nucleosomal compactness and provide access to regulatory factors (1). The typical H1 structure consists of a central globular winged-helix domain flanked by 2 basic, randomly structured segments: a short N-terminal stretch and an ≈100-residue C-terminal region (CTR) (1). The primary sequence of CTR, crucial for DNA binding and chromatin condensation, diverges among different species and isoforms but its amino acid composition is fairly constant in most of the higher eukaryotes: ≈40% K, 20–35% A, ≈15% P, few S, T, G, or V, and virtually no aromatic residues (1–5). Proteins of the HMG superfamily are the second-most abundant components of chromatin after histones. Of these, high-mobility group A (HMGA) are small (≤107 residues), intrinsically disordered proteins characterized by multiple repeats of the conserved RGRP or “AT-hook” DNA-binding motif embedded in a less conserved cluster of basic and proline residues, and flanked by an acidic region that modulates protein stability and DNA binding (6, 7). The AT-hooks, 3 of which occur in mammalian HMGA, bind in a defined conformation to the minor groove of AT-rich sequences occurring in at least 2 appropriately spaced tracts of 4 to 8 bp in length (8–11).

H1 and, more so, HMGA-like domains are rare in prokaryotes. The only known examples of bacterial proteins with HMGA-like domains occur in myxobacteria: Myxococcus xanthus CarD and its Stigmatella aurantiaca ortholog, CarD_Sa [(12, 13); supporting information (SI) Fig. S1A]. Myxobacteria are unique among prokaryotes in carrying out a programmed multicellular developmental process on starvation (14–16). Temporal and spatial control of gene expression during M. xanthus development depends on eukaryotic-like signal transduction proteins and transcription factors of which CarD is one (17–19). Additionally, CarD is involved in regulating other processes, such as light-induced carotenogenesis (19, 20). Its 136-residue C-terminal region, with a highly acidic stretch preceding 4 AT-hooks, resembles human HMGA1a in its physical, structural, and DNA-binding properties (21). This HMGA-like segment is linked to a 180-residue N-terminal domain, CarDNter, essential for function and absent in eukaryotic HMGA (13). Every known CarD-regulated activity requires physical association of CarDNter to CarG; the latter binds zinc via a H/C-rich motif of the type found in archaemetzincins (a class of metalloproteases) but lacks protease or DNA-binding activity (22). The myxobacterium Anaeromyxobacter dehalogenans also expresses sequence analogs to both CarD (CarD_Ad) and CarG (CarG_Ad). Strikingly, whereas CarD_Ad has an N-terminal domain very similar to that in CarD, its C-terminal end lacks the hallmark AT-hooks. Instead, its K/A/P content is typical of H1 CTRs.

Whether an H1 domain can functionally replace the HMGA one in CarD is studied here. Remarkably, we find that an H1-CTR like domain, whether from CarD_Ad or, even more notably, from human histone H1.2, can supply the DNA-binding activity essential for CarD function in 3 distinct processes in M. xanthus just as well as the naturally occurring AT-hooks. Our study provides insights into the modular design of this rather singular bacterial transcriptional factor. More importantly, it reveals that AT-hooks and H1 CTR can function equivalently in this system, and provides molecular insights that may be relevant in understanding their modes of action.

Results

CarD_Ad Alone Cannot Replace CarD in M. xanthus.

CarD_Ad and CarD share 58% sequence identity (70% similarity) in their N-terminal region and 37% (45%) in their C-terminal part (Fig. S1). Furthermore, CarD_Ad, like CarD, has a C-terminal region with a highly acidic stretch preceding a very basic one, with the most abundant amino acids in the latter region being K, A, and P in both proteins (Fig. S1D). However, the basic region of CarD_Ad stands out from that of CarD in lacking the hallmark AT-hooks, besides having K/A/P contents closer to those observed in H1 CTRs (Fig. 1A; Fig. S1D). This and the reported competition between H1 and HMGA for binding to DNA in eukaryotes (1) led us to examine whether CarD_Ad could replace CarD in M. xanthus. For this, a plasmid bearing carD_Ad was electroporated into a ΔcarD M. xanthus strain, where it integrates into the chromosome at the carD locus by homologous recombination resulting in merodiploids. Haploid strains with the carD_Ad allele replacing carD were then isolated as described (see SI Materials and Methods) to test how CarD_Ad affects the following distinct processes dependent on CarD (and CarG): (i) activation of P_QRS, the promoter that on illumination drives transcription of the regulatory carQRS operon, thereby leading to expression of the structural genes for carotenoid synthesis; (ii) starvation-induced fruiting-body formation; and (iii) expression of the ddvA locus, unrelated to the light- or starvation-induced responses, whose function remains to be defined (20, 22, 23).

Fig. 1.

Functional replacement of M. xanthus carD by carD_Ad in the absence or presence of carG_Ad. (A) Schematic representation of CarD, human HMGA1a, CarD_Ad, and human H1 (H1.2 subtype). The AT-hooks, the acidic region, and the H1 CTR are shown as boxes with “+,” “−,” and “X,” respectively. Black boxes in CarD and CarD_Ad, and the line in human HMGA represent N-terminal domains. In human H1, the short N-terminal domain is shown by the unfilled box and the globular domain by the gray box. Residue numbers demarcate the domains. (B) Color phenotypes (Top) of cell spots after overnight growth on CTT plates in the dark (“D”) or in the light (“L”) vertically aligned with the corresponding expression levels of carQ′::lacZ in strains bearing the indicated carD and carG alleles (Bottom). Cell cultures were grown to early exponential phase in the dark, divided into 2, grown for a further 14 h, one in the dark (filled bars) and the other in the light (empty bars), and specific β-galactosidase activities were estimated. (C) Developmental phenotype (Top) after 5-day incubation of 15-μL cell droplets (1.25 × 10⁸ cells ml⁻¹) spotted on TPM agar vertically aligned with the corresponding expression levels of the Ω4435 developmental marker in strains bearing the indicated carD and carG alleles (Bottom). Specific β-galactosidase activities were determined for samples collected after 24-h development on TPM agar. (D) Expression of ddvA::Tn5-lac in exponentially growing cultures (OD₅₅₀ = 0.7) of strains containing the indicated carD and carG alleles. In B, C, and D, β-galactosidase activities (in nanomoles of o-nitrophenyl β-D-galactoside hydrolyzed/min/mg protein), shown as a percentage of the wild-type values (128 ± 20 in B, 240 ± 30 in C, and 2900 ± 300 in D), are from 2 or more independent measurements.

In contrast to wild-type M. xanthus, which is yellow in the dark and turns red in the light due to carotenogenesis (the Car⁺ phenotype), a strain with carD replaced by carD_Ad remained yellow in the light (Car⁻), like those with deletions of carD, carG, or both (Fig. 1B; ref. 22). Consistent with this, expression levels of the carQ′::lacZ reporter probe, where lacZ is under P_QRS control, were close to those observed for strains with carD, carG, or both deleted (Fig. 1B). Furthermore, the strain with carD_Ad in place of carD did not produce normal, mature spore-filled fruiting bodies on starvation (Fru⁻ phenotype), in contrast to the wild-type Fru⁺ cells (Fig. 1C). In agreement with this, the strain with carD_Ad failed to express the Ω4435 Tn5lac transposon insertion, which is one of the developmental markers known to depend on CarD (19). Finally, a reporter lacZ transposon probe at the ddvA locus (ddvA::Tn5lac) was expressed in the strain with carD_Ad at levels far lower than in wild-type and similar to those in the ΔcarD ΔcarG strain (Fig. 1D). Thus, CarD_Ad on its own cannot compensate for the lack of CarD in M. xanthus. This is unlikely to be the result of low levels of carD_Ad expression in M. xanthus, because it maintains both the high GC content and the codon third-position G + C bias of carD. Other possible causes for why CarD_Ad differs from CarD in vivo might be in its binding to DNA and/or to CarG. These were therefore examined next.

Human HMGA binds to the minor groove of appropriately spaced AT-rich tracts, as noted above (8–11). We have demonstrated that the CarD HMGA-like domain resembles human HMGA1a not only physically and structurally, but also in binding to specific human HMGA DNA sites, such as the IFN response element (21). In M. xanthus, CarD binding to DNA has been well characterized only at the light-inducible P_QRS promoter, where its HMGA-like domain binds to 2 appropriately spaced AT-rich tracts located in the segment between −65 and −77 upstream of the transcription start site (12, 21–23). Human HMGA1a also exhibits preferential binding to this site in vitro (21) and yields a DNase I footprint that resembles that of CarD (Fig. S2), showing that HMGA1a and CarD bind at P_QRS similarly. Two DNA probes, of lengths 37 bp and 169 bp, containing the AT-rich tracts at P_QRS, bind CarD in gel mobility shift assays (lanes 2–4, Top and Bottom gels, respectively, in Fig. 2A). CarD_Ad also bound to these probes, but at over a 10-fold higher concentration (lanes 5–10, Top and Bottom gels in Fig. 2A). In DNase I footprinting assays with the 169-bp probe, the segment between −65 to −77 containing the CarD-binding site was protected by CarD (lanes 2–5, Fig. 2B) but not by CarD_Ad at comparable concentrations (lanes 7–10, Fig. 2B), consistent with the lower affinity of CarD_Ad inferred from gel-shift data. Higher CarD_Ad levels yielded protection extending throughout the DNA probe, indicative of nonspecific binding (data not shown). Thus, CarD binds to its site at P_QRS with a higher affinity in vitro than CarD_Ad, reminiscent of eukaryotic HMGA1a binding to a variety of DNA substrates such as AT-rich B-form DNA, 4-way junctions, or kinked, distorted structures with significantly greater affinities than H1 (24, 25). This could rationalize, at least in part, why CarD_Ad does not effectively replace CarD in vivo. However, as described next, a more critical determinant appears to be the lack of its cognate CarG_Ad partner.

Fig. 2.

DNA binding of CarD and CarD_Ad in vitro. (A) Gel mobility shift assay with ³²P-labeled DNA probes of 37 bp (Top gel) and 169 bp (Bottom gel) containing the CarD-binding site at the P_QRS promoter region. Lane 1: no protein; lanes 2 to 4: 0.19, 0.75 and 1.5 μM CarD; lanes 5 to 10: 0.19, 0.38, 0.75, 1.5, 3, and 6 μM CarD_Ad. The 37-bp and 169-bp probes (coding strand) are shown schematically on Top, the numbers indicating positions relative to the transcription start site of the carQRS operon. The CarD-binding sequence between positions −77 and −65 is also shown with the 2 AT-rich tracts in uppercase. (B) DNase I footprint of the 169-bp probe with the coding strand labeled. Protein concentrations used were: 0, 3.75, 7.5, 11.25, and 15 μM of CarD (lanes 1 to 5) or CarD_Ad (lanes 6 to 10). The CarD-binding site is marked on the Left.

CarD_Ad Can Replace CarD in M. xanthus if CarG_Ad Is also Present.

All known CarD-dependent processes require CarG, a zinc-bound protein with no DNA-binding activity, which interacts physically with CarDNter to form a stable regulatory complex (22). In M. xanthus, carD and carG are contiguous genes in an operon, as are carD_Ad and carG_Ad in A. dehalogenans, where overlap of the carD_Ad stop codon and the carG_Ad start codon suggests translational coupling. CarG_Ad is 46% identical (62% similar) to CarG and conserves the structurally critical Zn-binding motif (Fig. S1). To test if it is the presence of CarG rather than the cognate CarG_Ad that makes CarD_Ad unable to replace CarD in M. xanthus, a plasmid bearing the contiguous carD_Ad and carG_Ad genes was constructed and electroporated into the ΔcarD ΔcarG strain. Chromosomal integration of the plasmid by homologous recombination resulted in merodiploids that were Car⁺, suggesting that the simultaneous presence of carD_Ad and carG_Ad can complement the lack of carD and carG. Consistent with this, the haploid strain with carD_Ad and carG_Ad (generated as described in SI Materials and Methods) was also Car⁺, in marked contrast to the Car⁻ phenotype observed on replacing only carD by carD_Ad. Accordingly, equivalent levels of reporter carQ′::lacZ expression were observed in the light for the wild-type strain and for that with the carD_Ad–carG_Ad pair (Fig. 1B). Furthermore, the latter strain was Fru⁺, with normal expression of the Ω4435 developmental marker (Fig. 1C), and of the ddvA::Tn5lac probe (Fig. 1D). Thus, despite the apparently lower DNA-binding affinity of CarD_Ad relative to CarD in vitro, CarD_Ad can functionally replace CarD in regulating not just 1 but several distinct processes in M. xanthus, so long as the cognate CarG_Ad partner is also present.

CarD_Ad Forms a Stable Complex with CarG_Ad but Not with CarG.

Why the CarD_Ad–CarG_Ad pair, but not the mixed CarD_Ad–CarG pair, functions in M. xanthus may be the result of differences in interactions between cognate and noncognate CarD–CarG pairs. This was checked next. A bacterial 2-hybrid system in which interaction between 2 test proteins leads to functional complementation between the T25 and T18 fragments of the catalytic domain of Bordetella pertussis adenylate cyclase (26) demonstrated that CarD or CarDNter interacts with CarG (22). The same analysis showed this to be true for CarD_Ad and CarG_Ad as well (Fig. 3A). By contrast, CarG did not interact with the CarD_Ad N-terminal domain (Fig. 3A) or CarD_Ad (results not shown).

Fig. 3.

Physical interactions between cognate and noncognate CarD–CarG pairs. (A) Two-hybrid analysis in E. coli, showing reporter lacZ expression for the different pairs as indicated. “C⁻” is the negative control, where cells contain the vectors without fusions. Values are from 2 or more independent measurements. (B) Analytical gel filtration using a Superdex200 column. Protein concentrations were ≥10 μM. Elution profiles as tracked by absorbance at 220 nm in arbitrary units (a.u.) are shown. Bottom panel: pure CarD (black solid line), CarG (black dashed line), CarD_Ad (gray solid line displaced vertically for easy viewing), and CarG_Ad (gray dashed line). Top panel: a mixture of CarD and CarG (black solid line), of CarD_Ad and CarG_Ad (gray solid line), and of CarD_Ad and CarG (black dashed line). On Top is shown M_r (in kDa) aligned with the corresponding elution volume.

Analytical gel filtration using purified proteins further confirmed these findings in vitro. We have shown elsewhere that native CarD is an elongated dimer with a globular N-terminal dimerization domain and an extended C-terminal domain, and although its calculated molecular weight (MW) is 34 kDa it elutes off a Superdex200 column with an apparent MW, M_r, of (118 ± 7) kDa (21). CarG (calculated MW = 19 kDa) elutes as a compact monomer with M_r of (17 ± 1) kDa but, when mixed with ≥2-fold excess of CarD, it coelutes with the latter as a stable complex with M_r of ≈129 kDa, with the pure CarG peak undetected (22; Fig. 3B). Purified CarD_Ad, whose calculated MW is ≈42 kDa, eluted off Superdex200 with a 4-fold higher M_r of (166 ± 8) kDa. This behavior, which parallels that of CarD, suggests that CarD_Ad is also elongated with a globular N-terminal dimerization domain and an extended C-terminal domain; CarG_Ad (calculated MW ≈ 20 kDa) eluted with M_r of (18 ± 1) kDa and so appears to be compact like CarG (Fig. 3B, Bottom panel). The pure CarG_Ad peak disappeared on mixing with excess CarD_Ad, and both proteins coeluted with M_r ≈ 200 kDa (Fig. 3B, Top panel). In contrast, CarG with excess CarD_Ad yielded only the pure protein peaks (Fig. 3B, Top panel). Therefore, the cognate CarD_Ad–CarG_Ad pair forms a stable complex (like CarD and CarG) but the mixed CarD_Ad–CarG pair does not.

CarD Chimeras Containing the CarD_Ad C-Terminal Segment Function in M. xanthus.

CarD_Ad, despite its lower DNA-binding affinity in vitro, could replace CarD in vivo if CarG_Ad was also present. This led us to infer that so long as CarD–CarG interactions are maintained (which occurs only with the cognate pairs), differences resulting from the HMGA-like domain of CarD versus the H1-like one of CarD_Ad do not significantly affect function. In other words, the CarD HMGA-like domain and the CarD_Ad H1-like domain may be interchangeable in vivo. We therefore tested if CarD chimeras that retained the N-terminal part (and thus, interactions with CarG), but in which the HMGA domain was replaced by the CarD_Ad H1 CTR-like segment, could function in M. xanthus. Two such chimeras were examined: in 1 (C1), the CarD_Ad segment spanning the H1-like basic region and the preceding highly acidic region replaced the entire CarD HMGA-like domain; in the other (C2), the basic H1-like domain in CarD_Ad substituted the basic AT-hooks of CarD (Fig. 4A). An acidic region therefore exists in both C1 and C2, since lack of the acidic region decreases CarD stability (21) and impairs function in vivo (Fig. S3). To test if C1 and C2 function in M. xanthus, plasmid vectors bearing the corresponding genes were independently introduced into the ΔcarD strain, where they incorporate into the chromosome by homologous recombination to generate merodiploids. Both types of merodiploids turned red in the light, suggesting that C1 and C2 can complement the Car⁻ phenotype provoked by the ΔcarD allele. Haploid strains expressing C1 or C2 were isolated as described (see SI Materials and Methods) and confirmed to be Car⁺ (Fig. S4A). Both strains also recovered the Fru⁺ developmental phenotype (Fig. 4C). Furthermore, the strains with C1 or C2 showed levels of reporter lacZ expression within ±20% of the wild-type for the carQ′::lacZ (Fig. 4B), Ω4435 (Fig. 4C), and ddvA::Tn5lac (Fig. 4D) probes, confirming that the CarD_Ad H1-like domain can replace the CarD HMGA domain in vivo.

Fig. 4.

CarD chimeras with a CarD_Ad C-terminal domain are functional in M. xanthus. (A) Schematic representation of CarD chimeras C1 and C2. CarD and CarD_Ad segments in each chimera are indicated with numbers and boxes are coded as in Fig. 1A. (B) Expression of carQ′::lacZ in the dark (filled bars) and in the light (empty bars) for strains bearing the wild-type carD (C⁺), ΔcarD (C⁻), C1, or C2 alleles. (C) Developmental phenotype on TPM agar for strains as in A, vertically aligned with the corresponding expression levels of Ω4435. (D) Expression of ddvA::Tn5-lac for strains as in A. β-galactosidase activities in B, C, and D are shown and estimated as in Fig. 1.

CarD Chimeras with Human HMGA or Histone H1 Are Functional in M. xanthus.

The above results were sufficiently intriguing to prompt us to examine whether other variants of HMGA or H1 domains in CarD are equally capable of functioning in M. xanthus. We checked this using human HMGA1a or histone H1 (its H1.2 subtype) in chimeric CarD proteins. Human HMGA1a, as noted previously, resembles the CarD C-terminal HMGA-like domain in its physical, structural, and DNA-binding properties (21). However, human HMGA1a has 1 less AT-hook and its acidic region (significantly shorter than in CarD) flanks the C- rather than the N-terminal side of the AT-hook segment (Fig. 1A). Likewise, although the C-terminal region of CarD_Ad is rich in K/A/P like H1 CTRs, its sequence and overall amino acid composition vary from that in human H1.2, and the latter's globular and N-terminal domains do not occur in CarD_Ad. The globular domain has been implicated in positioning H1 to the nucleosome after the initial binding of CTR to linker DNA, while the N-terminal region may modulate the binding affinity of H1 to chromatin (1).

Fig. 5A depicts schematically the CarD chimeras with segments corresponding to human HMGA1a or histone H1.2 subtype that were examined. The CarD segment spanning residues 1 to 179 was fused to HMGA1a in chimera C3, while in C4, C5, and C6, the CarD stretch from residues 1 to 226 was linked to the HMGA1a AT-hook segment between residues 2 and 90, the entire H1.2, or the H1.2 CTR part (residues 110–213), respectively. C4, C5, and C6 thus retain the CarD acidic segment (residues 180–226) while C3 has the HMGA1a acidic segment that, unlike in CarD, is at the C-terminal end and follows the AT-hook segment. Plasmid vectors bearing the coding sequences for each chimera were constructed using the natural coding sequence for human HMGA1a and a synthetic coding sequence for H1.2, which was optimized for high GC content and G + C bias at the third codon position. These were then introduced into the ΔcarD M. xanthus strain by electroporation, where each construct incorporates into the chromosome by homologous recombination at the carD locus, as described earlier. Unlike the Car⁻ ΔcarD recipient strain, the merodiploids with each of these chimeras turned red on illumination with blue light. Haploid strains expressing each of these chimeras were isolated as described (see SI Materials and Methods) and confirmed to be Car⁺ (Fig. S4A). Moreover, they were all Fru⁺ (Fig. 5C). Expression of the reporter lacZ probes carQ′::lacZ (Fig. 5B), Ω4435 (Fig. 5C), and ddvA::Tn5lac (Fig. 5D) in the strains expressing any 1 of the 4 chimeras was within ±20% of the wild type. Thus, in all 3 CarD-dependent processes examined, human HMGA1a or H1 could replace the CarD HMGA domain in vivo. The results demonstrate that CarD remains functional despite 1 less AT-hook or a different juxtaposition of the acidic and AT-hook segments. Moreover, given that H1 CTR alone appears to function as well as whole H1.2, we can conclude that the H1 globular domain together with its N-terminal region does not contribute to CarD activity in vivo, consistent with the roles of these 2 domains being more pertinent in the context of eukaryotic chromatin.

Fig. 5.

CarD chimeras with human HMGA1a or histone H1.2, but not protamine 2 or cytochrome C, function in M. xanthus. (A) Schematic representation of CarD chimeras C3 to C8. Segments corresponding to CarD, HMGA1a, or H1.2 in each chimera are indicated with numbers and boxes and are coded as in Fig. 1A. (B) Expression of carQ′::lacZ in the dark (filled bars) and in the light (empty bars) for strains bearing the wild-type carD (C⁺), ΔcarD (C⁻), or the indicated chimeric alleles. (C) Developmental phenotype on TPM agar for strains as in A, vertically aligned with the corresponding expression levels of Ω4435. (D) Expression of ddvA::Tn5-lac in strains as in A. In B, C, and D, β-galactosidase activities are shown and estimated as in Fig. 1.

We next checked if any basic polypeptide could function as the CarD AT-hook region by constructing 2 additional chimeras, C7 and C8 (Fig. 5A). In C7, the AT-hooks are replaced by the 102-residue, R-rich human protamine 2 (Prt2; theoretical pI = 11.9), which lacks a defined structure and binds DNA in vitro like H1 CTR but with higher affinity (27, 28). In C8, the AT-hooks are replaced by the 105-residue human cytochrome C (CytC; theoretical pI = 9.58), which has defined structure and can bind DNA because of its positive charge although it is not a proper DNA-binding protein (29). Plasmid vectors bearing DNA coding for CarD residues 1 to 226 plus synthetic sequences coding for Prt2 or CytC (both optimized for the high GC content and codon usage in M. xanthus, as with H1.2 above) were introduced into the ΔcarD M. xanthus strain, and merodiploids and haploids expressing chimeras C7 or C8 were isolated as before. Unlike strains bearing chimeras with H1 CTR, those with C7 or C8 were Car⁻ Fru⁻ (Fig. 5C, Fig. S4B). In agreement with this, expression of the carQ′::lacZ and the Ω4435 reporter probes in strains with C7 or C8 was low or negligible relative to wild type (similar to the ΔcarD strain for C7, and ≈10–20% for C8; Fig. 5 B and C). Expression of the ddvA::Tn5lac probe in strains with C7 or C8 was also significantly lower than in those with C5 or C6 (<20% for C7 and ≈35% for C8, relative to wild type; Fig. 5D). We checked whether lack of complementation by C7 and C8 could be the result of their instability or deficient expression (despite optimizing codon usage in M. xanthus) of the corresponding cell extracts in Western blots using a monoclonal antibody that specifically recognizes the CarD region present in all chimeras. C8, but not C7, could be detected at levels comparable to functional chimeras (Fig. S4C). The apparent lack of C7 could therefore account for its inability to function in M. xanthus. On the other hand, CytC could not functionally replace the CarD AT-hook region despite stable expression of the corresponding chimera. The basic regions in CarD and its chimeras although comparable in size (90–105 residues) vary in net positive charge besides amino acid composition (Fig. S1D). That different HMGA and H1-like domains, but not CytC, can be interchanged with no loss of CarD function in CarD suggests that net charge alone may not be sufficient to determine function. Overall, our data indicate that although CarD can tolerate a remarkable plasticity in the nature of the AT-hook/H1 region and its arrangement relative to the acidic region, the AT-hook region cannot be replaced by just any basic domain.

Discussion

The unique 2-domain architecture of CarD, consisting of an N-terminal domain of exclusively bacterial origin and a C-terminal domain similar to eukaryotic HMGA, led us to propose that it may have evolved by lateral gene transfer (LGT) of the HMGA domain followed by its fusion to a preexisting N-terminal bacterial module (13). That a CarD analog with a C-terminal domain akin to H1 CTR exists in A. dehalogenans, another myxobacterium that is most closely related to the suborder that includes M. xanthus and S. aurantiaca (30), is intriguing both from evolutionary and functional standpoints. It has been argued that the evolutionary origin of H1 CTR can be traced to eubacteria and that eukaryotes subsequently acquired it by LGT, with the globular and N-terminal domains evolving much later (2). CarD_Ad could then have evolved by the fusion of 2 preexisting bacterial modules. Hence, given the relatedness between A. dehalogenans and M. xanthus, and that both CarD proteins have a K/A/P-rich C-terminal region, the most parsimonious explanation for how M. xanthus CarD acquired the AT-hooks would be convergent evolution from an ancestral H1-like domain (as in CarD_Ad), rather than LGT. This rapid evolution toward AT-hooks should have occurred before M. xanthus and S. aurantiaca branched out, as both species possess an HMGA-like domain. Less probable would be the alternative that, in 2 closely related bacteria, 2 independent fusion events involving the N-terminal CarD module occurred, one to an H1-like domain (as in CarD_Ad) and the other to an HMGA-like domain (as in CarD). Interestingly, a naturally existing domain swap between AT-hooks and H1 CTR occurs in plants, where a globular domain like the one linked to H1 CTR is also found associated with the AT-hook region of plant HMGA (31).

A bacterial origin for H1 CTR was based on similar domains existing in various bacterial species but not in archaea (2). These bacterial H1 CTR-like domains have been implicated in DNA compaction, gene regulation, and other DNA transactions but molecular details on their action or interaction with other proteins are scarce. They all differ from CarD_Ad in lacking the characteristic N-terminal domain similar to CarDNter. Accordingly, the corresponding bacteria also lack a CarG analog. Specific processes in A. dehalogenans subject to CarD_Ad–CarG_Ad control remain an open question, as this bacterium lacks genes for fruiting body development or carotenogenesis that are CarD–CarG dependent in M. xanthus (30). However, our study indicates that CarD–CarG pairs in M. xanthus and A. dehalogenans share the same molecular architecture and interactions and thus, very likely, employ conserved molecular mechanisms of action.

The role of CTR in histone H1 function has long been an enigma given its high sequence diversity and intrinsically disordered structure. It was supposed that this very basic region functions simply as a polycation that condenses DNA (5, 32). The winged-helix globular domain of H1, which is its most highly conserved part, has been more amenable for structure–function studies (33–36). Recently, however, CTR is emerging as a critical determinant of H1 function, and specific subdomains and length differences in CTR have been invoked to explain the functional heterogeneity observed among its subtypes (5, 27, 32). The importance of CTR is underscored by the existence of single-domain H1 proteins in protists, which lack a globular domain and have compositions very similar to CTR, as is exemplified by the Tetrahymena thermophila macronuclear linker histone (2). Interestingly, although this H1 is indeed involved in chromatin condensation that could lead to global transcriptional repression, it was shown to activate or repress specific genes, thus linking H1 CTR to specific control of gene expression (37).

A key finding of our present study is that H1 CTR can substitute the HMGA AT-hooks in CarD with no loss of function in vivo. Thus, we find that CarD_Ad can replace CarD in M. xanthus so long as CarG_Ad, with which it interacts specifically, is also present. In line with this, chimeras with the CarD AT-hooks replaced by the CarD_Ad H1-like domain or even human histone H1.2 CTR, but which retain CarDNter and thus specific interactions with CarG, are functional in vivo. We argued elsewhere that the essential role of CarG, which does not bind DNA directly, maybe as an adaptor that bridges CarD to the transcriptional machinery and/or other factors that remain to be identified (22). Such interactions involving CarG could conceivably override differences in affinity between CarD and CarD_Ad, as both cause similar expression of 3 distinct reporter fusions, even though CarD_Ad binds DNA with a lower affinity in vitro, at least to the well-defined site at the light-inducible P_QRS promoter. A CarD-binding site like the one at P_QRS is not discernible at other promoters that are also regulated by CarD and CarG. Thus, a rather broad DNA-binding specificity may be tolerated by CarD, which would be compatible with its ability to accommodate HMGA to H1 domain swaps with no apparent loss of function. Our findings on the functional equivalence of these 2 domains in a bacterial transcriptional factor are echoed in the case of the human Epstein-Barr virus protein EBNA1, whose first 378 residues could be replaced by HMGA1a (residues 1 to 90) or H1.2 to support replication and maintenance of oriP-containing plasmids in metazoan cells (38).

Both in H1 CTR and in the analogous CarD_Ad domain, ≈76% of the residues are K, A, or P (Fig. S1D). However, whereas H1 CTR has several S/TPXK motifs (X is any amino acid), which are subject to phosphorylation and occur within subdomains implicated in DNA binding and chromatin condensation (1, 5, 32), these are absent from the CarD_Ad H1-like domain (or the CarD AT-hook region). Given that different chimeras function essentially as well as CarD in vivo despite the diversity in amino acid compositions (Fig. S1D) and primary sequences, defined functional subdomains in CarD CTR, if they do exist, may not be readily discernible.

It is remarkable that 2 apparently unrelated DNA-binding domains like the AT-hooks and CTR are functionally interchangeable. Both have regions of high content in basic amino acids and their intrinsically disordered structure has been hypothesized as providing the conformational malleability required in the biological need to interact with many different targets (4, 10, 21, 32, 39–43). Also, H1 and HMGA both bind preferentially to DNA of scaffold-associated regions or SARs (proposed to be implicated in delimiting DNA loops and in chromosome dynamics) (25). The preferential binding of H1 has been mapped to its CTR domain, which appears to recognize the narrow minor groove of AT-rich tracts in SARs, as had been shown for the HMGA AT-hooks (8, 9, 25, 27). Thus, the common features of a basic, intrinsically disordered region, coupled to their similar DNA-binding modes, could explain the functional interchangeability that we observe between these 2 DNA-binding domains. A requirement for structural disorder might account for why CytC, with a defined structure, could not replace the AT-hooks/H1 CTR in CarD function, despite its basic nature. In sum, our findings indicate that an AT-hook region or H1 CTR, but not just any basic domain, can supply the DNA-binding activity for CarD function.

In eukaryotes, where they are implicated in multiple cellular functions, histone H1 and HMGA proteins co-exist and can be present as a variety of isoforms. Pinning down the specific functions and understanding the molecular basis for their modes of action remain among the most perplexing questions in chromosome biology. There is significant evidence, however, that HMGA and H1 are functionally distinct in eukaryotes, where DNA is packed into nucleosomes. The functional equivalence of HMGA AT-hooks and H1 CTR that we find in bacterial transcriptional regulation may, nevertheless, provide insights that could be relevant in understanding the molecular basis underlying the modes of action of HMGA AT-hooks and H1 CTR.

Materials and Methods

Strains, Plasmids, and Growth Conditions.

Table S1 lists the plasmids and Escherichia coli/M. xanthus strains used in this study. Vegetative growth was carried out in rich CTT medium at 33 °C, and fruiting body development was induced on TPM agar and examined with a Zeiss dissecting microscope as described before (22). E. coli DH5α (for plasmid constructions; SI Materials and Methods) and BL21-(DE3) (for protein overexpression) were grown in Luria broth at 37 °C.

Complementation Analyses.

Complementation analyses were performed using the ΔcarD and ΔcarD ΔcarG strains as recipients for electroporation with a given plasmid construct bearing carD_Ad or the gene encoding a specific CarD chimera, or carD_Ad and carG_Ad. These plasmids do not replicate in M. xanthus but integrate into the chromosome by homologous recombination. From the resulting merodiploids haploid strains bearing the desired allele were generated and verified as detailed in SI Materials and Methods.

Protein Purification, Protein–Protein, Protein–DNA Interactions, and Western Blots.

Procedures for protein purification and analysis are described in SI Materials and Methods.