A Cyanobacterial Component Required for Pilus Biogenesis Affects the Exoproteome

Cyanobacteria, environmentally prevalent photosynthetic prokaryotes, contribute ∼25% of global primary production. Cyanobacterial biofilms elicit biofouling, thus leading to substantial economic losses; however, these microbial assemblages can also be beneficial, e.g., in wastewater purification processes and for biofuel production.

along with the observation that the inactivation of Synpcc7942_2071 abrogates pilus formation, we rename this gene product PilB and the corresponding mutant pilB::Tn5. Here, we provide further evidence suggesting the involvement of the T4P machinery of S. elongatus in protein secretion, as previously shown for several heterotrophic bacteria (35), and identify a component of the complex that is conserved specifically in the cyanobacterial clade.
Robust biofilm development is observed when the pilB::Tn5 mutant is grown in fresh medium; however, when inoculated into CM from a WT culture, the mutant grows planktonically. CM supplemented with nutrients still suppresses biofilm formation, negating the possibility that nutrient starvation prevented biofilm formation. Furthermore, suppressive fractions identified from the CM implicate a small (,3 kDa) heat-resistant substance as the inhibitor (30). The inhibitory effect of WT CM on biofilm development by the pilB::Tn5 mutant is in accordance with a model that assigns a role for the S. elongatus T4P machinery in the deposition of a biofilm inhibitor to the extracellular milieu.
Components that promote biofilm development in S. elongatus include four small proteins, EbfG1 (enable biofilm formation with a GG motif 1) to EbfG4, which are secreted by a mechanism employing PteB (32) (peptidase transporter essential for biofilm) and the microcin-processing peptidase-like protein EbfE (36). The extracellular biofilm inhibitor represses the transcription of the genes ebfG1-4 and pteB, thereby prohibiting biofilm development (31,32).
Here, we describe additional components that are essential for the biofilm self-suppression mechanism in S. elongatus. A genetic approach uncovered a highly conserved cyanobacterial protein (EbsA [essential for biofilm suppression A]) that does not exhibit sequence similarity to proteins of known function. Structurally, EbsA belongs to the a1b class and has a roll architecture; however, comparisons with other protein structures did not provide significant insights into its role in biofilm suppression (37). We revealed a tripartite complex of EbsA, PilB, and the Hfq RNA chaperone. Mutants defective for any one of these components develop biofilms and are impaired in protein secretion, pilus assembly, and competence for DNA uptake. Together, the data from this study reveal previously unrecognized components of pilus assembly, and possibly of a secretion system, thereby suggesting a novel mode of operation of these cyanobacterial complexes.

RESULTS
Synpcc7942_0862 encodes a highly conserved cyanobacterial protein required for biofilm self-suppression. The biofilm-forming pilB::Tn5 mutant is unpiliated and characterized by a high rate of sedimentation compared to the WT (33). This correlation between fast cell sedimentation and biofilm development in the pilB::Tn5 mutant encouraged us to test additional rapidly sedimenting mutants for biofilm development. Four of the 68 transposon insertion mutants that were classified as fast sedimenting in a genetic screen of an S. elongatus mutant library (38) exhibited biofilm formation (33). Three of these mutants bear an inactivation of genes related to T2S or T4P systems (pilA2, pilC, and pilN) (33). The fourth mutant is inactivated in Synpcc7942_0862 (Fig. 1A), dubbed ebsA (essential for biofilm suppression A) here. Homologs of EbsA that share 44 to 52% identity and 68 to 73% similarity (see Fig. S1 in the supplemental material) are widespread among diverse cyanobacteria but are not found outside this clade. In contrast to the WT strain that grows planktonically ( Fig. 1B and Fig. S2), the Synpcc7942_0862 mutant (ebsA::Tn5) formed robust biofilms, as shown by adhesion to the growth vessel (Fig. 1B) and by confocal fluorescence microscopy ( Fig. 1C and Fig. S2).
Additionally, measurement of the relative amount of chlorophyll in the planktonic fraction of the culture as a proxy for biofilm development (39) demonstrated that less than 10% is present in suspended cells of the mutant (Fig. 1D, ebsA::Tn5), indicating that the majority of the cells were in the sessile fraction. The introduction, in trans, of an intact copy of the ebsA gene preceded by an upstream sequence, presumed to include its promoter, restored planktonic growth to the ebsA::Tn5 mutant ( Fig. 1A and D and Fig. S2, ebsA::Tn5/ebsA), validating that biofilm formation is due to the inactivation of this gene. The inactivation of ebsA did not affect growth as measured by the increase in cell density at the planktonic stage of the cultures (Fig. S3A) and by total chlorophyll following biofilm formation (Fig. S3B). Thus, it is unlikely that biofilm development by this mutant is the consequence of a stress response elicited by the inactivation of ebsA. Together, the data support the involvement of ebsA in the biofilm selfsuppression mechanism that operates in S. elongatus.
Previously, we demonstrated that the double mutant in which both pilB and Synpcc7942_1134 (encoding EbfG4) are insertionally inactivated grows planktonically, in contrast to the biofilm-forming pilB mutant (30). Likewise, the ebsA::Tn5 ebfG4X double mutant grew planktonically, in contrast to the biofilm-forming phenotype of the ebsA::Tn5 mutant (Fig. 1D). These observations suggest that both the pilB::Tn5 and ebsA::Tn5 mutants employ a mechanism of biofilm development that requires EbfG4.
Coimmunoprecipitation reveals a tripartite complex of EbsA, PilB, and the RNA chaperone Hfq. To gain insight into EbsA function, we aimed at the identification of its interactors by coimmunoprecipitation (co-IP). To this end, the protein was tagged at All indicated genes are annotated as "conserved hypothetical proteins." Unigene Set (UGS) vector 22D7 was used to insert Tn5 encoding a kanamycin resistance cassette (Km R ) to generate the ebsA:: Tn5 gene inactivation mutant. Arrows indicate PCR primers used to amplify a DNA fragment that was introduced in trans into ebsA::Tn5 to generate the ebsA::Tn5/ebsA strain to test complementation (see Table S1 in the supplemental material for details). (B) Cultures of WT, ebsA::Tn5, and ebsA::Tn5/ebsA strains. The arrowhead indicates the direction of light (biofilms are formed away from the light source). (C) Image of ebsA::Tn5 biofilm using confocal fluorescence microscopy. Imaging is based on autofluorescence (excitation at 630 nm and emission at 641 to 657 nm). The color scale represents biofilm depth. Different regions of the biofilm differ in depth (Fig. S2). (D) Percentage of chlorophyll in suspension as a proxy for biofilm formation. Strains analyzed were the WT, the ebsA::Tn5 mutant, the ebsA::Tn5/ebsA strain, and the double mutant in which ebsA and gene Synpcc7942_1134 are inactivated (ebsA::Tn5 ebfG4). Data represent averages and standard deviations (SD) from 3 independent experiments. The asterisk indicates significance (P , 5.7E230 by a two-tailed t test).
its C terminus with a triple-FLAG epitope tag (here, EbsA::FLAG). The expression of the gene that encodes this tagged protein under the control of its native regulatory region complemented the biofilm-forming phenotype of the ebsA-inactivated strain (Fig. S4A, ebsA::Tn5/ebsA::flag). Cell extracts from the ebsA::Tn5/ebsA::flag and ebsA::Tn5/ebsA strains were used for co-IP experiments followed by MS analysis. Protein enrichment was calculated as the amount of protein in the "bait strain" (ebsA::Tn5/ebsA::flag) relative to its cognate control (ebsA::Tn5/ebsA). These experiments indicated high enrichment of PilB ( Fig. 2A), the ATPase homolog required for the biofilm self-inhibitory mechanism. This observation, which indicates a physical interaction between EbsA and PilB, is consistent with a biofilm-inhibitory role for EbsA, as evidenced by ebsA::Tn5 biofilm development.
To validate the interactions between EbsA and PilB, the latter was FLAG tagged (Table S1), and the tagged protein, PilB::FLAG, was demonstrated to complement the pilB mutant (Fig. S4B, pilB::Tn5/pilB::flag). Co-IP experiments using PilB::FLAG indicated coprecipitation of EbsA, further confirming a PilB-EbsA interaction (Fig. 2). We also noted that both EbsA and PilB interact with the protein encoded by Synpcc7942_2070, a homolog of the PilT ATPase of T4P systems (Data Set S1). These interactions suggest a role for EbsA in pilus assembly.
Intriguingly, a homolog of the RNA chaperone Hfq (encoded by Synpcc7942_1926) was also highly enriched by co-IP with both EbsA::FLAG and PilB::FLAG baits (Fig. 2). Hfq is a central bacterial regulator that acts at the posttranscriptional level by mediating mRNA-small RNA interactions (40)(41)(42)(43). These co-IP data suggest a tripartite complex of EbsA, PilB, and Hfq. Indeed, EbsA and PilB were two of only seven highly enriched proteins when Hfq::FLAG was used as a bait for co-IP ( Fig. 2A and Data Set S1), further substantiating the presence of the tripartite EbsA-PilB-Hfq complex.
Studies of Synechocystis demonstrated that Hfq interacts with PilB (44), suggesting that the association of this RNA chaperone with T4P machinery is a more general cyanobacterial trait. RNA substrates of cyanobacterial Hfq proteins, however, have not been identified (44).
Co-IP data revealed 39 proteins interacting with EbsA (fold change of .6; P value of ,0.05) (Data Set S1), of which 33 did not interact with PilB or Hfq (Fig. 2B). Analysis for motifs using Multiple Em for Motif Elicitation (MEME) (45) did not reveal protein motifs that are significantly enriched in EbsA interactors. Analysis by STRING, however, indicated that molecular functions related to magnesium ion binding and protein domains related to different ATPase families are enriched among EbsA interactors (Data Set S2). Novel Protein Links T4P System and Exoproteome ® Hfq is essential for biofilm self-suppression. The presence of the Hfq RNA chaperone in a complex with two components, PilB and EbsA, that are both required for biofilm self-suppression in S. elongatus suggests that it is also involved in this mechanism. Indeed, the hfq knockout mutant (hfqX) resulting from the insertion of an omega cassette into Synpcc7942_1926 (Fig. 3A) develops robust biofilms, apparent as cells adhering to the growth tube (Fig. 3B). Additionally, fluorescence microscopy ( Fig. 3C and Fig. S2) and measurements of relative chlorophyll in suspended cells (Fig. 3D, hfqX) indicated biofilm formation by the hfqX mutant. Furthermore, the insertion of Synpcc7942_1926 in trans into the hfqX mutant abolished biofilm development, confirming that Hfq is required for the biofilm self-inhibition process ( Fig. 3D and Fig. S2; see also Fig. S4C for phenotypic complementation with FLAG-tagged Hfq). Also, similar to the pilB::Tn5 ebfG4X and ebsA::Tn5 ebfG4X mutants, the hfqX ebfG4X double mutant grew planktonically (Fig. 3D). Thus, the three mutants share a mechanism of biofilm development in which EbfG4 takes part.
Impairment of EbsA or Hfq abrogates pilus formation. The pilB mutant lacks cell pili ( Fig. 4) (33), indicating that the ATPase homolog impaired in this strain has a role in pilus assembly. Given the physical association between PilB, EbsA, and Hfq ( Fig. 2), we tested whether the latter two proteins take part in pilus assembly. Transmission electron microscopy (TEM) indicated that similar to the pilB::Tn5 strain (33), the ebsA::Tn5 and hfqX strains are completely devoid of cell pili, in contrast to the piliated WT strain (Fig. 4), thus demonstrating the requirement of EbsA and Hfq for pilus assembly. Moreover, this observation further supports the involvement of these proteins in the same cellular machinery as PilB, in agreement with the physical interaction among the proteins (Fig. 2). The involvement of Hfq in pilus assembly and its association with PilB were previously reported for the cyanobacterium Synechocystis sp. strain PCC 6803 (44,46).
T4P systems are commonly involved in DNA transfer (47)(48)(49)(50)(51); accordingly, we noted previously that the pilB::Tn5 mutant lost the natural ability of the WT strain to take up DNA (30). Therefore, we tested the transformability of the newly isolated mutants. The pilB::Tn5 and ebsA::Tn5 strains are absolutely nontransformable, whereas the transformation efficiency of the hfqX strain is about 100-fold lower than that of the WT (Fig. S5), in accordance with a recent publication (52).
Impairment of EbsA or Hfq affects the exoproteome. Components of the T4P assembly apparatus share homology with those of T2S systems (1,7,(53)(54)(55). Furthermore, in particular cases, T4P machineries have been demonstrated to be involved in the secretion of proteins to the extracellular milieu (35). The inactivation of pilB affects the levels of extracellular proteins (here, the exoproteome) (33). Therefore, the association of EbsA and Hfq with PilB encouraged us to examine the exoproteomes of the ebsA::Tn5 and hfqX strains. CM from planktonic (2-day-old) cultures of the WT, pilB::Tn5, ebsA::Tn5, and hfqX strains were concentrated by lyophilization and analyzed by electrophoresis and silver staining. This analysis indicated that several proteins are present in substantially reduced amounts or are absent from the exoproteome of the pilB::Tn5 mutant compared to the WT (Fig. 5A, left), in agreement with previous studies (30,33). Furthermore, lower levels of some proteins were observed in extracellular fluids from ebsA::Tn5 and hfqX cultures than in pilB::Tn5 cultures (Fig. 5A, left). The most prominent band in WT CM (;13 kDa) (Fig. 5A, left, arrowhead), which corresponds to the pilin subunit PilA1, represents detached pili (33).
Exoproteins are released to the extracellular milieu by regulated secretion processes; however, cell lysis may also contribute to the exoproteome. To test whether differences in cell lysis between the strains contribute to the observed changes in extracellular proteins, 2-day-old cultures (the growth stage at which the exoproteome was analyzed) were assessed by live/dead staining. SYTOX green nucleic acid stain enters cells with compromised membranes; accordingly, heat treatment of WT cells resulted in ;96% positive cells (Fig. 5B, control). In cultures of the pilB::Tn5, ebsA::Tn5, and hfqX mutants, however, staining revealed only a minute fraction of positive cells (,1%), and differences among the strains were not significant (Fig. 5B). Thus, differential cell lysis does not contribute to the observed changes among the exoproteomes.
Additionally, quantitative analyses of extracellular fluids were performed using MS. The ratio of the abundance of a protein in CM of the WT versus that of a given mutant was used to identify proteins that are less abundant in the mutant exoproteome ( Fig. 5C and Data Set S3) (fold change of .2; false discovery rate [FDR] of ,0.1). As commonly done, samples were normalized based on the protein amount; namely, equal amounts of protein from each strain were digested by trypsin and subsequently analyzed by MS. Because the prominent ;13 kDa protein (PilA) is missing from extracellular fluids of the mutants (Fig. 5A), equal total protein samples result in an Novel Protein Links T4P System and Exoproteome ® exaggerated representation of the non-PilA protein complement in the exoproteome of the mutants. Thus, the normalization step underestimates the quantitative difference of proteins that are less abundant in the mutant exoproteomes. Nevertheless, MS analyses identified numerous exoproteins that are less abundant in the mutants, as specified below.
The inactivation of pilB decreased the complexity of the exoproteome (Fig. 5C), as previously reported (33); furthermore, many proteins are less abundant in the exoproteome of the ebsA::Tn5 or hfqX mutant than in the WT (Fig. 5C). Of the 98 proteins that are less abundant in the pilB::Tn5 exoproteome than in the WT, 47 (48%) are shared among the three mutants (Fig. 5C). This overlap is consistent with our suggestion that PilB, EbsA, and Hfq take part in the same protein secretion machinery, in agreement with co-IP data that indicate physical interactions among these proteins (Fig. 2). Numerous proteins, however, are less abundant in the ebsA::Tn5 or hfqX mutant but not in the pilB::Tn5 mutant (Fig. 5C), suggesting an additional effect of the inactivation of either ebsA or hfq on the exoproteome via a process that does not involve PilB.
Analysis of cell extracts was performed to examine whether the inactivation of either pilB, ebsA, or hfq substantially affects the overall amount of cellular proteins. This analysis revealed indistinguishable protein profiles with the exception of an ;13 kDa protein that is much more intense in the WT than in any of the three mutants (Fig. 5A, right) and is likely to be PilA, in accordance with the absence of cell pili from the pilB:: Tn5, ebsA::Tn5, and hfqX mutants (Fig. 4). Of note, 14 ribosomal subunit proteins were detected at low levels in the exoproteomes of the mutants, 4 of which are less abundant in the exoproteomes of all three mutants (Data Set S3, sheet "Less abundant-Ribosomal"). Ribosomal subunits are ubiquitous cellular components, and thus, drastic changes in their intracellular amounts should have been apparent by gel electrophoresis. For example, RplJ (175 amino acids [aa]) is ;28-, ;12-, and ;7-fold less abundant in the ebsA::Tn5, pilB::Tn5, and hfqX mutants, respectively, than in the WT. The intensities of the bands in the corresponding regions of the total protein gel, however, are similar among the strains (Fig. 5A, right). Additionally, the levels of RpmE (77 aa) are ;137and ;213-fold lower in the exoproteomes of the ebsA::Tn5 and hfqX mutants, respectively; however, such changes are not observed in cellular extracts (Fig. 5A,  right). Thus, the data suggest that the identified changes in the extracellular levels of ribosomal proteins do not merely represent variations in their cellular levels. Moreover, if changes in extracellular levels of ribosomal proteins are due to lower intracellular abundance, one would expect a substantial impact on protein synthesis and, consequently, reduced growth of the mutants, which was not observed (Fig. S3). Together, these data support the suggestion that at least some of the changes detected in the exoproteomes represent the involvement of EbsA and Hfq in secretion processes.
Analysis using MEME (45) revealed the enrichment of particular motifs in proteins that are less abundant specifically in the exoproteomes of the ebsA::Tn5 and hfqX mutants (Table S2). A motif that includes two conserved cysteine residues (C-x-x-C), typical of thiol:disulfide oxidoreductases involved in redox reactions, is enriched (Table S2, (Table S2), is unknown. Enrichment analysis of less abundant proteins using STRING indicated the presence of signal secretion sequences in proteins less abundant in the ebsA::Tn5 and hfqX exoproteomes than in the WT exoproteome (Data Set S2). This analysis did not reveal signal secretion sequences in proteins that were less abundant only in pilB::Tn5 samples, in agreement with a previous comparative analysis of pilB::Tn5 and WT exoproteomes (33).
In addition to the suggested abrogation of protein secretion, the inactivation of pilB resulted in the augmentation of particular proteins in the mutant's exoproteome compared to the WT ( Fig. 5D and Data Set S3), in agreement with our previous report (33). The "overabundant" proteins include some that enable biofilm development and carry a GG secretion motif (EbfGs). The eight proteins that are overabundant in the exoproteomes of all three mutants (Fig. 5D) include EbfG1, EbfG2, and EbfG4 (Data Set S3, sheet "More abundant in Mutants"). Note that on average, EbfG4 is 12-fold more abundant in hfqX than in WT samples (Data Set S3, sheet "Raw data"); however, due to the high variability between independent experiments, this difference is insignificant. The presence of EbfG proteins in the exoproteome correlates with biofilm development by the pilB::Tn5 mutant (32); thus, the overabundance of EbfG proteins in the exoproteomes of the ebsA::Tn5 and hfqX mutants is consistent with biofilm development by these mutants. Furthermore, the planktonic phenotype observed by the inactivation of ebfG4 in the pilB::Tn5 mutant (30) as well as in the ebsA::Tn5 (Fig. 1D) and hfqX (Fig. 3D) mutants further supports a biofilm development process that involves the biofilm-enabling protein EbfG4 and that is common to the three mutant strains.
Inactivation of the Synechocystis homolog of EbsA impairs protein secretion and motility. Homologs of EbsA are present in diverse cyanobacteria (Fig. S1). To examine whether the Synechocystis homolog serves similar functions, we inactivated the gene slr0038, encoding a protein that is 50% identical and 72% similar to EbsA. We suggest that the resulting mutant, slr0038::Tn5, is impaired in protein secretion, as revealed by protein electrophoresis followed by silver staining of concentrated CM (Fig. 6A).
Synechocystis, unlike the laboratory strain of S. elongatus, is motile on solid growth medium, and it has been shown that the abrogation of pilus formation impairs this motility (44,46,50,51,56). The lack of motility of the slr0038 mutant (Fig. 6B) supports the suggestion that Slr0038 is required for pilus formation. Taken together, data from the analysis of extracellular proteins and motility support a role for Slr0038 in protein secretion and pilus assembly in Synechocystis, analogous to the role of EbsA in S. elongatus. The inactivation of slr0038, however, did not result in biofilm development, suggesting differences in the regulation of biofilm formation between these two cyanobacteria.
Bioinformatics analysis suggests a single complex for secretion and T4P functions. Studies of heterotrophic bacteria revealed the presence of distinct complexes for T2S or T4P biogenesis, the subunits of which are homologous (1,7,(53)(54)(55). Recent analyses, however, indicate that distinct T2S complexes are largely confined to Proteobacteria (7,34). Based on the observation that the inactivation of solely pilB, ebsA, or hfq affects the exoproteome (Fig. 5) as well as pilus assembly (Fig. 4), we propose that S. elongatus possesses a single complex that serves both protein secretion and pilus biogenesis, as was demonstrated for some cases of heterotrophic bacteria (35). Bioinformatic analysis of the S. elongatus genome, which revealed only a single gene for each component of a putative T2S/T4P system, supports the existence of a single complex, for example, Synpcc7942_2450 (annotated gspD [general secretion protein D]), Synpcc7942_2451 (annotated pilO), Synpcc7942_2452 (annotated pilN), Synpcc7942_2453 (annotated pilM), and Synpcc7942_2069 (annotated pilC). The absence of a second set of genes encoding putative components of an independent T2S, together with data indicating that the inactivation of a single gene impairs pilus assembly and affects the exoproteome, led us to propose that a single complex serves overlapping T2S and T4P functions in S. elongatus. Likewise, bioinformatics analysis and the phenotype obtained from the inactivation of slr0038, encoding an EbsA homolog in Synechocystis, support the existence of such a shared complex in other cyanobacteria.

DISCUSSION
Several lines of evidence support the suggestion that two newly identified components, EbsA and Hfq, act along with PilB to affect biofilm formation via a common mechanism: (i) genetic evidence demonstrating the requirement of ebfG4 for biofilm formation by the pilB::Tn5, ebsA::Tn5, and hfqX mutants ( Fig. 1D and Fig. 3D); (ii) analysis of the exoproteome demonstrating high levels of biofilm-enabling proteins, including EbfG4, in CM from mutants (see Data Set S3 in the supplemental material); and (iii) the physical interaction among PilB, EbsA, and Hfq (Fig. 2). Nevertheless, the possibility that EbsA and Hfq also exert biofilm regulation independently of PilB should not be excluded.
A tripartite complex, PilB-EbsA-Hfq, of a shared pilus assembly and protein secretion machinery. Based on this study we suggest that the tripartite complex PilB-EbsA-Hfq is associated with machinery involved in several cellular functions: pilus assembly (Fig. 4), DNA uptake (Fig. S5), and protein secretion (Fig. 5). Although it is well established that numerous bacteria possess distinct complexes for these functions (1,7,(53)(54)(55), the T4P complexes in some bacteria are known to serve for the secretion of proteins other than the pilus subunits (35). For example, in Vibrio cholerae (57) and Dichelobacter nodosus (58), T4P systems are used for the secretion of a soluble colonization factor and a protease, respectively. Additionally, T4P-mediated secretion was demonstrated in Escherichia coli (59) and Francisella tularensis (60). The physical association of EbsA and Hfq with PilB, and the shared phenotypes of their mutants, strongly supports the suggestion that S. elongatus employs the same complex for pilus biogenesis and protein secretion (Fig. 7).
EbsA, a previously unknown protein that may be a component of T4P/secretion systems, is widespread in diverse cyanobacteria, and the inactivation of the Synechocystis homolog, slr0038, impairs protein secretion and, based on the lack of motility on agar, Novel Protein Links T4P System and Exoproteome ® pilus function (Fig. 6). Similar to the ebsA::Tn5 strain, the inactivation of hfq in Synechocystis blocks pilus formation and motility on agar plates (44). The involvement of Synechocystis Hfq in pilus formation, in analogy to its role in S. elongatus, together with the phenotype of the slr0038::Tn5 strain, further substantiates the presence of a mechanism that is common to divergent cyanobacteria.
Exoproteome analyses imply pleiotropic effects of the inactivation of ebsA or hfq. About half of the proteins that are less abundant in the pilB::Tn5 mutant (47 of 98) are also less abundant in the ebsA::Tn5 and hfqX mutants. However, a larger number of proteins are less abundant than in the WT for the ebsA::Tn5 and hfqX mutants compared to the pilB::Tn5 mutant. Furthermore, a data set of less abundant ebsA::Tn5 proteins is almost entirely encompassed by that group from the hfqX mutant. We propose that EbsA and Hfq also act in a secretion process that is independent of PilB, in agreement with enrichment analysis using STRING that showed secretion signal sequences in proteins that are less abundant in the ebsA::Tn5 and hfqX exoproteomes than in the WT exoproteome, which was not true for the less abundant proteins of the pilB mutant. The 79 proteins that are less abundant specifically in the hfqX mutant (Fig. 5C) may represent an indirect effect of the Hfq RNA chaperone on protein expression or secretion. Numerous studies of Hfq in heterotrophic bacteria revealed its involvement in the regulation of translation (40)(41)(42)(43). A role for cyanobacterial Hfq as an RNA chaperone has not been demonstrated yet; however, it is possible that it exerts translation regulation and that, consequently, the less abundant proteins unique to the hfqX mutant represent the outcome of the impact of Hfq on translation. Less abundant exoproteins of the ebsA::Tn5 mutant, which are not shared with the pilB::Tn5 mutant, may also represent a pleiotropic effect, although the particular mechanism in this case is enigmatic.
Some of the proteins that are underabundant in the mutants' exoproteomes are cellular proteins, e.g., ribosomal subunits and a nucleoid protein (Data Set S3). Intracellular proteins that serve another function when present extracellularly, socalled "moonlighting proteins," have been described in diverse organisms, including bacteria (61,62). Such protein multitasking, however, has not been documented in cyanobacteria, and the reason for the release of S. elongatus cellular proteins is unknown. However, our data argue against cell lysis or general permeability changes as an explanation for these factors outside the cell.
In summary, this study is consistent with a hypothesis assigning a secretion role to the complex that assembles pili in cyanobacteria, as has been shown for a few species of Proteobacteria, and also identified distinct features of the cyanobacterial complex. Cyanobacterial cells possess an elaborate internal network of photosynthetic membranes in addition to their inner and outer cell membranes. Such morphology imposes further challenges for protein sorting and delivery, compared to heterotrophic bacteria. Conceivably, cyanobacteria evolved to accommodate these challenges by recruiting EbsA and Hfq to the T4P/secretion machinery.

MATERIALS AND METHODS
Strains, culture conditions, biofilm quantification, and assessment of DNA competency. The growth of all Synechococcus elongatus PCC 7942 and Synechocystis PCC 6803 strains was performed in BG-11 medium as described previously (39). Quantification of biofilms based on chlorophyll measurements was performed as described previously (39). Briefly, suspended cells were carefully removed, and chlorophyll from the biofilm cells remaining in the growth tube was extracted with acetone. Additionally, acetone extraction was used to quantify chlorophyll in the suspended fraction. The percentage of chlorophyll in suspended cells from the total culture chlorophyll served for quantitative assessment of biofilms. Assessment of DNA competence was performed essentially as described previously (63). Exponentially growing cells were centrifuged (5,000 Â g for 8 min at room temperature), washed once with 10 mM NaCl, and resuspended to an optical density at 750 nm (OD 750 ) of 4.0. A shuttle vector (1,000 ng) was added to 600 ml of cells, which were gently agitated overnight at 28°C in the dark. Transformants were selected by plating on selective solid growth medium (50 mg/ml spectinomycin) supplemented with NaHCO 3 (5 mM) and sodium thiosulfate (0.3%, wt/vol). Additionally, serial dilutions were plated onto nonselective medium. The transformation efficiency was calculated as the number of CFU per milliliter obtained on selective medium normalized to that observed on nonselective medium. The shuttle vector replicates autonomously, thus allowing the assessment of DNA uptake without a possible impact of the efficiency of DNA integration into the chromosome.
Microscopy. Examination of negatively stained cells by transmission electron microscopy was conducted at the Irving and Cherna Moskowitz Center for Nano and Bio-Nano Imaging at the Weizmann Institute of Science as described previously (33).
To observe biofilms by fluorescence microscopy, a sterile microscope slide was inserted into a growth tube upon culture inoculation, and biofilms formed on the glass tube wall as well as on the microscope slide. Following 7 days of growth, the microscope slide was removed with forceps and washed once by dipping into double-distilled water. Autofluorescence-based images were collected using a Leica SP8 confocal microscope (excitation at 630 nm and emission at 641 to 657 nm).
IP followed by MS. IP was performed essentially as described previously (44), with the following modifications. A culture (250 ml; OD 750 of 2.5 to 3) was concentrated by centrifugation to 2.2 ml, and a freshly prepared protease inhibitor cocktail (catalog number P8465-5ML; Sigma) was added to 3.44 mg/ ml. Aliquots (1 ml) of the concentrated cell culture were combined with ;1 g of glass beads (catalog number 11079101; Biospec) in 2-ml Eppendorf tubes. Cells were broken using a mixer mill (catalog number MM400; Retch) at a frequency of 30 s 21 for 2 min in prechilled holders (5 times, with 1 min of incubation on ice between the cycles). Cell lysates were centrifuged (relative centrifugal force [RCF] of 835 Â g for 5 min at 4°C) to pellet the glass beads. The cell lysate (1 ml) was transferred to a 1.5-ml Eppendorf tube containing 100 ml washed anti-FLAG magnetic beads (catalog number M8823; Sigma) and incubated for 2 h at room temperature and overnight at 4°C with mixing (RotoFlex; Argos Technologies). Beads were washed 4 to 5 times according to the manufacturer's instructions, and elution was performed with 1 ml of 100 mM triple-FLAG peptide (catalog number A6001; APExBIO).
MS analysis at the de Botton Institute for Protein Profiling at The Nancy and Stephen Grand Israel National Center for Personalized Medicine (Weizmann Institute of Science) was performed as previously described (32), except that trypsin digestion was not followed by chymotrypsin digestion and "discovery mode" was used rather than "targeted analysis." Raw data were processed with MaxQuant v1.6.0.16. The data were searched with the Andromeda search engine against the cyanobacterial proteome database appended with common laboratory protein contaminants. Quantification was based on the label-free quantitative (LFQ) method. Data were analyzed to identify proteins that were significantly enriched in a "bait strain" (triple-FLAG-tagged protein) compared to its nontagged negative control (P value of #0.05 by a t test; fold change of .6).
Analysis of extracellular fluids and total cell extracts. For collection of CM, cultures were centrifuged (5,000 Â g for 10 min) at room temperature, and the supernatant was removed and passed through a 0.22-mm filter. MS analysis was performed as described above. Data were analyzed to identify proteins that are significantly more or less abundant in a particular mutant's exoproteome than in the WT (FDR of #0.1), with a cutoff of at least a 2-fold change. Protein motifs were searched using the MEME package (45) (parameters -nmotifs 10, -minw 6, -maxw 50, and -mod anr). Significant relevant motifs were scanned against the S. elongatus proteome using Find Individual Motif Occurrences (FIMO) (45) to reveal motif occurrences in the total proteome. Enrichment was calculated for each significant motif by comparing its occurrence in a specific protein list with its FIMO background output data (twotailed P value of ,0.05 by Fisher's exact test).
Additionally, exoproteins as well as cell extracts were analyzed by using Tricine gel (64) followed by silver staining (SilverSNAP stain kit II, catalog number 24612; Pierce). For this analysis, CM collected as described above was lyophilized and resuspended in 1/200 of the original volume in a solution containing 10 mM Tris (pH 8.0) and 1 mM EDTA supplemented with 0.86 mg/ml of a freshly prepared protease inhibitor cocktail (catalog number P8465-5ML; Sigma). Cell extracts were prepared essentially as described above for the IP analysis; however, a 50-ml culture was concentrated to 0.5 ml, the protease inhibitor concentration was 0.86 mg/ml, and 0.2 g of glass beads was added for breakage by the mixer mill (catalog number MM400; Retch).
Live/dead staining. SYTOX green dead cell stain (Molecular Probes), a dye that enters cells with compromised membranes and binds to nucleic acids, was employed for viability assessment as described previously (65). For SYTOX staining, cells were diluted with phosphate saline buffer to an OD 750 of 0.0004. Following the addition of SYTOX (40 nM), cells were incubated in the dark for 15 min and analyzed by flow cytometry using Becton, Dickinson Fortessa (excitation at 488 nm and emission at 530 6 25 nm). Heat-treated (80°C for 1 min) WT cells served as positive controls. Fluorescence was plotted versus forward scattering (FSC).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.