Gene clusters encoding putative outer membrane electron conduits have specific roles during metal and electrode respiration in Geobacter sulfurreducens

At least five gene clusters in the Geobacter sulfurreducens genome encode putative outer membrane ‘electron conduits’, which are redox active complexes containing a periplasmic multiheme c-cytochrome, integral outer membrane β-barrel, and outer membrane redox lipoprotein. Single gene-cluster deletions and all possible multiple deletion mutant combinations were constructed and grown with graphite electrodes poised at +0.24 V and -0.1 V vs. SHE, Fe(III)- and Mn(IV)-oxides, and soluble Fe(III)-citrate. Different gene clusters were necessary for reduction of each electron acceptor. For example, only the ΔextABCD cluster mutant had a severe growth defect on graphite electrodes at all redox potentials, but this mutation did not affect Fe(III)-oxide, Mn(IV)-oxide, or Fe(III)-citrate reduction. During metal oxide reduction, deletion of the previously described omcBC cluster caused defects, but deletion of additional components in the ΔomcBC background, such as extEFG, was necessary to produce defects greater than 50% compared to wild type. Deletion of all five gene clusters was required to abolish all metal reduction. Mutants containing only one cluster were able to reduce particular terminal electron acceptors better than wild type, suggesting routes for improvement by targeting substrate-specific electron transfer pathways. Our results show G. sulfurreducens utilizes different membrane conduits depending on the extracellular acceptor used.

severe growth defect on graphite electrodes at all redox potentials, but this mutation did not 54 affect Fe(III)-oxide, Mn(IV)-oxide, or Fe(III)-citrate reduction. During metal oxide reduction, 55 deletion of the previously described omcBC cluster caused defects, but deletion of additional 56 components in the ∆omcBC background, such as extEFG, was necessary to produce defects 57 greater than 50% compared to wild type. Deletion of all five gene clusters was required to 58 abolish all metal reduction. Mutants containing only one cluster were able to reduce particular 59 terminal electron acceptors better than wild type, suggesting routes for improvement by 60 targeting substrate-specific electron transfer pathways. Our results show G. sulfurreducens 61 utilizes different membrane conduits depending on the extracellular acceptor used. 62

INTRODUCTION 64
one cluster grew as poorly as the Δ5 mutant, further indicating that under these conditions, 219 extEFG, extHIJKL, and omcBC did not contribute to electron transfer to electrodes (Fig. 2B). 220 221 A 5-conduit deletion mutant expressing extABCD has a faster growth rate on electrodes 222 than wild type. To further investigate the specific effect of extABCD on electrode growth, 223 extABCD was provided on a vector in the Δ5 strain. The 3-gene omcB conduit cluster (ombB-224 omaB-omcB) was also placed in the Δ5 strain using the same vector, and both were compared 225 to wild type cells containing the empty vector. While the plasmid is stable for multiple 226 generations, routine vector maintenance requires growth with kanamycin, and kanamycin carry-227 over into biofilm electrode experiments is reported to have deleterious effects on electrode 228 growth Chan et al., 2015). Thus, we re-examined growth of the empty vector 229 strain. When selective levels of kanamycin (200 μg·ml -1 ) were present in electrode reactors, 230 colonization slowed and final current production decreased 74%. At levels resulting from carry-231 over during passage of cells into the electrode reactor (5 μg·ml -1 ) growth rate was not affected, 232 but final current was decreased up to 30%, suggesting interference with biofilm thickness rather 233 than respiration (Fig. 3A). All subsequent complementation was performed in the presence of 5 234 μg·ml -1 residual kanamycin and compared to these controls. 235

236
Expressing the omcB conduit cluster in the ∆5 strain failed to increase growth with electrodes as 237 electron acceptors. These data were consistent with the lack of an effect seen in ∆omcBC 238 deletions, as well as the poor growth of omcBC + mutants containing both the OmcB and OmcC 239 clusters (Fig. 3B). However, when extABCD was expressed on the same vector in the ∆5 240 background, colonization was faster and cells reached a higher final current density compared 241 to wild type (421 ± 89 µA/cm 2 vs. 297 ± 11 µA/cm 2 , n=3) (Fig. 3B). This enhancement was 242 similar to the positive effect observed in the extABCD + strain, and further supported the 243 hypothesis that extABCD played a central role during electron transfer to electrodes (Fig. 2B). 244 245 Growth of intermediate two-conduit deletion mutants were unchanged from single-cluster strains 246 (Fig. S1). Just as the mutant lacking extABCD produced the same phenotype as the ∆5 strain 247 (Fig. 2), deletion of second clusters from the ∆extABCD strain produced similar results as 248 ∆extABCD, and no other two-cluster combination of omcBC, extEFG or extHIJKL mutants showed defects to suggest they were utilized or expressed during these electrode growth 250 conditions. 251 252 Cells lacking single gene clusters have partial reduction defects with Fe(III)-and Mn(IV)-253 oxides. In contrast to the dominant effect of extABCD on electrode respiration, no single cluster 254 deletion eliminated the majority of growth with particulate Fe(III)-or Mn(IV)-oxides. The most 255 severe defect was observed in the ∆omcBC cluster mutant, which reduced 68% of Fe(III)-oxide 256 compared to wild type (Fig. 4A). Minor defects were observed for ΔextEFG and ΔextHIJKL, 257 while ΔextACBD reduced Fe(III)-oxide near wild-type levels. None of the single mutants 258 displayed defects with Mn(IV)-oxides (Fig. 4C). These results suggested that multiple clusters 259 were active during metal oxide reduction. 260 261 Any one gene cluster is sufficient for partial Fe(III)-or Mn(IV)-oxide reduction, while 262 deletion of all 5 clusters eliminates electron transfer to these metal oxides. Unlike 263 electrode respiration, deletion of the full suite of clusters eliminated all residual electron transfer 264 to Fe(III)-and Mn(IV)-oxides ( Fig. 4B and D). When multiple-deletion strains containing only one 265 cluster were tested for Fe(III)-oxide reduction, results supported key roles for omcBC and 266 extEFG in metal oxide reduction, and little involvement by extABCD. For example, Fe(III)-oxide 267 reduction by omcBC + was nearly 80% of wild type, extEFG + was over 60%, but the extABCD + 268 strain reduced less than 30% of wild type. The omcBC + , extEFG + , and extHIJKL + strains 269 achieved about 80% of wild type Mn(IV)-reduction at 80 hours, but the extABCD + strain again 270 displayed poor growth with Mn(IV)-oxide. However, in contrast to Fe(III)-oxides, the ΔextABCD strain showed near wild-type reduction. 304 The ∆5 strain lacking all omcBC and ext clusters failed to reduce Fe(III)-citrate (Fig. 7B). Also 305 unlike Fe(III)-oxide reduction, strains with only omcBC + or extABCD + clusters had near wild-type 306 Fe(III)-citrate reduction rate, while extEFG + and extHIJKL + reduced Fe(III)-citrate to just 20% of 307 wild type. 308

309
Since the ∆omcBC ΔextEFG strain showed the largest defect in Fe(III)-oxide reduction, this 310 strain was analyzed with Fe(III)-citrate as well. However, this double cluster deletion mutant 311 showed little difference compared to the parent ∆omcBC strain (Fig. 7C). In contrast to Fe(III)-oxides, where deletion of extABCD had little effect, ∆omcBC ∆extABCD was the only conduit 313 deletion combination that severely affected growth with Fe(III)-citrate (Fig. 7C). Compared to 314 growth of extEFG + and extHIJKL + (Fig. 7B), the ΔomcBC ΔextABCD mutant (containing both 315 extEFG and extHIJKL) reduced Fe(III)-citrate to the same level (Fig. 7C). These data suggest 316 that when both extEFG and extHIJKL remained in the genome in the ΔomcBC ΔextABCD 317 mutant, their activity was not additive. Plasmids containing either ombB-omaB-omcB or 318 extABCD restored Fe(III)-citrate reduction in a Δ5 strain to levels within 90% of the respective 319 omcBC + and extABCD + strains (Fig. 7D).   conduit, adjusted to wild type performance. Many of the recently described ext gene clusters are 330 necessary for wild-type metal reduction, yet few are sufficient. For example, extEFG and 331 extHIJKL were necessary for Fe(III)-citrate reduction, as strains lacking these clusters only 332 reduced ~65% of wild type levels. But when only extEFG or extHIJKL was present, they were 333 not sufficient to reduce Fe(III)-citrate to more than 25% of wild type levels. 334

335
In contrast, the omcBC cluster or the extABCD cluster alone was sufficient for Fe(III)-citrate 336 reduction, and the extABCD cluster alone was sufficient for electrode growth. These phenotypes 337 could be due to electron acceptor preferences of each complex, or differential expression driven 338 by each electron acceptor, but in either case, each gene cluster was linked to phenotypes only 339 under specific conditions. Deletion of all five conduits resulted in complete elimination of metal 340 reduction abilities, while activity remained when the Δ5 strain was grown using electrodes as The genetic analysis presented here confirms a role for these unstudied conduits in extracellular 374 respiration. All mutants still containing at least one cluster retained at least partial activity 375 towards metals, and deletion of the omcBC region, plus all three ext clusters, finally was able to 376 eliminate metal reduction. This need to delete more than one conduit cluster helps explain prior 377 variability and rapid evolution of suppressors in ∆omcB-only mutants. In the case of electrodes 378 at both high and low potentials, only deletion of extABCD affected growth. Since residual 379 electron transfer to electrodes was still detected after deletion of all clusters, additional 380 mechanisms remain to be discovered. Overall, these data support the conclusion that for all 381 tested metal and electrode acceptors, more than one conduit is functional and capable of 382 participating in electron transfer. We found no pattern of specific gene clusters being required at 383 particular redox potentials, suggesting that periplasmic proteins act as a 'translator' to interface 384 the array of outer membrane complexes with the energy conserving inner membrane 385 cytochromes ImcH and CbcL. 386 387 More difficult to resolve is whether each putative conduit is designed for interaction with specific 388 extracellular substrates. The fact that single cluster mutants performed differently with each 389 substrate, along with evidence that omcB could not complement electrode growth while 390 extABCD could, supports the hypothesis of substrate specificity. Promoters more active in the 391 presence of Fe(III) vs. Mn(IV) could create some of these phenotypes, but differential 392 expression still suggests cells prefer to use each cluster under specific conditions. Some 393 complexes may preferentially interact with secreted extracellular proteins who carry electrons to 394 the final destination, and activity from a complex is masked in the absence of its partner protein. 395 While many extracellular proteins are known to be involved in electron transfer, such as OmcS, 396 OmcE, OmcZ, PgcA, and pili, a lack of secreted proteins encoded within omcBC or ext gene 397 clusters argues against co-evolution of dedicated partners. The availability of strains containing 398 only one gene cluster will enable easier purification, engineered changes in expression levels, 399 and protein-protein interaction studies to test these hypotheses. saline, subsurface, and fuel cell environments (Fig. 8). In about 1/3 of cases, the entire cluster is 421 conserved intact, such as extABCD in G. anodireducens, G. soli, and G. pickeringii (Fig. 8B). 422 However, when differences exist, they are typically non-orthologous replacements of the outer 423 surface lipoprotein, such as where extABC is followed by a new cytochrome in G. 424 metallireducens, Geoalkalibacter ferrihydriticus, and Desulfuromonas soudanensis. 425 Conservation of the periplasmic cytochrome coupled to replacement of the outer surface redox 426 protein also occurs in the omcB and extHIJKL clusters (Fig 8A and D).  Table S1), as 492 well as longer-read or single molecule sequencing, were we able to verify and isolate strains in 493 which complete loss of the omcBC cluster occurred, and dispose of hybrid mutants. Whole-494 genome resequencing was also performed on strains containing only one cluster, such as the 495 strain containing only extABCD, especially since this strain has an unexpected phenotype 496 where it produced more current than wild type. Thorough verification by PCR and whole 497 genome sequencing are recommended to confirm deletions of large and repetitive regions such 498 as the omcBC cluster. 499 500 Mutants lacking a single gene region were used as parent strains to build additional mutations. 501 In this manner, six double gene-cluster deletion mutants, four triple-cluster deletion mutants and 502 one quintuple-cluster deletion mutant lacking up to nineteen genes were constructed ( Fig. 1; 503 Table 1). For complementation strains, putative conduits were amplified using primers listed in 504 Table S1 and  sequence length and 40% identity based on amino acid sequence within the 537 Desulfuromonadales. A higher percent identity was demanded in this search due to the high 538 heme binding site density with the invariable CXXCH sequence. Only ExtJ and ExtL were 539 excluded from the search and the OmcBC region was collapsed into a single cluster due to the 540 high identity shared between the two copies. The gene neighborhood around each homolog hit 541 was analyzed. With a few exceptions (see Table S2), all homologs were found to be conserved 542 in gene clusters predicted to encode cytochrome conduits and containing several additional 543 homologs to each corresponding G. sulfurreducens conduit. The proteins within each 544 homologous cytochrome conduit that did not fall within the set cutoff were aligned to the amino 545 acid sequence of the G. sulfurreducens component they replaced using ClustalΩ (Sievers et al., 546 2011). 547   Schematic representation of cytochrome conduits from the Desulfuromonodales with homologs 838 to either A) OmcBC, B) ExtABCD, C) ExtEFG, or D) ExtHIJKL. Red arrows = putative outer 839 membrane products with a predicted lipid attachment site, yellow arrows = predicted periplasmic 840 components, and green arrows = predicted outer membrane anchor components.               Growth using 55 mM Fe(III)-citrate as an electron acceptor by A) single conduit cluster deletion mutants, B) triple mutants lacking all but one cytochrome conduit, as well as the Δ5 strain lacking all five cytochrome conduits, C) mutants in an ΔomcBC background strain, and D) Δ5 mutants expressing omcB or extABCD or carrying an empty expression vector as control. All experiments were conducted in triplicate and curves are average ± SD of n ≥ 3 replicates.  Schematic representation of cytochrome conduits from the Desulfuromonodales with homologs to either A) OmcBC, B) ExtABCD, C) ExtEFG, or D) ExtHIJKL. Red arrows = putative outer membrane products with a predicted lipid attachment site, yellow arrows = predicted periplasmic components, and green arrows = predicted outer membrane anchor components. Complete clusters with all components sharing >40% identity to the corresponding G. sulfurreducens cytochrome conduit are represented in boxes to the left of each gene cluster. Clusters in which one or more proteins are replaced by a new element with <40% identity are listed on the right side of each gene cluster. Proteins with numbers indicate the % identity to the G. sulfurreducens version. a OmcBC homologs in these gene clusters also encoding Hox hydrogenase complexes. b Gene clusters have contiguous extBCD loci but extA is not in near vicinity, as extA were found un-clustered in separate parts of the genome for some of those organisms (see Supplemental  Table S2). c Gene cluster has additional lipoprotein decaheme c-cytochrome upstream of extE. d Lipid attachment sites corresponding to ExtJL could not be found but there is an additional small lipoprotein encoded within the gene cluster. For ExtHIJKL encoding clusters, homologs depicted above extH are found in gene clusters containing only extI, whereas homologs depicted below extH are found in gene clusters containing full extHIJKL loci. Upstream and on the opposite strand to all gene clusters homologous to extHIJKL there is a transcription regulator of the LysR family, except e , where there is no transcriptional regulator in that region, and f , where there are transcriptional regulators of the TetR family instead.