Deg-family protease homologs were identified from sequenced genomes of 11 marine Synechococcus strains (WH8102, CC9605, BL107, CC9902, WH78903, WH7805, RS9917, CC9311, RS9916, WH5701, and RCC307) and 12 Prochlorococcus strains (MIT9303, MIT9313 CCMP1375, MIT9211, MIT9812, MIT9301, AS960, MIT9215, MIT9515, CCMP1986, NATL2A, and NATL1A). Deg-family proteases are characterized by the presence of both a C-terminal PDZ domain and a trypsin domain and identified in NCBI as COG0265. Homologs were first identified using the MicrobesOnline database (Price et al., 2009) to search for paralogs and homologs of SYNW2176 and COG0265. These results were then verified using published data of genome-wide comparisons of the 11 Synechococcus strains conducted by Dufresne et al. (2008) where the three groups were separately clustered for Synechococcus (Clusters 8096, 8095, and 13). The genes were also searched against the ProPortal database (Kelly et al., 2012) where the genes are grouped in one cluster (CyCog4045). Finally, a reciprocal BLAST was done for each gene found against its own respective genome to identify any paralogs that might have been missed. From this search a total of 59 Deg-family proteases were identified in the 23 genomes searched. The amino acid sequences were then aligned using MUSCLE (Edgar, 2004), trimmed manually and curated using Gblocks 0.91 (Talavera and Castresana, 2007). A maximum likelihood tree was then built using PhyML (WAG substitution model, 4 substitution rate categories) and branch support values were calculated with an approximate likelihood ratio test (aLRT SH-like) (Anisimova and Gascuel, 2006; Dereeper et al., 2008; Guindon et al., 2010). The alignment was later run through ProtTest (Abascal et al., 2005) which determined that an LG substitution model was best. The LG-based tree and the WAG-based tree were very similar, with one difference that included a change in position for the two group 3 Prochlorococcus genes within the group 3 branch. Since this was of uncertain significance and the support values for the WAG-based tree were better, the WAG-based tree is shown. TreeDyn was used for initial tree rendering. Conserved neighborhoods were assessed using MicrobesOnline (Price et al., 2009).

Three-liter batch cultures of WH8102 were grown to mid-exponential phase in chelexed synthetic ocean water (SOW) (Dupont et al., 2008; Stuart et al., 2009) and 10 μM CuEDTA (pCu 10.1),1 μM CuEDTA (pCu 11.1) or water (control) was added. For the 10 μM addition, 500 ml of culture was harvested by centrifugation at 1, 2, 4, 6, and 8 h after addition in three biological replicates. For the 1 μM addition samples were harvested at 2 and 24 h after addition in two biological replicates. RNA was extracted from the cell pellet, cDNA synthesized and QRTPCR performed as described in Stuart et al. (2009), with the exception that QRTPCR was normalized to total RNA, 250 ng of total RNA was used in triplicate cDNA reactions for each sample. DNA contamination controls with no reverse transcription were performed with every reaction. Primers for genes assayed are listed in Table S2. Three technical replicates were ran for each of the three biological replicates. A genomic DNA standard curve was used to calculate average transcript abundance for each sample, with subtraction of DNA-contaminant control. Biological triplicates were then averaged and log₂ fold changes between no-Cu controls and Cu-treated samples were calculated for each gene and time point. Statistical significance of the fold change was calculated by a one-sample t-test and a p-value less than 0.05 was considered significant.

Gene inactivations were conducted as described in Brahamsha (1996b) with several modifications in vector construction for the selection of double recombinants. Synechococcus sp. strain SYNW2176- in which the gene SYNW2176 is interrupted by an antibiotic resistance cassette, kanamycin (kan^R) was constructed as follows. A 1997-bp region surrounding the 1125-bp gene was PCR-amplified (Figure S2, Table S2, 2176 dbl primers) and cloned into pCR®2.1-TOPO® (Life Technologies, Carlsbad CA) according to the manufacturer's specifications. The vector was then digested with EcoRI and the 1997 bp + 10 bp insert was ligated into the EcoRI site of pBR322 (New England Biolabs, Ipswich, MA). Next the vector was digested with SmaI and the kanamycin resistance cassette (kan^R) from pUC4K (Taylor and Rose, 1988) was cut out of the pUC4K (using BamHI digestion, gel purification and blunting) and ligated into the SmaI cut site. The construct was then digested with EcoRI and the insert was ligated into pRL271, a vector for sacB-mediated positive selection for double recombinants in gram-negative bacteria (Cai and Wolk, 1990). The new vector, pRL271_kanR_2176 was then transformed into E. coli MC1061 (pRK24, pRL528) and introduced via conjugation into Synechococcus sp. strain WH8102 as described in Brahamsha (1996b). The pour plates contained 5% sucrose to select for double recombinants, and 25 μg/ml kanamycin. Mutants were verified using PCR (Figure S2, Table S2).

One liter batch cultures of Synechococcus sp. strain WH8102 (WH8102), and SYNW2176- were grown in chelexed SOW with stirring as previously described in Stuart et al. (2009). For mutant cultures, kanamycin was added at a final concentration of 20 μg/ml to 50 ml stock cultures that were transferred weekly, but antibiotics were not added to 1 L growth assay cultures, because kanamycin may bind copper. For growth assays, cultures were grown to early log phase and 50 ml were aliquoted into acid-washed glass flasks and placed on a shaker plate at 23°C. Cultures were given 1 day in these conditions to adjust to the transfer, after which CuEDTA was added in triplicate at 10, 4, 1, or 0.4 μM and sterile milliQ water was added to the controls. Using the MINEQL algorithm (Westall et al., 1976) this approximately corresponds to equilibrium free copper levels equivalent to pCu 10.1, pCu 10.5, pCu 11.1, and pCu 11.5. Flow cytometry samples were collected daily for up to 7 days and were fixed and analyzed as described in Stuart et al. (2009).

Outer membrane protein preparation was done as described in Brahamsha (1996a) and McCarren and Brahamsha (2007) with the exception that SOW medium was used for culturing and washing instead of SN. Briefly, 1 L of WH8102 or 2176- was grown up to the mid-exponential phase. For copper stress conditions, CuEDTA was added during mid-exponential at pCu 10.5 (4 μM) and samples were incubated for 6 h before harvesting. 10 ml of culture was filtered through a 0.2 μm polycarbonate filter and the spent medium was frozen at −80°C. Cells were harvested by centrifugation, washed with sterile medium, and resuspended in 50 mM Tris pH 8.0, 50 mM EDTA and 15% sucrose and incubated on ice for 30 min. Samples were subsequently spun down at 10,000 × g for 10 min at 4°C. The cell pellet from which the outer membrane had been stripped was frozen and the supernatant containing the outer membrane fraction was spun at 28,000 × g for 90 min at 4°C. The resulting high speed pellet (HSP), which consists of the outer membrane, was resuspended in sterile water. The high-speed supernatant (HSS), which contains periplasmic proteins as well as cell-surface proteins solubilized by EDTA, and the spent medium fraction were concentrated with an Amicon Ultra 4 30,000 MWCO filter (Millipore, Billerica, MA). Protein fractions were run on NuPAGE Novex Tris-Acetate 3–8% gels (Life Technologies, Carlsbad, CA) under denaturing conditions according to the manufacturer's specifications. For imaging, gels were stained with SYPRO Ruby stain (Life Technologies, Carlsbad, CA) according to the manufacturer's specifications. For band excision, Coomassie Blue R-250 (Thermo Scientific) was used to stain according to the manufacturer's specifications. Excised bands were sequenced by the UCSD Biomolecular/Proteomics Mass Spectrometry Facility.

Alkaline phosphatase activity assays and phosphate analysis were carried out using colorimetric assays as described in Tai et al. (2009). Briefly, for alkaline phosphatase activity 180 μl of sample of standard were put in a microplate and 20 μl 135 mM p-nitrophenyl phosphate in 1 M Tris (pH 8.0) added. Plates were read at 405 nM absorbance at 0, 30, 60, and 90 min. Phosphate concentrations were measured using a standard ammonium molybdate assay, samples were diluted 20-fold and 200 μl was added to a microplate with 20 μl reagent, incubated for 30 min and the absorbance at 880 nM was measured. Duplicate 50 ml cultures were inoculated with either wild-type WH8102 or SYNW2176- into SOW media and assayed for 14 days. Cell counts were determined by flow cytometry using a BD FACSort as described in Stuart et al. (2009).

Eleven sequenced genomes of marine Synechococcus all contain three Deg-family proteases, which can be separated phylogenetically into three main groups. The picocyanobacterial deg-family proteases have only one PDZ domain, compared with the E. coli family, which has two. Of the 12 Prochlorococcus genomes included in the analysis only two (strains MIT9313 and MIT9303, underlined) had a group 3 Deg protease (Figure 1A). MIT9313 and MIT9303 have much larger genomes than the other 10 Prochlorococcus strains analyzed (2.4 and 2.6 Mbp vs. 1.6–1.8 Mbp). In addition, the group 1 proteins from MIT9313 and MIT9303 grouped with the Synechococcus sequences in contrast with a more typical phylogenetic placement at the base of Prochlorococcus gene trees as seen in their group 2 proteins. The three Deg-family proteases previously studied from the freshwater cyanobacterium Synechocystis PCC6803 were most similar to each other. Based on pairwise BLASTp scores group 1 member SYNW2176 was more similar to HhoA (DegQ) and group 2 member SYNW0608 was more similar to HtrA (DegP). However, the alignment based trees using all members of the group did not support this. It is thus difficult to predict whether group 1 is homologous to HtrA/DegP or HhoA/DegQ. Although there was no clear predicted operon for any of the three Deg protease groups, the surrounding genes for each of these groups in Synechococcus were also well conserved (Figure 1B).

We used QRTPCR to investigate induction over time in response to copper stress (pCu 10.1 addition). In previous microarray expression work on copper stress, we found that SYNW2176 was induced after 2 h of copper stress at both pCu 10.1 (Cu²⁺ 10⁻¹⁰ M; 10 μM CuEDTA) and pCu 11.1 (Cu²⁺ 10⁻¹¹ M; 1 μM CuEDTA). The other two paralog Deg-proteases (SYNW0608, SYNW1537) were not differentially regulated by copper stress in either strain WH8102 or CC9311 (Stuart et al., 2009). We looked at expression of three genes including SYNW2176, SYNW0807 (a predicted sensor histidine kinase thought to be involved in osmoregulatory response) and SYNW0514 (predicted GroEL chaperonin). The putative specialized stress response genes, SYNW2176 and SYNW0807, both sustained induction over at least 6 h in response to copper stress whereas for GroEL, the general stress response gene, induction was only detected at 2 h after copper stress (Figure 2). We also looked at expression following a lower copper addition (pCu 11.1) over 24 h and found that SYNW2176 was induced at 2 h and maintained induction over 24 h (Figure S1). In contrast, GroEL was downregulated and SYNW0807 was not induced at 2 h by this lower level of copper, but by 24 h had increased (Figure S1). Based on expression data which showed sustained induction to both high and moderate copper additions, we expected that a SYNW2176 mutant would be more sensitive to copper.

We inactivated Deg-family protease SYNW2176 and investigated growth in response to copper. Insertional inactivation of SYNW2176 through double recombination (designated strain “SYNW2176-”) was validated using PCR (Figure S2, Table S2). Surprisingly, growth assays with copper additions showed that SYNW2176- exhibited a clear increased tolerance to copper additions relative to the wild-type (Figure 3). The wild-type strain showed decreased growth rates relative to the no-Cu control in response to copper levels as low as pCu 11.5 (0.4 μM CuEDTA, Cu²⁺ at 10^−11.5 M) and failed to grow at all at pCu 11.1 (1 μM CuEDTA), whereas SYNW2176- only started to show a significantly decreased growth rate at pCu10.5 (4 μM CuEDTA).

Since envelope stress may affect the outer membrane composition we looked at copper induced changes in wild-type outer membrane fractions (Brahamsha, 1996a; McCarren and Brahamsha, 2007). The technique generates two outer membrane fractions, an insoluble HSP consisting of the outer membrane and a soluble high speed supernatant (HSS) fraction largely consisting of cell surface proteins solubilized by the EDTA treatment. Using HPLC tandem mass spectrometry (“shotgun proteomics”) we sequenced single samples of insoluble outer membrane protein preparations (HSP) from both a control and 6-h copper stress (pCu 10.5) in WH8102. A total of 49 proteins were detected (with at least two unique peptide hits above 95% confidence) with 47 proteins detected in the control condition and 26 under copper stress (Table S1). The two proteins found only under copper stress were SYNW0418, a hypothetical protein of unknown function, and SYNW2077, a 50S ribosomal protein (Table S1).

WH8102 has four genes annotated as possible porins, of which only two were detected in our insoluble outer membrane fractions, SYNW2224, and SYNW2227. Under copper shock these two porins (SYNW2224 and SYNW2227) appeared to be downregulated because peptide spectral count percentages (proportion of total spectral counts) from both porins combined were lower under copper (7.4%) than the control condition (10.4%). Also notable under copper shock were higher levels of the periplasmic phosphate binding ABC transporter protein (SYNW1018) and undetectable levels of SwmB, a cell surface motility protein (Table S1). SYNW2176 itself was not detected in either sample and given that it has transmembrane domains this may indicate that it is located in the inner membrane or conversely its abundance is below the detection limit. Copper treatment clearly caused changes in relative protein abundance in the cell surface/outer membrane compartment of the cells.

We examined the outer membrane protein composition of SYNW2176- in two ways, first using gel electrophoresis and then shotgun proteomics. We compared the outer membrane profiles of the wild type and the SYNW2176- mutant under exponential growth. The gels showed a distinctly different pattern of bands in SYNW2176- relative to the wild-type in the insoluble HSP fraction (Figure 4). Most notably, there was an additional band in the mutant at around 120 kDA. This band was also detectable in the soluble HSS fraction (Figure S3A). These differences remained unchanged 6 h after copper additions (Figure S3B). There was also a distinct difference in the ~80 kDA region in the mutant in the soluble and insoluble fractions. In the mutant the band was sharper and darker relative to the wild-type.

Duplicate biological replicates of the soluble (HSS) and insoluble (HSP) fractions of outer membrane samples from SYNW2176- and wild-type in exponential growth were also analyzed with shotgun proteomics. In the insoluble fraction a total of 11 proteins were detected in both SYNW2176- biological replicates (with at least two unique peptide hits above 95% confidence) and 12 proteins were detected in both wild-type replicates. In the soluble fraction 13 proteins were detected in SYNW2176- and 12 in wild-type. There were a total of 22 proteins detected in at least one set of biological replicates (Table 1).

Several differences in protein levels were detected between the wild-type and mutant, a few of which were also induced in wild-type under copper stress (Table 1). One difference was in the abundant porin, SYNW2224, which was lower in both the mutant and under copper stress compared to wild-type. Also, higher levels of SYNW1018 (an ABC transporter phosphate-binding protein) were found in both the mutant and under copper stress compared to wild-type.

The less abundant porin SYNW2227 was only detected in the mutant, and it was not detected following 6 h of copper stress to the wild-type. Comparing the two porins revealed that predicted isoelectric points of SYNW2227 and SYNW2224 were drastically different. Modeling of the outer membrane loops using published methods (Bagos et al., 2004) revealed far fewer positive residues in the outer loops of SYNW2227 (Figure S4, Table S4). SYNW2227 is also missing two cysteine residues on the predicted outer loop portion of the protein that are present in SYNW2224 and one of the other undetected porins, SYNW2223. These differences between the porins suggest that SYNW2227 may have reduced permeability to copper or reduced sensitivity to copper due to the fewer cysteines potentially sensitive to copper binding.

There was little overlap between proteins upregulated in the SYNW2176- mutant and genes identified from our previous transcriptional work on Cu-stress that we predicted would be upregulated in the mutant given its copper tolerance (Table 1). Instead, there was a strong correlation with phosphate-stress inducible genes (Table 1). Several studies have looked at the phosphate limitation response in Synechococcus WH8102, including two microarray expression studies by Tetu et al. (2009) and Thomas et al. (2009) and a study by Mazard et al. (2012) that compared the phosphorus stress response in several Synechococcus strains. Of the 22 identified proteins in our study, 10 are differentially transcriptionally regulated under phosphate-starvation and similarly differ in protein abundance in the SYNW2176- mutant relative to wild-type (Table 1).

The most apparent difference was the presence of an additional protein in the mutant, SYNW0762, which is encoded by a phosphate-stress inducible gene. SYNW0762 is annotated as a hypothetical protein whose closest BLAST hit is to an atypical alkaline phosphatase with a phytase domain in Synechococcus WH7803 (Table 1). The protein has two large repeat regions that are identified as WD40/YVTN repeat-like domains. When the additional band seen in the 120 kDa region of mutant outer membrane preparations was excised and sequenced, this protein was the most abundant hit (Figure S3B, labeled “1,” Table S3). Also, SYNW0165, a phosphate-inducible protein of unknown function, was almost 3-fold higher in the mutant insoluble fraction and in the soluble fraction it was only detected in the mutant. Two putative alkaline phosphatases, SYNW2391, and SYNW0196, both of which are induced by phosphate-stress (Tetu et al., 2009) were also detected at higher levels in the mutant.

Two of the proteins identified were both phosphate inducible and copper stress responsive—the phosphate ABC transporter SYNW1018 and SwmB were both more abundant in the mutant. SYNW1018 was 2.2-fold more abundant in the mutant insoluble fraction and SwmB was 4-fold higher in the mutant soluble fraction. Additionally, two proteins that are down-regulated transcriptionally by phosphate stress in the wild-type (Tetu et al., 2009), SYNW2009 and SYNW1565, were less abundant in the mutant. It should be noted that five phosphate-stress inducible proteins did not follow this pattern: the aforementioned porin SYNW2224, SwmA, and SYNW0759, were more abundant in wild-type; SYNW0406 for which no difference in protein abundance was detected; and SYNW1886 which was only detectable in the wild-type. SYNW2224, the porin that is down-regulated in the mutant, is an interesting exception since it is one of only two proteins detected that was upregulated under phosphate limitation but downregulated under copper stress. All outer membrane preparations were obtained from exponentially growing cultures (in media with 86 μM PO₄) that were not expected to be phosphate-limited at this stage of growth.

Considering the abundance of phosphate-stress inducible proteins expressed in SYNW2176- we compared alkaline phosphatase activity, a hallmark of the onset of phosphate limitation, between SYNW2176- and the wild-type. Mutant and wild-type cultures were inoculated in phosphate-replete media, and alkaline phosphatase activity and phosphate levels were measured during growth. The mutant exhibited alkaline phosphatase activity that did not change significantly over the growth period (Figure 5). By contrast, in wild-type, as expected, alkaline phosphatase activity was very low until day 11, when phosphate levels began to decrease, and then increased dramatically. In the mutant, low activity was detected by day 6, when phosphate levels were still high, and activity stayed low but detectable throughout the rest of the experiment. The mutant grew slightly more slowly than the wild-type under these conditions, but by day 6 after inoculation both strains were in exponential phase (Figure 5B). Since the alkaline phosphatase assay has a low sensitivity threshold, to confirm the early time points when cell numbers were low we replicated the experiment and concentrated the samples 50-fold before measurement. From this we were able to confirm that while in exponential growth (days 6 and 7), the mutant does have slightly higher activity (8.3 × 10⁻¹⁰ μmoles/min/10⁶ cells/ml ± 4.0 × 10⁻¹⁰) than the wild-type (3.58 × 10⁻¹⁰ μmoles/min/10⁶ cells/ml ± 4.6^*10⁻¹¹) while both had high phosphate levels (wild-type 50 μM ± 1.4 and mutant 74 μM ± 8.3). The mutant maintains alkaline phosphatase activity around 1.0 × 10⁻⁹ μmoles/min/10⁶ cells/ml never having levels above 2.0 × 10⁻⁹ μmoles/min/10⁶ cells/ml, whereas the wild-type has very low activity until phosphate levels decrease after which it induces activity sharply to around 6.0 × 10⁻⁹ μmoles/min/10⁶ cells/ml. From both these experiments it is clear that the mutant has constitutive, low alkaline phosphatase activity that is not induced by phosphate levels. This suggests either a constitutive stress response or an inability to sense phosphate levels.

The phylogenetic grouping of the Deg-family proteases in marine Synechococcus shows three distinct groups. The group 1 Deg-proteases appear to have an inducible function distinct from the other 2 groups. Our transcriptional data show that SYNW2176 exhibits a sustained high level of induction following copper stress for at least 24 h and is quite sensitive to copper as it was induced by 2 h at the lower copper level as well (Figures 2, S1). This is in contrast to GroEL chaperonin, which is induced only at 2 h following the high copper addition and suggests a more specialized stress response for SYNW2176, perhaps to envelope stress, as in E. coli (Meltzer et al., 2009). Published expression data also shows inducible activity in the group 1 members with SYNW2176 and sync_2523 both upregulated in response to copper addition and SYNW2176 slightly upregulated in response to early phosphate stress and downregulated in response to late phosphate starvation (Stuart et al., 2009; Tetu et al., 2009). Additionally, the group 1 member SynWH7803_2187 is up in response to methyl viologen and H₂O₂, also oxidative stressors (Blot et al., 2011). In contrast, neither group 2 nor 3 members in CC9311 or WH8102 are upregulated in response to copper, and in WH7803, the group 3 member is only slightly upregulated in response to methyl viologen and H₂O₂ (Blot et al., 2011).

Most of the Prochlorococcus strains do not have a group 3 Deg-protease suggesting that group 3 may serve a redundant function that can be compensated for by the other two in the more streamlined Prochlorococcus genomes. Prochlorococcus strains that have been tested are more sensitive to copper than Synechococcus. Strain MED4 (which lacks a group 3 member) is in fact more tolerant to copper than MIT9313 and SS120 (Mann et al., 2002). There currently are not equivalent transcriptional stress response experiments on Prochlorococcus. However, from data available on high light stress in Prochlorococcus strain MED4, there is upregulation of the group 2 member in response to high light and blue light, but no regulation of the group 1 member in response to light, nitrogen, phosphate, or iron stress (Martiny et al., 2006; Steglich et al., 2006; Tolonen et al., 2006; Thompson et al., 2011). In a manner more similar to Synechococcus, Prochlorococcus MIT9313, which has three Deg proteases, upregulates its group 1 member in response to nitrogen limitation but does not appear to upregulate either of its other Deg proteases in response to nitrogen, phosphate, or iron stress (Martiny et al., 2006; Tolonen et al., 2006; Thompson et al., 2011). From these limited data it appears that the Prochlorococcus strains lacking a group 3 member may have different induction responses to stress for the remaining two Deg proteases.

Our data on the SYNW2176- mutant reveals a complex phenotype. This includes increased copper tolerance and a significantly altered outer membrane composition with a putative alkaline phosphatase/phytase (SYNW0762) not detected under in wild type. The mutant outer membrane also has an extra porin and lower levels of the main porin (which is downregulated by copper in wild type cells.). The phenotype also includes higher abundance of a copper-inducible outer membrane protein (SYNW1018); higher levels of many phosphate-limitation-induced proteins; a longer lag phase after transfer; and a lack of phosphate limitation-inducible alkaline phosphatase activity. While diverse in nature, all of these could point to a constitutive stress response.

Several hypotheses can be put forward to explain this complex phenotype. SYNW2176 could be part of a regulatory cascade that responds to envelope stresses and regulates a certain set of outer membrane proteins in response to these stresses. In its absence some of these proteins are constitutively expressed whereas others are not induced. One complication for this hypothesis is that it does not correlate directly with previous transcriptional microarray data. If SYNW2176 is a regulatory protein, we would expect to see other genes responding when it is up-or down-regulated. However, although SYNW2176 is up in response to copper and early phosphate stress and down in response to late phosphate stress, we do not see a correlation with other genes, like SYNW0762 (the atypical alkaline phosphatase/phytase protein only detectable in the mutant) between these datasets.

A simpler hypothesis is that SYNW2176 functions in protein quality control, as does DegP in E. coli. If this is the case, the mutant phenotype could be the result of a buildup of damaged membrane and periplasmic proteins, that would otherwise be degraded. This buildup could be sensed by other cellular systems, perhaps other Deg-family proteins, leading to a specific stress response regulating porins and other membrane proteins. Indeed, in E. coli ΔdegP mutants have activated σ^E and Cpx pathways, likely a result of envelope stress from the buildup of damaged proteins usually degraded by DegP (Leiser et al., 2012). Additionally, activation of the σ^E pathway can result in a reduction of the major OMP protein levels (Johansen et al., 2006). This buildup of damaged proteins and active envelope stress response could lead to a decreased permeability or transport of copper.

How copper crosses the outer membrane is still not entirely clear but porins can be regulated by copper, can bind copper, and may allow passive import of unchelated copper (Yoneyama and Nakae, 1996; Caille et al., 2007; Balasubramanian et al., 2011). Moreover, a porin-deficient (most likely OmpC and OmpF) mutant in E. coli is resistant to elevated copper levels, indicating such changes can be sufficient to lead to increased copper resistance (Lutkenhaus, 1977). However, this is complicated by the fact that a single OmpC mutant is copper sensitive (Egler et al., 2005), indicating that porins may vary in their ability to let copper into the periplasm. There is some evidence that the charged residues in the outer loops of porins contribute to voltage gating in Haemophilus influenzae (Arbing et al., 2000). Our comparison of the outer loop charged residues of the two porins suggest they may allow diffusion of very different sets of substrates. In addition, copper could potentially react with cysteines on SYNW2224 so one could speculate that its decrease in SYNW2176- and the new presence of the SYNW2227 porin lacking cysteines could be a mechanism increasing copper resistance. Given this, our data suggests that the two porins may have different permeability to copper ions and changing the levels of these porins may change copper permeability and copper resistance.

There could also be a role for SYNW0762, SYNW1018 or SwmB, all of which were much higher in the mutant than wild type although it is not clear what these would do to increase Cu resistance. Phosphatases and phytases have been linked to uranium resistance in bacteria (Martinez et al., 2007; Nilgiriwala et al., 2008; Yung et al., 2014) and copper resistance in fungi (Tsekova and Galabova, 2003), likely through biomineralization of the metals with phosphate. While the marine chemical environment is quite different it is possible that these extra alkaline phosphatases like SYNW0762 may play a role in copper detoxification directly. SYNW1018 is part of a well-conserved group of phosphate-binding proteins, PstS, found in bacteria and in E. coli is part of the high-affinity phosphate transport system (Makino et al., 1989). Oligotrophic Synechococcus strains including WH8102 tend to have multiple paralogs of this protein and it is unclear what different functions the four present in WH8102 may have (Scanlan et al., 2009).

The high number of phosphate-stress inducible proteins in the mutant indicates constitutive phosphate stress. High affinity phosphate transport in Synechococcus involves both a membrane-bound ABC transporter and a periplasmic phosphate-binding protein as well as a phosphate-sensing histidine kinase, PhoR, that may be membrane bound (Scanlan et al., 2009). Thus, a buildup of damaged periplasmic and membrane proteins in the mutant could disproportionately affect phosphate transport as compared to nitrate transport, for example, which is not transported by ABC family transporters but by a permease (Sakamoto et al., 1999). Reduced transport of phosphate could in turn lead to phosphate limitation in the presence of high phosphate levels and induction of a phosphate limitation response. Indeed, Tetu et al. (2009) found changes in expression of many outer membrane proteins in response to phosphate starvation as well as regulation of SYNW2176 (up slightly under early phosphate stress and down under late phosphate stress), so the mutant's distinct phenotype regarding phosphate acquisition is not surprising. During exponential growth SYNW2176- had a higher level of alkaline phosphatase activity than wild-type, which had virtually undetectable levels until phosphate levels decreased (Figure 5). It also has higher levels of many phosphate-inducible proteins than wild-type in exponential growth as well as in low-phosphate growth (Table 1). Despite higher levels of proteins involved in phosphate stress response, or perhaps because of it, the mutant had a longer lag time (Figure 5), all of which may indicate the costs of constitutively activated stress.

While quantification of damaged proteins in the periplasm is beyond the scope of this work, our data nevertheless shows the major role that SYNW2176 plays in determining outer membrane composition, perhaps through protease activity on specific membrane proteins. Our data clearly show that the presence/absence of SYNW2176 affects expression of many outer membrane proteins potentially playing a role in outer membrane permeability to a number of substrates. The Deg-proteases are moreover well-conserved in marine Synechococcus and their phylogeny appears to inform function. Our data also reveal the potential importance of porins in copper tolerance in marine Synechococcus. As is clear from the dramatic increase in copper tolerance in the mutant and the significantly altered outer membrane protein composition, Synechococcus outer membrane composition is key to understanding the copper sensitivity of this group. The outer membrane functions as both a barrier to toxins and a gateway for nutrients and copper is at the intersection of these functions as it is both a toxin and a micronutrient. Identifying proteins that contribute to copper permeability can provide insight into outer membrane function and how marine Synechococcus maintains a balance between importing limiting nutrients and keeping out toxins.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.