To determine the functions of the identified putative protein secretion systems, we generated derivatives of strain 3841 carrying insertion mutations in prsD (A895), toaD (A913), tobD (A896), virB6 (A897), autA (A1010), autB (A1011), autC (A1012) and tatC (A1052). The mutants were tested for polygalacturonase (PGase) activity, glycanase activity, motility, biofilm formation and growth on different carbon sources (mannitol and succinate). All mutants were able to grow on TY, Y-succinate and Y-mannitol plates.

None of the mutants was affected in cell-associated PGase activity, suggesting that this activity is independent of the identified secretion systems. The prsD (A898) mutant lacked extracellular glycanase activity, failed to form biofilm rings in shaking culture, and produced very mucoid colonies and viscous culture supernatants (not shown). Such phenotypes were observed previously with a prsD mutant of R. l. bv. viciae strain A34 12 59 . No such effects were observed with any of the other mutants including A913 (toaD), or A896 (tobD). To rule out the possibility of additive phenotypes or redundancy between the different Type I systems, we constructed a toaD/tobD/prsD triple mutant (A912) and toaD/prsD (A911) and tobD/prsD (A903) double mutants, but all had the same phenotypes as A898 (prsD).

Whereas the wild type (and all the other mutants tested) swarmed outward from the point of inoculation on 0.2% agar TY plates, the tatC mutant (A1052) formed a tight colony and showed no sign of swarming behaviour. Growth of the tatC mutant in liquid TY medium was slightly impaired (it had a doubling time of 4.5 h compared with 3 h for WT and the other mutants). Microscopy revealed that the tatC mutant was much more rounded in shape than the wild type.

To examine the role of the identified secretion systems in nodulation, pea seedlings (cv. Frisson) were inoculated with wild type R. l. bv. viciae 3841, A895 (prsD), A913 (toaD), A896 (tobD), A897 (virB7), A1010 (autA), A1011 (autB), A1012 (autC) and A1052 (tatC). After three weeks growth on nitrogen-limited medium, the plants inoculated with the tatC mutant (A1052) looked nitrogen starved, and the nodules formed were unable to fix nitrogen based on measurements of acetylene reduction; staining of these nodules with SYTO 13 67 revealed bacteria near the tip of the nodule, but they did not extend as far into the nodule as wild type bacteria (Fig. 3). All of the other mutants formed normal numbers of nitrogen-fixing nodules (data not shown). Nodulation experiments carried out with the alternative host plants broad bean (Vicia faba cv. The Sutton) and hairy vetch (Vicia hirsuta) mirrored the results obtained on pea plants.

A comparison by SDS-PAGE of the proteins in the culture supernatant of the tatC mutant (A1052) and the wild type grown in Y-mannitol liquid medium showed clear differences in the protein pattern (Fig. 4). Although there were many bands in common, A1052 consistently contained more total protein in the culture supernatant as estimated by loading proteins precipitated from equal amounts of culture supernatant. The abundant proteins of around 36 kDa produced by the wild type but missing from the supernatant of A1052 (tatC) were identified by MALDI-ToF as flagellins (RL0718 and RL0720). Although the TAT pathway secretes proteins to the periplasm and not directly to the extracellular space, the absence of flagellar proteins from the culture supernatant of a tatC mutant was previously shown for A. tumefaciens 54 .

Periplasmic proteins from R. l. bv. viciae 3841 and A1052 (tatC) were isolated by cold hypo-osmotic shock and analysed by gel electrophoresis. The absence of cytoplasmic contamination of the periplasmic protein preparation was checked by measuring glucose-6-phosphate dehydrogenase (G6PDH) activity 68 . Specific G6PDH activity in the periplasmic preparations was less than 4% of the specific G6PDH activity in whole cell extract. Although the TAT machinery exports proteins to the periplasm, there were no major differences in the periplasmic protein profile of R. l. bv. viciae 3841 and the tatC mutant A1052 (Fig. 4.).

Culture supernatant proteins of R. l. bv. viciae 3841and the prsD (A895), toaD (A913), tobD (A896), virB6 (A897), autA (A1010), autB (A1011) and autC (A1012) mutants grown in Y-mannitol medium were analysed by SDS-PAGE. Only the supernatant protein profile from the prsD mutant (A895) differed from that of the wild type and was missing seven protein bands (Fig. 5). The seven proteins corresponding to the indicated bands in Fig. 5 were identified by MALDI-ToF MS fingerprinting (Table 2). Six of these were either known Type I substrates (PlyB and RapA2) 62 or were predicted above (RL2961, pRL100309, pRL90140 and RL2412) to be secreted using bioinformatic analyses.

One of the proteins missing from the culture supernatant of the prsD mutant (A895) was a predicted nucleoside diphosphate kinase (NDK, RL1580). NDK is generally a cytoplasmic enzyme involved in nucleotide metabolism 69 and catalyses the general reaction N_xTP + N_yDP ↔ N_xDP + N_yTP, where N_{x or y} can be any base. Therefore, its presence in the supernatant of the wild type was unusual, although NDK has been shown to be secreted via a Type I system in P. aeruginosa 70 . The absence of NDK from the culture supernatant of A898 (prsD) indicates specific export by the PrsDE Type I system. To rule out cytoplasmic contamination of the culture supernatant, the small (~22 kDa) and highly expressed cytoplasmic protein RhiA was used as a marker. RhiA is a protein of unknown function 71 72 , and its high abundance and the availability of a specific antiserum 73 made it a good choice as a marker protein, because cell lysis would release large amounts of RhiA into the culture medium. RhiA could not be detected in the culture supernatant of bacteria grown for 4 days to mid-stationary phase (data not shown).

Double and triple mutants in toaD, tobD and prsD were made but no effects on supernatant proteins were seen unless the strain also carried a mutation in prsD, in which cased the same changes as in a prsD single mutant were observed. A mutation in bltD similarly had no observed effect on the profile of proteins isolated from the culture supernatant (data not shown).

Although the prsD mutant differed from the wild type in culture supernatant protein profile (Fig. 5), this only accounted for seven of the many culture supernatant protein bands observed by SDS-PAGE. Most of these bands were also observed in periplasmic fractions of wild type R. l. bv. viciae 3841, and many were indeed very abundant in this fraction, suggesting that these proteins might have reached the culture supernatant through unspecific leakage. Nevertheless, three protein bands remained which were exclusive to the culture supernatant. To identify these other proteins, bands 7, 11 and 12 were excised from the gel for MALDI-ToF MS analysis. Select bands present in both the culture supernatant and the periplasmic fraction were also excised from the gel to confirm their periplasmic origin. Table 2 and Figure 6 list 23 different proteins (including all seven bands missing from the supernatant of the prsD mutant) that were identified from the culture supernatant. Of these, proteins RL0718, RL0720 and RL0728 (Table 2) are predicted to be extracellular components of the flagellum and expected to be present in the culture supernatant 74 75 76 . Out of the 23 identified supernatant proteins (Table 2), 13 were predicted to be periplasmic by the SignalP 3.0 software 66 (Table 3). The distinction between signal peptide proteins and leaderless proteins is very clear, with proteins either having a signal peptide probability of 0.999 or better or less than 0.033. The signal peptides are predicted to be between 20 and 32 amino acid residues in length with probabilities between 0.803 and 1.000. One of the predicted signal peptide proteins (RL0518) is also predicted to be a TAT substrate by the TATP 1.0 77 , but not by the TATFIND 78 algorithm. The predicted signal peptide of RL0518 carries a twin-arginine motif, and the surrounding residues also suggest a TAT leader peptide. The absence of this protein could not be confirmed in the culture supernatant of the tatC mutant A1052 (Fig. 4) this could either be due to the abundance of other proteins in the culture supernatant masking its absence or the possibility that it is not a TAT substrate. Five of the supernatant proteins were predicted to be periplasmic substrate-binding components of ABC uptake transporters, two each were basic membrane proteins or conserved proteins of unknown function and one was predicted to be peptidyl-prolyl cis-trans isomerase. None had similarity to any known extracellular proteins from other bacteria. Table 3 lists all proteins identified in the culture supernatant and gives their predicted function and likely secretion system.

The only Type I secretion systems studied in rhizobia so far have been the PrsDE system 12 13 62 and systems dedicated to the export of bacteriocin 79 80 . This study reports the presence of a total of four predicted Type I secretion systems in R.l. bv. viciae 3841, encoded by the prsDE, toaDE, tobDE and bltDE genes. Of these, the prsDE-encoded system has been characterised previously 12 13 62 and is conserved in S. meliloti 15 , R.l. bv. trifolii 81 , R. etli and A. tumefaciens. Of the fourteen predicted Type I substrates identified in R.l. bv. viciae 3841, seven (NodO, PlyA, PlyB, PlyC, RapA2, RapB and RapC) were already known or predicted to be secreted by the PrsDE system based on previous work 12 13 62 82 . The other seven predicted Type I substrates were previously unknown; we demonstrated that four of these (RL2412, RL2961, pRL90140 and pRL100307) were secreted in a prsD-dependent manner, while another (RL0790) was subsequently shown to be secreted in complex (TY) growth medium in a prsD-dependent manner (A. Edwards and J. A Downie unpublished data). In addition to these predicted Type I substrates, we found that the presence of RL1580 (a predicted nucleoside diphosphate kinase) in the growth medium requires prsD. This suggests that the PrsDE system secretes at least thirteen proteins encoded by genes in diverse regions of the genome. This is a much broader specificity than typically observed with Type I secretion systems in other bacteria, where the secretion system is usually limited to the secretion of only one or a narrow range of closely-related substrates 83 84 .

A role for Type I secretion in the symbiosis has been shown previously, since the nodulation signalling protein NodO is secreted via the PrsDE system, 11 12 . In addition, prsD mutants of R.l. bv viciae strain A34 and R.l. bv. trifolii strain TA1 form nodules that are infected but do not fix nitrogen 12 81 . This is not the case with R.l. bv viciae strain 3841 and the reason for the difference is not known, although a prsD mutant of S. meliloti also forms nitrogen-fixing nodules 15 . Although the PrsDE secretion system in R.l. bv. viciae 3841 is not essential for nodulation, the proteins secreted by this system could nevertheless have an auxiliary function in nodulation as has already been demonstrated for NodO 5 6 . Several of the proteins whose secretion is prsD-dependent are predicted to be involved in surface attachment. This includes the

R

a

p

Two of the PrsDE-dependent secreted proteins are the glycosyl hydrolases PlyA and PlyB, which cleave the extracellular acidic exopolysaccharide 13 . A plyA plyB double mutant retained some glycosyl hydrolase activity, indicating that other extracellular glycosyl hyrolases were present 59 . It is possible that the other two predicted glycosyl hydrolases (PlyC and RL2412) could cleave rhizobial exopolysaccharides or could play a role in plant cell wall degradation similar to that observed previously in R. l. bv. trifolii 17 .

RL0790 is predicted to be a secreted protease. The observation that it was present in complex medium (containing peptone), but not in minimal medium, suggests it may be induced if the bacteria are relying on peptides for growth. This is unlikely in the rhizosphere, but infection threads and senescent nodules are environments rich in proteins, and so it might be anticipated that RL0790 could contribute to growth of bacteria in infection threads or in senescing plant cells.

Nucleoside diphosphate kinase (NDK) was not predicted to be a Type I substrate by the bioinformatics approach, but was absent from the culture supernatant of the prsD mutant. Exported NDKs are not without precedent in mammalian pathogens; P. aeruginosa, Vibrio cholerae and Mycobacterium tuberculosis secrete NDKs 70 87 88 89 , as do the parasitic nematodes Trichinella spiralis and Haemonchus contortus 90 91 . In P. aeruginosa, secretion of NDK into the culture supernatant has been shown to be dependent on the presence of the sequence DTEV near its C-terminus and it has been suggested that NDK secretion occurs via a Type I secretion system 70 . A very similar sequence (DTEI) is also present near the C-terminus of NDK from R.l. bv. viciae 3841. Extracellular ATP plays a role in macrophage apoptosis as a response to pathogen attack. In Arabidopsis thaliana, auxin and ATP are both transported by the metabolite transporter MDR1. Removal of extracellular ATP leads to FB1-induced cell death in a number of plants including legumes 92 . Extracellular NDK may possibly have an effect on modulating these responses, but as with the other PrsDE-dependent secreted proteins its role is probably minor, since the prsD mutant lacking secretion of NDK does not appear to differ from the wild type in nodulation phenotype. A role for Type I secretion in the symbiosis has been shown previously because the nodulation signalling protein NodO is secreted via the PrsDE system, although this is not essential for nodulation 5 12 . In addition, prsD mutants of R.l. bv viciae strain A34 and R.l. bv. trifolii strain TA1 form nodules that are infected but do not fix nitrogen 12 81 . This is not the case with strain 3841 and the reason for the difference is not known, although a prsD mutant of S. meliloti also forms nitrogen-fixing nodules 15 .

The predicted BltDE secretion system seems likely to be a bacteriocin/lantibiotic type exporter that secretes pRL90159 and pRl90161, large proteins encoded by genes upstream of the bltDE genes. Proteins similar to those encoded by bltDE are present in R. etli and a related system (Rzc) is involved in the export of medium bacteriocins in other rhizobia 79 80 implying a possible role for pRL90159 and pRl90161 as bacteriocin-like proteins. However, the sequences of pRL90159 and pRl90161 are more similar to hypothetical proteins from non-rhizobia such as Nitrobacter spp. than to proteins from other rhizobia

The other two predicted Type I secretion systems are enigmatic because we did not identify any secreted proteins that require toaD or tobD for their secretion, even though we made double and triple mutants (with toaB, tobD and prsD) to test for redundancy of function. This could mean that we have not identified appropriate growth conditions, the resolution of the protein gels is inadequate to identify the products (too low in amount or too small), or even that these are not protein secretion systems, but possibly secrete some other component. Genes similar to toaD are present in syntenic locations in S. meliloti (80% identity) 56 and A. tumefaciens (86% identity) 93 . A protein from Bradyrhizobium japonicum has similarity to TobD (52% identity). The genes surrounding the genes encoding these proteins also display similarity to those surrounding toaD and tobD in R.l. bv. viciae 3841, respectively, suggesting synteny and likely orthology. It is evident that these genes are not essential for growth or for the symbiosis with the tested host legumes pea, broad bean and vetch.

The R. l. bv. viciae 3841 tatABC-encoded secretion system genes are as expected from related bacteria such as A. tumefaciens 54 and R. l. bv. viciae UPM 791 38 . The tatC mutant exhibited phenotypes expected from previous work in A. tumefaciens and R. l. bv. viciae UPM791 in that the A. tumefaciens tatC mutant was also shown to be largely immotile, lacking flagellar proteins in the culture supernatant and the R. l. bv. viciae UPM791 tatC mutant also formed white nodules on peas and did not fix nitrogen. However, while the R. l. bv. viciae UPM791 tatC mutant only formed small nodules, the nodules formed by a R. l. bv. viciae 3841 tatC mutant were on average larger than those formed by the wild type. As in R. l. bv. viciae UPM791, the R. l. bv. viciae 3841 tatC mutant did not progress beyond the infection zone in the nodule, but unlike in R. l. bv. viciae UPM791 38 , viable bacteria could be isolated from nodules formed by the R. l. bv. viciae 3841 tatC mutant. The lack of growth and symbiotic nitrogen fixation in nodules could result from the inability of the tatC mutant to export to the periplasm electron-transport components required for respiration in the low oxygen environment in the nodule. Periplasmic proteins isolated from the tatC mutant and R. l. bv. viciae 3841 grown in TY medium give very similar patterns when separated by SDS-PAGE, suggesting that most TAT substrate proteins are either not expressed under the observed conditions or are expressed at such a low level that they are not visible on the SDS-PAGE gel.

Another major difference between the wild type and the tatC mutant is the increased quantity of periplasmic proteins in the supernatant. A similar increased leakage of periplasmic contents into the extracellular space has been shown in tatC mutants of E. coli 94 . In this strain, outer membrane stability is compromised by the mislocalisation of two cell wall amidases, AmiA and AmiC, which are TAT substrates 95 . The AmiC protein from R. l. bv. viciae 3841 (RL1742) is also predicted to be a TAT substrate, suggesting that the increased leakiness of the outer membrane is due to the mislocalisation of this protein.

Previous studies on the extracellular proteome of rhizobia such as that by Suss et al. 96 have successfully used a proteomic approach to identify novel substrates of specific secretion systems, but have largely ignored proteins still present in the culture supernatant of the mutant studied. In this study, we show for the first time that many of the culture supernatant proteins identified by MALDI-Tof-MS are predicted to be periplasmic, most of them being similar to well known periplasmic proteins such as substrate binding proteins that are unlikely to function extracellularly. Type II protein secretion system secrete periplasmic proteins across the outer membrane in many Gram-negative bacteria, but such a system is absent from R. l. bv. viciae 3841. Bands corresponding to the predicted periplasmic proteins in the culture supernatant were also observed in periplasmic preparations, arguing against a specific secretion mechanism and for unspecific leakage. Periplasmic proteins have also been identified in the culture supernatant of H. pylori 74 , but the mechanism of secretion, if any, was also unknown. The absence of the cytoplasmic protein RhiA suggests leakage from the periplasm rather than cell lysis as the reason for the presence of periplasmic proteins in the culture supernatant. Periplasmic proteins are found in the culture supernatant of R. l. bv. viciae 3841 at all stages of the cell cycle, and indeed are highly abundant during exponential growth. This indicates leakage from the periplasm during active growth, rather than cell death and lysis, as the likely origin of periplasmic proteins in the culture supernatant. Periplasmic proteins in the culture supernatant have previously been reported in R. leguminosarum RBL5523 97 . The same study identified a mutant in which one of these proteins was absent from the culture supernatant and discussed the possible presence of a specific secretion system for this protein. To our knowledge, this is the only possible example of a periplasmic protein being secreted in rhizobia, and it seems to be limited to a single protein. Preliminary studies in our laboratory with an equivalent mutant in R. l. bv. viciae 3841 did not reveal any difference in extracellular protein profile.

Only one Type III-like system was found in the genome sequence of R. l. bv. viciae 3841, namely the flagellar biosynthetic cluster (Table 1). Although not classically regarded as a protein secretion system, the flagellar apparatus in Salmonella has also been shown to actively secrete proteins unrelated to flagellar biosynthesis 25 , highlighting the possibility of secretion of non-flagellar proteins by the flagellar apparatus. It is not known whether R. l. bv. viciae 3841 utilises the flagellum for protein secretion as well. Three components of the flagellar apparatus were identified in the culture supernatant. All three are part of the flagellum itself, and so are directly exposed to the extracellular medium. As the flagellum is easily sheared from the bacterial surface, flagellar proteins are common components of the culture supernatant in many bacteria 74 75 76 96 . Type IV secretion systems are involved in nodulation in M. loti R7A in a host-specific manner 20 . In contrast, the virB6 mutant of R. l. bv. viciae 3841 was not affected in the nodulation process of any of the three host legumes tested in this study, showing that the putative Type IV protein secretion system in this strain is not essential for nodulation in these hosts.

Strains of Rhizobium were grown at 28°C in Tryptone-Yeast (TY) medium 98 or Y 99 medium supplemented with succinate or mannitol as indicated in the text. Solid media contained 1% agar unless stated otherwise. Antibiotics were used at the following concentrations: ampicillin 400 μg ml^-1; gentamicin 20 μg ml^-1; kanamycin μg ml^-1; spectinomycin 400 μg ml^-1; streptomycin 400 μg ml^-1. PGase and glycanase activity were assayed on plates as described by Finnie et al. 12 . Motility was assayed by observing the absence or presence of swarming on 0.2% agar TY plates.

Nodulation assays on Pisum sativum cv. Frisson and Vicia faba cv. The Sutton were done as described previously for peas by Beynon et al. 100 . Briefly, seeds were rinsed with 70% ethanol, surface-sterilised by treatment with 5% sodium hypochlorite for 10 minutes and germinated on agar medium in the dark. Vicia hirsuta seeds were scarified with sulphuric acid and germinated on agar slants as described by Knight et al. 101 .

Discontinuous sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) was used to separate denatured proteins according to molecular weight 102 . The discontinuous system consisted of a stacking gel (4% (w/v) acrylamide, pH 6.8) and a separating gel (12% (w/v) acrylamide, pH 8.8). The gels were cast and run in the vertical Biorad MiniProtean II system. Samples were prepared by adding an equal volume of 2 × SDS sample buffer and incubating for 10 min in a boiling water bath. Gels were run at 25 mA per gel using 50 mM Tris/glycine pH 8.6, 0.1% (w/v) SDS as running buffer.

Cells grown in 100 ml of liquid Y mannitol medium for four days until stationary phase were pelleted by centrifugation for 45 min at 15,000 × g at 4°C. 20% (w/v) trichloroacetic acid (TCA) was added to the supernatant to a final concentration of 5% (w/v). After 30 min incubation on ice, the supernatant proteins were pelleted by centrifugation for 1 h at 15,000 × g at 4°C. The pellet was washed twice with 80% (v/v) acetone, 50 mM Tris/HCl pH 8.0 to remove residual TCA and air-dried prior to resuspension in SDS-PAGE sample buffer. Other methods to precipitate culture supernatant such as acetone precipitation, ammonium sulphate precipitation or lyophilisation of the culture supernatant were found to be ineffective for this strain due to the presence of large amounts of exopolysaccharide that were co-precipitated by these methods.

Mutants are listed in Table 4 and all except toaD and tatC were identified by PCR from an arrayed random Tn5-insertion library using gene- and Tn5-specific primers. The locations of insertion of Tn5 were confirmed by DNA sequencing. Strains carrying an insertion in two of the genes of interest, tatC and toad, were not present in this random library and were therefore constructed by double homologous recombination. For the construction of a toaD mutant, a 2 kb fragment carrying the toaD gene was cloned into pJQ200SK 103 and a spectinomycin resistance cassette ligated into the unique NcoI site within toaD. This construct was then recombined into the genome by a double homologous recombination. The tatC mutant was generated using a similar approach to that used to inactivate toaD. PCR and DNA sequencing were used to verify that all insertions interrupted the gene of interest.

To avoid the problem of possible redundancy between the different members of Type I protein secretion systems, a prsD/toaD/tobD triple mutant and all double combinations were generated. To achieve this, the kanamycin resistance in the prsD mutant A895 was exchanged for gentamicin resistance by homologous recombination within the Tn5 sequence using plasmid pJQ175 104 , resulting in strain A898. Gentamicin resistance from this prsD mutant (A898) and spectinomycin resistance from the toaD mutant (A913) were then transduced consecutively into the kanamycin resistant tobD mutant (A896) to produce the prsD/toaD/tobD triple mutant A912. The double mutants toaD/prsD (A911) and tobD/prsD (A903) were constructed similarly. The mutants were checked by PCR and DNA sequencing. Although also part of a putative Type I secretion system, the BltD protein clearly belonged to a different class of proteins than the other three Type I ABC protein exporters. Therefore double or triple mutants protein exporter mutants carrying the bltD mutation were not made.

Nucleotide sequence data was curated and annotated using the CloneManager or Artemis software packages. Versions of BLAST 53 were used to carry out similarity searches in various databases. Sequences were aligned using ClustalX 1.8 105 , TreeView 106 was used to visualise dendrograms and phylograms. GEP signal sequences were predicted using the programs PrediSi and SignalP 3.0, TATFIND and TATP 1.0 were used for the prediction of TAT substrates. The RL and pRL numbers refer to gene identifiers used in the genome sequence of R.l. bv. viciae strain 3841 107 .

The proteins contained in the gel slices (approx. 2 × 8 mm) were submitted to tryptic digest and MALDI ToF MS fingerprinting. The MS data were queried against a database containing the predicted R. l. bv. viciae 3841 protein sequences using the MASCOT software. Proteins achieving a MOWSE score of more than 51 (representing more than 95% confidence) were regarded as positively identified. Most proteins were identified with more than 99% confidence. All MASCOT searches were carried out on the complete database without molecular weight constraints, but the theoretical molecular weights of the predicted proteins all lay within 15 kDa of the molecular weight observed on the gel. Restricting the upper molecular weight limit adds additional data to the database search parameters and would result in even higher MOWSE scores, especially for relatively small proteins. As the molecular weight of many proteins can be estimated from an SDS-PAGE gel, this would be a legitimate constraint under the assumption that there is no major posttranslational processing. The MOWSE score for any individual proteins in a protein mixture is reduced, as many of the fragments in a mixture would remain unmatched to any single protein. For this reason, proteins identified from band 12 (which comes from a region of the gel with many closely-spaced protein bands) tend to have lower MOWSE scores than proteins matched to clear single bands. Two proteins from band 12 (RL0713 and RL0489) actually achieve MOWSE scores below the statistical cutoff at 51. Considering their molecular weights and the fact that they clearly come from a mixed band, they were nevertheless included. Removing fragments from the positively identified proteins pRL100451 and RL0196 from the band 12 dataset lifts the MOWSE score above the cutoff. This is not the case for low scoring proteins in the other datasets.