Study of mural painting isolates, leading to the transfer of 'Bacillus maroccanus' and 'Bacillus carotarum' to Bacillus simplex, emended description of Bacillus simplex, re-examination of the strains previously attributed to 'Bacillus macroides' and description of Bacillus muralis sp. nov.

Gottheil (1901) described Bacillus simplex and believed that Bacillus carotarum (Koch, 1888) and Bacillus cohaerens were possibly synonyms. Bacillus capri, Bacillus cobayae and Bacillus musculi were mentioned as other possible synonyms of B. carotarum by Stapp (1920). Since the original B. carotarum strains were lost, Gibson & Gordon (1974) proposed B. capri strain NRS 607 as the representative of the species. Hunger & Claus (1981) showed that the DNADNA relatedness between B. carotarum NRS 607 and B. simplex was around 70 %, the borderline for species delineation. Priest et al. (1988) studied 368 strains, including B. carotarum and B. simplex, by 118 phenotypic tests. In the resulting clustering, B. simplex and B. carotarum strains grouped together. Despite this close relatedness and regardless of the DNADNA hybridization data of Hunger & Claus (1981), Priest et al. (1988) concluded from their study that B. carotarum and B. simplex were distinct taxa and gave a species description of B. simplex (later validated; Priest et al., 1989).

Bacillus maroccanus (Delaporte & Sasson, 1967) was included, together with B. simplex, in the phylogenetic studies of Ash et al. (1991) and Xu & Côté (2003). In both studies, B. maroccanus and B. simplex clustered closely together. However, to our knowledge no DNADNA hybridization studies between these two taxa have been published to date.

B. macroides was originally described as Lineola longa (Pringsheim & Robinow, 1947; Pringsheim, 1950), until Bennett & Canale-Parola (1965) demonstrated endospore formation and a positive Gram-stain in this ostensibly Gram-negative, non-sporulating taxon. Two different B. macroides strains are available: ATCC 12905 (= DSM 54), on which the original description is based, and NCIMB 8796 (=NCDO 1661) of which a 16S rRNA gene sequence is available (GenBank/EMBL accession number X70312). The sequence of the latter shows high similarity to sequences of B. simplex and B. maroccanus, while Xu & Côté (2003) recently demonstrated that strain ATCC 12905 shares highest similarity with Bacillus fusiformis and Bacillus sphaericus based on comparison of 3'-end 16S rRNA gene and 5'-end 16S23S ITS sequences. Since the phenotypic characters of B. macroides also conform to those of B. sphaericus, except for sporangial morphology (Claus & Berkeley, 1986), it can be concluded that ATCC 12905 is the original B. macroides strain.

The finding of isolates from deteriorated wall paintings with high 16S rRNA gene sequence similarity to B. macroides strain NCIMB 8796 (Heyrman & Swings, 2001) and consequently the need for their classification, led in this study to the revision of the taxonomic position of strains previously attributed to B. macroides, B. maroccanus and B. carotarum and to the proposal of the novel species Bacillus muralis sp. nov. to accommodate five mural painting isolates.

Strains and media.
The mural painting isolates originate from two severely damaged mural painting sites, the Servilia tomb of the Roman necropolis of Carmona (Spain) and the Lutheran church of Greene-Kreiensen (Germany). The Carmona strains were isolated from a sample of powdery growth on the painted wall at the left of the main entrance of the tomb, at a height of 1·8 m. The Greene strains originate from sampling of ochre-coloured, round, dry spots taken at the west wall of the church (depicting the Adoration of the Magi). Further isolate designations and methods applied are shown in Table 1. Isolates were maintained in Microbank tubes at 80 °C. For phenotypic characterization studies, including microscopy, cultures were maintained on slopes of trypticase soy agar (TSA) containing 5 mg MnSO₄ l¹ (to enhance sporulation).

Table 1. Strains studied and the methods used for characterization

DNA preparation.
Total genomic DNA was purified for 16S rRNA gene sequencing, ARDRA and rep-PCR using a slight modification of the method of Pitcher et al. (1989), as described by Heyndrickx et al. (1996). For determination of the G+C content and DNADNA hybridization, approximately 1 g biomass was harvested from agar plates. DNA was purified by a combination of the protocols of Marmur (1961) and Pitcher et al. (1989), as described by Logan et al. (2000).

16S rRNA gene sequencing and phylogenetic analysis.
Sequence analysis was performed as previously described by Heyrman & Swings (2001). For partial sequencing, two primers were used (reverse 358339 and reverse 536519; Heyrman & Swings, 2001) to obtain the first 400500 base pairs of the 16S rRNA gene, which according to Goto et al. (2000) includes the hypervariant region for the genus Bacillus. A phylogenetic tree was constructed using the BioNumerics 3.0 software (Applied Maths) by applying the neighbour-joining method of Saitou & Nei (1987) on a multiple alignment similarity matrix. The stability of relationships was assessed by a bootstrap analysis of 1000 datasets.

ARDRA.
Amplified 16S rRNA genes were obtained by PCR and analysed by restriction digestion with five restriction enzymes (HaeIII, DpnII, RsaI, BfaI and Tru9I) as described previously by Heyndrickx et al. (1996). The restriction patterns were analysed as described by Heyndrickx et al. (1998).

rep-PCR genomic fingerprinting.
PCR was performed with the (GTG)₅ primer (Versalovic et al., 1994) using the PCR conditions described previously by Rademaker & de Bruijn (1997). Electrophoretic conditions for the (GTG)₅-PCR of the mural painting isolates (see Fig. 1a) were those described by Heyrman et al. (2003); those for the additional (GTG)₅-PCR (see Fig. 1b) were described by Scheldeman et al. (2004). The resulting patterns were digitized and a Pearson correlation of the resulting band patterns was performed using the BioNumerics and the GelCompar software, respectively.

(42K): Fig. 1. (a) Grouping of normalized (GTG)5 patterns (using the electrophoretic conditions described by Heyrman et al., 2003) of 24 mural painting isolates. (b) Grouping of normalized (GTG)5 patterns (using the electrophoretic conditions described by Scheldeman et al., 2004) of representative mural painting isolates together with the B. simplex type strain, B. macroides LMG 18474 (=DSM 54=ATCC 12905), B. maroccanus, strains belonging to B. carotarum and additional isolates R-15936 and R-15943. Both dendrograms are based on a UPGMA clustering of Pearson's correlation similarity coefficients.

G+C content and percentage DNADNA binding.
The G+C content of DNA was determined by HPLC (Mesbah et al., 1989) using the further specifications given by Logan et al. (2000). DNADNA hybridization was performed using a modification of the microplate method of Ezaki et al. (1989), as described by Willems et al. (2001). A hybridization temperature of 37 °C (calculated with correction for the presence of 50 % formamide) was used.

Chemotaxonomic characterization.
Gas chromatographic analysis of fatty acid methyl esters was performed starting from strains grown on TSA for 24 h at 28 °C. A quantitative analysis of cellular fatty acid compositions was further performed by the gas-liquid chromatographic procedure as previously described (Mergaert et al., 1993). Computer analysis of the resulting profiles was performed as described by Heyrman et al. (1999). For SDS-PAGE of whole-cell proteins, cells were obtained as described by Heyndrickx et al. (1998). SDS protein extracts were prepared and electrophoresed according to Pot et al. (1994), and the data were collected and interpreted as described by Vauterin & Vauterin (1992).

Phenotypic characterization.
The strains were phenotypically characterized by the methods of Logan & Berkeley (1984); other characters were determined, and the data numerically analysed, as described by Logan et al. (2000). For strains that give very few positive reactions in these phenotypic tests, the inclusion of negative results in similarity calculations tends to overemphasize the similarities of these organisms; data for unreactive strains in this study were therefore analysed by omitting negative matching characters from the calculations. Vegetative cells and sporangia were observed by phase-contrast microscopy. Strains were also examined for catalase and oxidase production and casein and starch hydrolysis using the methods of Gordon et al. (1973). Maximum and minimum growth temperatures were determined by incubating 10 ml TSA cultures in a water bath; pH ranges for growth were determined using 10 ml TSA cultures adjusted to pH 5·0, 7·0 and 9·0; both series were examined for turbidity at 24 h intervals. Anaerobic growth was tested for by incubating cultures on TSA plates in a GasPak jar, with aerobically incubated plates as controls. A small number of strains were also tested for assimilation of substrates as sole carbon sources, using the API Biotype 100 kit as described by Heyndrickx et al. (1997).

Genomic profiling and 16S rRNA gene sequencing
Firstly, (GTG)₅-PCR fingerprinting was performed on the mural painting isolates to determine their genotypic diversity and to select representative strains for further analyses. The resulting electrophoretic patterns were used to construct a dendrogram based on a Pearson correlation (Fig. 1a). In this dendrogram, four clusters can be distinguished with an internal correlation above 65 %. These clusters of strains are called (GTG)₅-groups A to D. The mural painting strains had clustered according to geographical origin, with the Carmona (Spain) strains divided between (GTG)₅-groups B and C, and the Greene (Germany) strains between groups A and D.

Six representative strains were chosen for partial sequencing (463 bases) of the hypervariable region for Bacillus of the 16S rRNA gene (Goto et al., 2000): one of (GTG)₅-groups A and B (LMG 21002 and LMG 22045, respectively) and two of groups C and D (LMG 19489 and LMG 22046; LMG 21004 and LMG 20238^T, respectively). The strains of groups A, B and C, had identical partial sequences, while the two strains of group D were almost identical to each other (one extra base in the sequence of LMG 20238^T). Comparison of LMG 21002 (group A) and LMG 20238^T (group D) revealed four base mismatches; these two strains and LMG 19489 (group C) were chosen for determination of the nearly complete 16S rRNA gene sequence. LMG 21002 showed respective sequence similarities of 99·7 and 99·9 % to LMG 19489 and LMG 20238^T, while the latter two strains had a sequence similarity of 99·5 %. A FASTA search (Pearson & Lipman, 1988) of the analysed sequences showed 99 % sequence similarity or higher for all three to sequences of B. simplex DSM 1321^T (AJ439078), B. macroides NCIMB 8796 (X70312), two sequences (AY030319 and AF157696) attributed to B. macroides on the basis of comparative analysis with X70312, and some additional unnamed strains (Fig. 2). The 16S rRNA gene sequences of B. simplex DSM 1321^T (X60638) and B. maroccanus NCIMB 10500 (X60626) analysed by Ash et al. (1991) could not be retrieved from the FASTA results, because these sequences contain large numbers of unknown bases (61 and 25, respectively). When their unknown bases were discarded, high sequence similarities (>99·5 %) to the sequences deposited as AJ439078 (B. simplex) and X70312 (B. macroides) were also demonstrated. Since the sequence AJ439078 for B. simplex was only recently deposited in the international sequence databases, several strains were attributed previously to B. macroides because of their high sequence similarity to strain NCIMB 8796 (X70312). However, B. macroides NCIMB 8796 differs importantly from ATCC 12905, the strain on which the original description of B. macroides was based. Xu & Côté (2003) compared the 3'-end 16S rRNA gene and 5'-end 16S23S ITS nucleotide sequences of 40 species of the Bacillaceae and showed that B. macroides ATCC 12905 showed highest similarity to B. sphaericus and B. fusiformis. In the present study, the nearly complete 16S rRNA gene sequence of strain LMG 18474 (=DSM 54=ATCC 12905) was determined and B. sphaericus and B. fusiformis were indeed retrieved as the closest related species on the basis of the respective sequence similarities of 98·4 and 98·7 %. In the original description, the phenotypic characters of B. macroides were similar to those of B. sphaericus, except for sporangial morphology (Claus & Berkeley, 1986). This led to the conclusion that strain NCIMB 8796 is taxonomically different from B. macroides as represented by LMG 18474 (=DSM 54=ATCC 12905). In a phylogenetic tree, strain NCIMB 8796 clustered closely with strains belonging to B. simplex and strains that were named B. maroccanus. In this tree, three sequences are present that are all presumed to belong to the B. simplex DSM 1321^T, although one of them (D78478) is clearly distinct from the other two. To assess which sequence was erroneously linked with the type strain of this species, we also determined the 16S rRNA gene sequence of B. simplex LMG 11160^T (AJ628743). The sequence with accession number D78478 remained separate and was therefore regarded as aberrant; the presence of such sequences in the databases that are erroneously attributed to type strains results in faulty interpretations. No 16S rRNA gene sequences of named B. carotarum strains were available in the international databases. Since a close relatedness could be expected from the literature between named B. carotarum strains and B. simplex (see Introduction), 16S rRNA gene sequences of LMG 17633 and LMG 17636 were determined. They were identical and showed 99·7 % sequence similarity to B. simplex (AJ439078 and AJ628743), corresponding to 4 base mismatches, confirming the close relatedness between B. carotarum and B. simplex.

(35K): Fig. 2. Unrooted phylogenetic tree derived from the analysis of 16S rRNA gene sequences of mural painting isolates (in bold) together with strains that were attributed to B. simplex and the species B. macroides, B. maroccanus and B. carotarum.

As a result of this sequence analysis that was supported by other data (see Introduction), a separate (GTG)₅-PCR fingerprinting experiment (using different electrophoretic conditions, see Methods) was performed on 14 representative mural painting isolates, the B. simplex type strain, strains previously attributed to B. macroides (strain DSM 54), B. maroccanus (LMG 12364) and B. carotarum and two additional isolates, R-15936 and R-15943 (Fig. 1b). Numerical analysis of this second set of (GTG)₅ data revealed that most strains named B. carotarum, the B. maroccanus strain and the mural painting isolates attributed to (GTG)₅-group A formed a tight cluster with a Pearson correlation above 70 %. Strains of (GTG)₅-group B clustered at more than 80 % correlation and were joined by clinical isolate R-15943 at 60 %. B. simplex LMG 11160^T, B. carotarum LMG 17638, B. macroides LMG 18474 (=DSM 54) and (GTG)₅-groups C and D clustered separately (<50 % relatedness to other profiles).

In addition, the named B. carotarum strains, B. macroides strains LMG 18474 and LMG 18508 and B. maroccanus were analysed together with B. simplex by ARDRA (see Supplementary Fig. A available in IJSEM Online). A cluster at 90 % or more pattern similarity was obtained for all B. carotarum strains, B. maroccanus LMG 12364 and B. macroides LMG 18508 (=NCIMB 8796) that was joined by the B. simplex type at 88 % similarity. The pattern of the named B. macroides strain LMG 18474 (=DSM 54) was only 70 % similar to that of the other strains.

The results above showed that the mural painting strains belonging to (GTG)₅-group A grouped together with B. carotarum and B. maroccanus. Furthermore, these strains were also similar to the type strain of B. simplex, B. macroides NCIMB 8796 and the other mural painting strains. However, B. macroides strains LMG 18474 (=DSM 54=ATCC 12905) and LMG 18508 (=NCIMB 8796) clearly belong to separate taxa. DNADNA hybridization experiments were performed to determine the exact taxonomic position of the investigated strains.

DNADNA hybridization and DNA base composition
On the basis of rep-PCR fingerprinting, 16S rRNA gene sequencing and ARDRA, strains were selected for DNADNA relatedness studies. The results (Supplementary Table A) show that the mural painting strains belonging to (GTG)₅-groups A, B and C, B. maroccanus and B. carotarum belong to B. simplex (relatedness values>70 %), according to the current species definition (Stackebrandt et al., 2002). For B. carotarum, these data are in accordance with those of Hunger & Claus (1981), who performed a DNADNA hybridization experiment between NRS 607, proposed as a representative strain of B. carotarum by Gibson & Gordon (1974), and B. simplex NRS 960^T and found a DNADNA relatedness value of 71 %.

Strain LMG 20238^T, representative of (GTG)₅-group D, showed less than 50 % DNADNA relatedness to B. simplex, B. maroccanus and the other mural painting representatives, and thus represents a novel genospecies. In the text, these strains, i.e. the (GTG)₅-group D strains, will be referred to as Bacillus muralis. As predicted from the previous analyses, B. macroides strains LMG 18474 and LMG 18508 do not belong to the same species (DNADNA relatedness below 10 %). Strain LMG 18508 (=NCIMB 8796), of which a nearly complete 16S rRNA gene sequence is available (accession no. X70312), can be attributed to B. simplex (mean DNADNA relatedness value of 73·4 %).

In this study, the G+C content of the B. simplex type strain (LMG 11160^T) was in the range of 39·541·6 mol%. All tested strains that were attributed to B. simplex by DNADNA hybridization study have a G+C content that falls within this range (Supplementary Table A). The G+C content of LMG 20238^T (B. muralis) is similar to that of B. simplex (40·541·8 mol%), while the G+C content of B. macroides LMG 18474 is slightly lower (37·638·9 mol%).

Chemotaxonomic characteristics
Fatty acid analysis was performed on the mural painting isolates, B. simplex LMG 11160^T, B. macroides strains LMG 18508 (=NCIMB 8796) and LMG 18474 (=DSM 54=ATCC 12905), B. maroccanus LMG 12364, B. carotarum LMG 17634 and strains R-15936 (clinical isolate) and R-15943 (Antarctic soil isolate). In Fig. 3 the resulting profiles are grouped together with the MIDI database reference profiles of B. simplex and B. macroides. The outgroup (at a Euclidean distance of about 86 %) is formed by B. macroides LMG 18474 and the database reference profile of B. macroides, which is in accordance with the earlier statement that strain LMG 18474 (=DSM 54=ATCC 12905) is the original B. macroides strain, while strain LMG 18508 (=NCIMB 8796) must be allocated to B. simplex. The remaining strains form a group that is divided into two subgroups at a Euclidean distance of 95 %. The first subgroup contains the mural painting isolates of (GTG)₅-groups B and D (B. muralis) together with profiles of the B. simplex type strain and of the atypical B. macroides strain, LMG 18508 (=NCIMB 8796). The second contains the mural painting isolates of (GTG)₅-groups A and C, two profiles of B. maroccanus, B. carotarum LMG 17634, R-15936, R-15943 and the MIDI reference profile of B. simplex. Except for R-15943, which showed the highest similarity to the database profile of B. simplex, the strains of the second subgroup were identified by the TSBA 4.0 MIDI database as Bacillus megaterium with similarities ranging from 0·48 to 0·89. Kämpfer (1994) already noted the close similarity of the fatty acid profiles of B. simplex and B. megaterium. The mural painting isolates that belong to the novel species B. muralis form a tight distinct FAME cluster. However, this cluster is embedded in a heterogeneous larger FAME group composed of strains that may all be attributed to B. simplex on the basis of genotypic analyses. It is thus not straightforward to distinguish B. muralis from B. simplex on the basis of fatty acid analysis, although small differences are clear: in comparison, B. muralis strains contain smaller amounts of the fatty acid anteiso-C_{15 : 0} and higher amounts of iso-C_{15 : 0} and C_{16 : 1}ω11c (Table 2).

(34K): Fig. 3. Simplified dendrogram obtained by UPGMA clustering of Euclidean distances calculated between fatty acid profiles of mural painting isolates, additional isolates R- 15936 and R-15943, B. simplex and named B. macroides, B. maroccanus and B. carotarum strains. For B. simplex and B. macroides, the MIDI reference profiles were also included.

Table 2. Comparison of the mean fatty acid profiles of the mural painting isolates and B. simplex B. simplex profiles obtained in this study and the MIDI reference are shown. The data given are mean percentages of total fatty acids (±SD if applicable). Only fatty acids accounting for at least 1·0 % of the total fatty acid content are listed. Strains belonging to (GTG)5-groups A and C are listed together since they have similar profiles (see Fig. 3).

Numerical analysis of SDS-PAGE patterns of whole-cell proteins (Fig. 4) revealed that, as expected from previous results, the B. simplex type strain, B. maroccanus, B. carotarum and representative mural painting isolates showed very similar patterns (similarity>90 %). B. macroides LMG 18474 (=DSM 54=ATCC 12905) showed a distinct pattern, which confirms our earlier findings. Again, the strains of the novel species B. muralis group together within the larger cluster of strains that must be attributed to B. simplex according to our polyphasic approach.

(70K): Fig. 4. Normalized computer profiles of protein patterns (PAGE) of mural painting isolates, additional isolates R-15936 and R-15943, B. simplex and named B. macroides (strain LMG 18474=DSM 54=ATCC 12905), B. maroccanus and B. carotarum strains. The group containing strains belonging to the novel species B. muralis is presented as a shaded triangle.

Phenotypic characterization
The strains listed in Table 1 were investigated by the API 20E and API 50CHB system (bioMérieux). The phenotypic patterns obtained were compared along with the profiles of other Bacillus species, using Gower's coefficient (S_G) and omitting negative matches. In this grouping (not shown), the strains were divided between three clusters, which will be called biovars 1 to 3 (biovar 1 comprised 14 strains, biovar 2 comprised 21 strains and biovar 3 comprised 5 strains; biovars 2 and 3 each possessed a single atypical strain; Table 1). Biovar 1 strains included the type strain of B. simplex (LMG 11160^T) and B. macroides LMG 18508 (=NCIMB 8796) and they differed from biovar 2 strains by generally weaker acid production from a slightly narrower range of carbohydrates (as listed in the emended description given below). Biovar 1 strains merged at only 70 % S_G and this rather heterogeneous cluster encompassed a strain of Bacillus firmus (B0823=NRS 854=R-15586). Strains attributed to biovar 2 formed a very loose cluster that merged at only 65 % S_G, with one atypical strain (LMG 17633) outlying at 60 % S_G. This cluster contained B. maroccanus LMG 12364. The mural painting strains belonging to (GTG)₅-groups A, B and C and the B. carotarum strains were distributed between both biovars 1 and 2. B. macroides strain LMG 18474 (=DSM 54=ATCC 12905), which is believed to represent the authentic B. macroides, was unreactive in the API 50CHB and 20E tests. These results again confirm that the two strains received as B. macroides (the unreactive LMG 18474, and LMG 18508, which was reactive in the phenotypic tests) do not belong to the same species. When the groupings based on API 50CHB and 20E test patterns were compared with those obtained by other analyses [(GTG)₅, ARDRA, SDS-PAGE, FAME], biovars 1 and 2 were not distinguishable, but appeared scattered among the groups that had been recognized by those other methods.

Strains belonging to biovar 3 showed generally stronger acid production reactions over a wider range of substrates in comparison with strains of biovars 1 and 2. Biovar 3 contained only strains belonging to (GTG)₅-group D (B. muralis), but B. muralis strain R-8204 was an outlier of this cluster. When positive and negative reactions were given equal weights in the similarity calculations, B. muralis strains LMG 20238^T, LMG 21004, R-8210 and R-8251 merged at 90 % S_G, and this group was joined by R-8204 at 85 %, but strains of Bacillus shackletonii and B. simplex (LMG 17636 and LMG 17643) also joined at this level. However, when individual API tests are considered, B. muralis R-8204 can be differentiated from B. simplex strains (including LMG 17636 and LMG 17643) by acid production from D-galactose, lactose, D-mannose and melibiose. In addition, all of the B. muralis strains can be separated from B. simplex on the basis of the ONPG test and the colour of the diffusible pigment produced on skimmed milk agar (brown for B. simplex strains and salmon-orange for B. muralis strains), and from B. shackletonii on the basis of several API tests and some non-API characters (Table 3). The differentiation of B. simplex and B. muralis from other related Bacillus species can also be achieved using the characters shown in Table 3.

Table 3. Characters that distinguish between B. simplex (divided into biovars 1 and 2), B. muralis sp. nov. and some phenotypically similar and phylogenetically related Bacillus species Reference species: 1, B. megaterium; 2, B. flexus (morphological characters, starch and casein hydrolysis data taken from Priest et al., 1988); 3, B. shackletonii; 4, B. firmus (starch hydrolysis data from Priest et al., 1988); 5, B. sporothermodurans; 6, B. oleronius. Symbols: +, >85 % positive; +/W, positive or weakly positive; W, weakly positive; V, variable (2674 % positive); V/W, variable and when positive weak; , 015 % positive. With the exception of microscopic observations, anaerobic growth, starch and casein hydrolysis, all characters were determined using tests in API 20E and 50CHB.

To investigate further the differentiation of B. muralis from B. simplex, two biovar 3 strains (LMG 20238^T and R-8204) and six biovar 1 and 2 strains (LMG 11160^T, LMG 17633, LMG 17636, LMG 17644, LMG 17646 and R-8208) were characterized using the Biotype 100 carbon source assimilation kit (bioMérieux). The patterns of substrate utilization of the tested B. simplex biovar 1 and 2 strains were similar, but they were quite distinct from those of the biovar 3 (B. muralis) strains. B. muralis could be differentiated from B. simplex by assimilation of D-galactose, α-lactose, D-mannose, α-D-melibiose and D-turanose, and failure to assimilate L-aspartate, caprylate, fumarate, gentisate, m-hydroxybenzoate, L-malate, mucate, D-saccharate and succinate. The authentic B. macroides strain LMG 18474 (=DSM 54=ATCC 12905) was unable to grow on most of the carbon sources. This again confirms its distinctness from B. simplex and B. muralis.

Taxonomy
The possible synonymy of the species B. carotarum (Koch, 1888) and B. simplex (Gottheil, 1901) was already recognized in the original description of the latter species. Several other possible synonyms were also mentioned in the early literature for B. carotarum: B. cohaerens (Gottheil, 1901), B. capri, B. cobayae and B. musculi (Stapp, 1920). Since the original B. carotarum strains were lost, Gibson & Gordon (1974) proposed B. capri strain NRS 607 as the representative strain of this species. For some time, B. carotarum and its synonyms were thought to belong to B. megaterium (Smith et al., 1946). B. megaterium was considered to consist of two main groups of strains with similar properties as well as some intermediate strains (Gordon et al., 1973). One main group of strains corresponded closely to the original description of B. megaterium (de Bary, 1884), while the other included strains of B. capri (including NRS 607), B. cobayae, B. simplex and Bacillus teres. Later, Hunger & Claus (1981) showed that the two main B. megaterium groups differed in G+C content and that the DNADNA relatedness between the groups was well below 70 %. Within the second group, these authors found DNADNA relatedness values of 70 % or higher between B. capri (strain NRS 607), B. cobayae and B. musculi. Between B. simplex strains and B. capri values of 6771 % were obtained. B. teres gave 57 and 64 % DNADNA relatedness values with B. capri and B. musculi, respectively. Thus, Hunger & Claus (1981) showed that B. carotarum (represented by B. capri) could be attributed to a single genospecies together with B. simplex, B. cobayae and B. musculi. Further, it was shown that this genospecies differed from B. megaterium. However, the authors did not propose taxonomic changes in their paper, but rather stated that more DNA pairing studies had to be performed. Priest et al. (1988) studied 368 strains, at that time attributed to Bacillus, by 118 phenotypic tests. This study included strains named as B. carotarum, B. simplex and B. teres. In the resulting clustering, B. simplex and B. teres grouped together and were closely joined by B. carotarum strains. Despite this close relatedness and regardless of the DNADNA hybridization data of Hunger & Claus (1981), Priest et al. (1988) concluded from their study that B. carotarum and B. simplex were distinct taxa and gave a species description of B. simplex (including the B. teres strain). The name B. simplex was later validated (Priest et al., 1989).

In the present study B. carotarum strain NRS 607 (=LMG 17637) was included together with other named B. carotarum strains and the B. simplex type strain. The data led to the conclusion that strains belonging to B. carotarum and its suggested synonyms B. capri, B. cobayae and B. musculi must be allocated to B. simplex. In addition, strains belonging to B. maroccanus and B. macroides NCIMB 8796 (=NCDO 1661=LMG 18508) should be reclassified as B. simplex, as could already be suspected from their close phylogenetic relatedness (Ash et al., 1991; Xu & Côté, 2003). Another B. macroides strain, ATCC 12905 (=DSM 54=LMG 18474), that can most probably be regarded as the original representative of this species, shares highest 3'-end 16S rRNA gene and 5'-end 16S23S ITS sequence similarity with B. fusiformis and B. sphaericus (Xu & Côté, 2003). Whether this B. macroides representative deserves separate species rank is still open to investigation. Most mural painting isolates investigated in this study, as well as isolates R-15936 and R-15943, were also attributed to B. simplex. In addition, a group of five mural painting isolates, recognized on the basis of genotypic and phenotypic analyses, is proposed as the novel species Bacillus muralis and a description follows.

Description of Bacillus muralis sp. nov.
Bacillus muralis (mu.ra'lis. L. masc. adj. muralis pertaining or belonging to walls).

Cells are rod-shaped, 0·70·8 µm in diameter, and occur singly, in pairs or chains. They are straight with round or occasionally slightly tapered ends. They are motile when occurring singly but non-motile when in chains. Endospores are ellipsoidal, lie centrally or paracentrally and do not swell the sporangia or swell them only slightly. Gram reaction is variable. Colonies grown for 2 days on nutrient agar at 30 °C are 36 mm in diameter, butyrous, pink-pigmented, glossy, with irregular margins, slightly raised and umbonate. Growth under anaerobic conditions on nutrient agar is weak. Optimal growth at 30 °C, weak growth at 20 °C and no growth at 45 °C. Strains grow at pH 9 less profusely than at pH 7 and growth at pH 5 is negative. Casein hydrolysis is variable and a diffusible salmon-pink tint appears in the medium. Starch is hydrolysed. Strains tolerate 7 % NaCl, but grow weakly. Oxidase and catalase reactions are positive.

In the API 20E strip, the ONPG reaction is positive. Arginine dihydrolase, lysine decarboxylase and ornithine decarboxylase reactions are negative, citrate is not utilized, hydrogen sulphide and indole are not produced, urease production and the VogesProskauer reaction are negative and gelatin is not hydrolysed. Nitrate is reduced to nitrite. In the API 50 CHB gallery, hydrolysis of aesculin is positive and acid is produced (sometimes weakly)without gas from the following carbohydrates: L-arabinose, arbutin, cellobiose, fructose, galactose, N-acetylglucosamine, glucose, glycerol, lactose, maltose, mannitol, mannose, melibiose, raffinose, ribose, trehalose and salicin. Results are variable for acid production from sucrose. Acid is not produced from the following carbohydrates: adonitol, amygdalin, D-arabinose, D- and L-arabitol, dulcitol, erythritol, D- and L-fucose, gentiobiose, gluconate, glycogen, inositol, inulin, 2- and 5-ketogluconate, D-lyxose, melezitose, methyl α-D-glucoside, methyl α-D-mannoside, methyl β-D-xyloside, rhamnose, starch, sorbitol, sorbose, D-tagatose, D-turanose, D- and L-xylose and xylitol.

The major cellular fatty acids are anteiso-C_{15 : 0} and iso-C_{15 : 0}, present at about 43 and 23 % of the total fatty acids, respectively. The following fatty acids are present at 1 %: iso-C_{14 : 0}, C_{14 : 0}, C_{15 : 0}, C_{16 : 1}ω7c alcohol, iso-C_{16 : 0}, C_{16 : 1}ω11c and C_{16 : 0}. The G+C content is 41·2 mol% for the type strain.

The type strain, strain LMG 20238^T (=DSM 16288^T), was isolated from mural paintings in the Lutheran church of Greene-Kreiensen, Germany. In the characters for which a variable result was obtained, the type strain does not hydrolyse casein and does not produce acid from sucrose. In the Biotype 100 kit, the type strain hydrolyses aesculin and is able to grow on the following substrates as sole carbon sources: cis- and trans-aconitate, D- and L-alanine, 4-aminobutyrate, 5-aminovalerate, L-arabinose, L-arabitol, D-cellobiose, citrate, β-D-fructose, D-galactose, β-gentiobiose, D-gluconate, D-glucosamine, α-D-glucose, glycerol, L-histidine, 3-hydroxybutyrate, 2-oxoglutarate, DL-lactate, α-lactose, maltose, maltotriose, D-mannitol, D-mannose, α-D-melibiose, methyl α-D-glucopyranoside, N-acetyl-D-glucosamine, palatinose, phenylacetate, L-proline, putrescine, D-raffinose, D-ribose, L-serine, sucrose, D-trehalose, L-tryptophan, D-turanose, L-tyrosine and D-xylose.

Emended description of Bacillus simplex (ex Meyer and Gottheil 1901) Priest et al. 1989
Bacillus simplex (sim'plex. L. masc. adj. simplex simple).

Rods are straight, 0·70·9 µm in diameter, round-ended or occasionally slightly tapered and occur in chains and sometimes singly or in pairs. Motile when occurring singly but non-motile when growing in chains. Endospores are ellipsoidal, occasionally spherical, lie centrally, paracentrally or subterminally, and do not obviously swell the sporangia. Gram reaction is variable amongst cells. Colonies, grown for 2 days on nutrient agar at 30 °C, are 36 mm in diameter, cream-coloured, glossy, with irregular margins, slightly raised and umbonate. Most strains are strictly aerobic, although some strains may grow weakly on nutrient agar under anaerobic conditions. They grow at 20 and 30 °C but are not able to grow at 45 °C. Strains grow at pH 9 as profusely as at pH 7 and growth at pH 5 is variable. Casein hydrolysis is variable and a brown tint, of varying strength, diffuses into the medium. Starch is hydrolysed. Tolerance to 5 % NaCl (w/v) is variable and no growth occurs in media supplemented with 7 % NaCl (w/v). Oxidase reaction is negative and catalase is positive. In the API 20E strip, the ONPG, arginine dihydrolase, lysine decarboxylase and ornithine decarboxylase reactions are negative, citrate is not utilized, hydrogen sulphide and indole are not produced, urease production and the VogesProskauer reaction are negative and gelatin hydrolysis is variable. Nitrate is reduced to nitrite. In the API 50 CHB gallery, hydrolysis of aesculin is variable and weak. For all strains tested, acid is produced, mostly weakly, without gas from the following carbohydrates: fructose, N-acetylglucosamine, glucose, inulin, trehalose and sucrose. Strains produce acid weakly and variably from salicin. Acid production is always negative from adonitol, amygdalin, D-arabinose, D- and L-arabitol, arbutin, dulcitol, erythritol, D- and L-fucose, galactose, gentiobiose, gluconate, 2- and 5-keto-D-gluconate, glycogen, lactose, D-lyxose, mannose, melezitose, melibiose, methyl α-D-glucoside, methyl α-D-mannoside, methyl β-D-xyloside, rhamnose, starch, sorbose, D-tagatose, D-turanose, L-xylose and xylitol. On the basis of reaction patterns for other substrates, two biovars may be recognized: strains belonging to biovar 1 produce acid weakly and variably from L-arabinose, mannitol, raffinose, ribose and sorbitol, while acid production is always negative from cellobiose, glycerol, inositol, maltose and D-xylose; strains of biovar 2 produce acid, weakly, without gas from the following carbohydrates: L-arabinose, mannitol, raffinose, ribose, sorbitol and D-xylose; biovar 2 strains are variable for weak acid production from cellobiose, glycerol, inositol and maltose.

The major cellular fatty acids are anteiso-C_{15 : 0} and iso-C_{15 : 0}, present (means±SD) at 59·03±5·88 and 15·55±2·95 % of total fatty acids, respectively. The following fatty acids are present (mean values) at 1 %: iso-C_{14 : 0} (5·94±2·82 %), C_{14 : 0} (1·56±0·59 %), C_{16 : 1}ω7c alcohol (2·97±0·74 %), iso-C_{16 : 0} (2·26±1·42 %), C_{16 : 1}ω11c (4·80±1·30 %), C_{16 : 0} (2·36±1·58 %), iso-C_{17 : 1} I and/or anteiso-C_{17 : 1} B (1·12±0·82 %), and anteiso-C_{17 : 0} (1·82±0·80 %). The G+C content is in the range of 39·541·8 mol%.

The type strain is strain NRS 960^T (=ATCC 49097^T=DSM 1321^T=LMG 11160^T). For the variable characters in the API kits, the results for the type strain are as follows: weak or moderate acid production for L-arabinose, mannitol, ribose and sorbitol. No acid is produced from raffinose. In the Biotype 100 kit, the type strain hydrolyses aesculin very weakly and is able to grow on the following substrates as sole carbon sources: cis- and trans-aconitate, D- and L-alanine, 4-aminobutyrate, 5-aminovalerate, L-aspartate, benzoate, caprylate, citrate, m-coumarate, ethanolamine, β-D-fructose, fumarate, gentisate, D-gluconate, D-glucosamine, α-D-glucose, D-glucuronate, L-glutamate, glutarate, DL-glycerate, glycerol, L-histidine, m-hydroxybenzoate, 3-hydroxybutyrate, myo-inositol, 2-keto-D-gluconate, DL-lactate, L-malate, maltose, maltotriose, D-mannitol, methyl α-galactopyranoside, mucate, N-acetyl-D-glucosamine, palatinose, phenylacetate, 3-phenylpropionate, L-proline, propionate, protocatechuate, putrescine, quinate, α-L-rhamnose, D-ribose, D-saccharate, L-serine, D-sorbitol, succinate, L-tartrate, D-trehalose, tricarballylate, L-tryptophan, L-tyrosine and D-xylose.

The authors acknowledge financial support from the European Commission (grant ENV4-CT98-0705). We are most grateful to bioMérieux for providing API materials and for supporting M. R.-D. The authors are grateful to Mrs Sandra Cots and Mrs Elly Engels for excellent technical assistance.