One can ask whether the sizes of proteins of a psychrophile differ from those of a mesophile. The sequences of all proteins of P. ingrahamii were compared to sequences of all proteins of three other bacteria: Shewanella oneidensis MR-1, Vibrio cholerae and Escherichia coli K-12 MG1655. 916 protein sequences were conserved among all four bacteria. The great majority of the conserved proteins were enzymes. The distribution of lengths of the 916 orthologous proteins were compared, revealing that the distribution was about the same for the comparable proteins in all four bacteria (Figure 1). Ideas that proteins that are required to function at low temperatures would be either shorter or longer than those of mesophiles are not borne out.

Amino acid composition of an organism's proteins is affected by nucleotide composition of the DNA. The GC content of P. ingrahamii DNA was determined experimentally to be 40% 8, verified by the composition of the total genome nucleotide sequence (40.1%). Overall amino acid content of the encoded proteins compared to those of V. cholerae, S. oneidensis and E. coli is shown in Table 1 (P. Sharp, personal comunication). The amino acids isoleucine, asparagine, lysine, phenylalanine and tyrosine all are present in higher percentage in P. ingrahamii.

However, like the genome, the codons for these residues are GC rich, a factor that must be taken into account. Codon usage for P. ingrahamii is shown in Table 2 for all CDSs (coding DNA sequences) and for the highly expressed genes tuf, tsf, fusA+RP genes (P. Sharp, personal communication). Thus when amino acid content is examined as a function of GC3s, correcting for GC content at synonymously variable third positions, the overall amino acid composition of P. ingrahamii was found not to be remarkable. At the value of 34.2 determined for P. ingrahamii, amino acid contents fall on the curves generated for GC3s dependency in 80 other organisms (P. Sharp, personal communication) (data not shown) 9.

Lobry and Chessel 10 examined datasets of amino acid content and codon usage in thermophilic and mesophic bacteria and found evidence that the amino acid composition of thermophilic proteins was under the control of a pressure at the nucleic acid level, not a selection at the protein level. An extended study 11 produced similar results of no connection. However, the authors identified the most discriminating codon as being AGG for arginine, present in many thermophiles, not in either mesophiles or a psychrophile. P. ingrahamii fits this observation in that it does not use the AGG codon (Table 2).

We analyzed the amino acid compositions of all individual proteins of P. ingrahamii by CA 12 after trimming the first 10 and last 5 residues because there is a strong nucleotide bias at these positions. Data for each protein is presented in Additional file 1: "Ping Correspondence Analysis.xls". Clustering of groups of proteins with similar composition was performed using a bayesian approach as proposed by Bailly-Bechet et al. 13 Figure 2 presents a plot of the first three most informative axes: hydrophobicity, aromaticity and asparagine content.

The proteins fell into six classes whereas most bacteria contain at most five well separated clusters with bayesian clustering and at most four with dynamic clouds clustering 14. Proteins of two bacterial proteomes examined to date fall into four clusters, Aeropyrum pernix and Thermoplasma acidophilum 14. Subsequently it was seen that psychrophilic P. haloplanktis has 5 clusters 7, exceeded now by P. ingrahamii, a more extreme psychrophile, which resolves into six. In P. ingrahamii, the bulk of the proteins fall into three clusters (clusters 1, 2, 6). The integral inner membrane proteins (IIMPs) fall largely into clusters 3 and 5 (Figure 2). In other bacteria the IIMPs are distinctly separate from the bulk of other proteins 14. In P. ingrahamii, however, separation is poor and there is some continuity between the IIMPs and the bulk proteins. Evidently the property of hydrophobicity is not distributed in psychrophile proteins as in mesophiles.

Not seen in any other bacteria viewed to date by CA is the sixth group, cluster 4. This group is of 57 proteins characterized by an excess of threonine that forms a bleb protruding from one of the core protein clusters. Examining the annotations of these proteins reveals that almost half are hypothetical proteins with no homologs in other bacteria examined at threshold Pam 150. It is tempting to suggest this group of proteins could be involved in facilitating low temperature growth.

As to amino acid content, one sees that asparagine occurs more frequently than expected for a random distribution, and on the other hand that cysteine, methionine, arginine and histidine are relatively rare (Figure 2). This same asparagine-driven bias has been seen in other psychrophiles, notably P. haloplanktis 7. Deamidation via cyclization into aspartate threatens integrity of asparagine, a process which is sensitive to higher temperature, providing a rationale for an asparagine excess in psychrophiles. The corresponding lower amounts of cysteine, methionine, arginine and histidine can be understood as decreasing proportions of these oxygen-sensitive residues. Oxygen concentrations are higher in the liquid medium at low temperatures. Although similar composition gradients were seen in another psychrophile, P. haloplanktis, they are stronger in P. ingrahamii, perhaps correlated with its lower growth temperatures.

Thus, the main features that emerged from the CA of P. ingrahamii protein compositons (Figure 2) are that (1) there are more classes of proteins than have been seen in other bacteria, (2) one of these classes, cluster 4, has a high proportion of "orphan" hypothetical proteins, (3) IIMPs merge into bulk proteins rather than occupying a separate space, possibly due to IIMPs having lower hydrophobic character, and (4) one notes the strong opposition between asparagine, sensitive to heat, and the amino acids methionine, arginine, cysteine and histidine, sensitive to oxygen.

From the above two sets of ortholog matches combined with the work of the JGI scientists, we constructed a table of P. ingrahamii gene numbers and corresponding best-guess annotated gene products available in Additional file 2: "Ping Annotations 2". Included in the table are the gene number, the type of gene product (enzyme, regulator, etc) and the name of the protein. (Gene number throughout this report is the number of the locus_tag which, as submitted December 2006 to Genbank, CP000510.1, has the prefix "Ping"). Altogether 217 proteins originally characterized as unknown gained an annotation by the manual process, some suggestive, others substantive.

Table 4 gives classification of all proteins of P. ingrahamii by imputed function. The distribution of sizes of classes is similar to that for Escherichia coli K-12 20 with enzymes being the largest class followed by transporters, then regulators, the rest divided into smaller categories.

Eighty-one genes were identifiable as currently known horizontally transferrable loci such as transposases of insertion sequences and integrases of phages. Thus clearly horizontal exchange of mobile elements into the chromosome has taken place. Among the bacteria tested, most such loci of P. ingrahamii matched elements of Shewanella oneidensis MR1. As is the case with so many genome sequences, we cannot know with currently available information how much of the genome of P. ingrahamii was formed by horizontal transfer and how much vertically inherited.

There are 10 ribosomal RNA clusters containing 5S, 16S, 23S RNAs (see Additional file 2). The relatively high number could reflect the need for high capacity of translation at cold temperatures and an ability to adapt quickly to changing conditions of nutrient availability 21. 86 tRNA genes support translation.

There are 100 contiguous clusters of two or more related genes such as subunits of an enzyme or pathway-related proteins. The largest cluster, for septum formation and peptidogylcan synthesis enzymes, comprises 14 genes. These are genes 1140 through 1153. Altogether 850, or 25% of CDSs, reside in clusters related by function.

Enumeration of sets of paralogs within a genome requires definition of degree of relatedness. In the initial analysis done at Oak Ridge National Laboratory, 1799 proteins were identified as belonging to 510 paralogous clusters. Using Darwin analysis 22 and a more conservative threshold of relatedness (Pam =< 175), 965 proteins were identified as grouped into 273 paralogous families of sizes ranging from two to 66. This threshold is comparable to that used to enumerate E. coli paralogs 23 and permits comparison. Functions of paralogous groups of size 7 or above are shown in Table 5. Just as had been found previously for E. coli, the largest paralogous groups are transporters and regulators. Even though enzymes are present in the genome in the largest numbers, they fall into smaller, more differentiated families. For transporters and also regulators, evidently a limited number of mechanisms have evolved for these functions, creating large groups of similar proteins of either transporters or regulators that differ in specificity but not in mechanism of action. Enzymes are more diverse, use a larger variety of mechanisms, thus they fall into a larger number of smaller groups of similar proteins.

Almost half of all enzymes annotated in P. ingrahamii (634/1317) were identified as enzymes of small molecule metabolism. Not all enzymes of every pathway were found, not unexpected since some orthologs may not reach the threshold of similarity employed, nevertheless there is strong evidence for standard pathways of intermediary metabolism being present.

P. ingrahamii is a facultative anaerobe capable of both respiratory and fermentative metabolism 24. In agreement we find in its enzyme sequences that pathways are present for fermentation, glycolysis, pentose phosphate pathway, the TCA cycle and gluconeogenesis. For respiration, electron transfer agents are present such as iron-sulfur centers, flavodoxin and flavoproteins, ubiquinone, and cytochromes; for anaerobic respiration, sequences of reductases for fumarate, nitrate, nitrite and sulfite are present (Table 6). Nitrate reduction has been observed experimentally 24. In addition there are 10 members of the short-chain oxidoreductase family and 8 oxidoreductases identified by domain. Any one of these could be either a primary dehydrogenase or a terminal reductase not yet characterized (Table 7). The varieties of oxidants that appear to be used by P. ingrahamii suggest that its capabilities in this regard are comparable to that of Shewanella species.