The tblastn searches, intron analysis, manual checking of the 19 conserved amino acid sites, and sequence alignment reveal that there are 57 bHLH members in H. saltator (Additional file 2). The alignment of all 57 PabHLH members is shown in Figure  1, and the phylogenetic tree generated utilizing amino acids of 57 PabHLH motifs and 59 DmbHLH motifs is illustrated in Figure  2. Both figures demonstrate that there are 23, 14, 10, 1, 8 and 1 PabHLH members in group A, B, C, D, E and F, respectively. It can be seen from Figure  1 that sites 23 and 64 of the bHLH motif are the most conserved sites among all PabHLH motifs. Besides, other ten sites, marked with asterisks on top of Figure  1, are also highly conserved. Additionally, Figure  1 indicates that one PabHLH motif PaDys2 has quite special amino acids. Whereas the alignments of all identified bHLH motifs in other insect species have no gaps in the basic and helix 1 regions and only one major gap in the loop region 13 14 , PaDys2 has two additional amino acids (T and P) in helix 1 region. The existence of these amino acids has created an additional gap among aligned PabHLH motifs (Figure  1), indicating certain difference between H. saltator and other insect species. From Figure  2, we found two cases, like PaAse1 and PaAse2 which can form a monophyletic clade with ase from fruit fly, that the two PabHLHs group together with high statistical support to the exclusion of any other sequences and are often orthologs of a single fruit fly motif. This may reveal relatively recent duplications specific to the H. saltator. Besides these characters on Figure  2, two PabHLHs, PaH1 and PaH2, can be related to two D. melanogaster families of orthologs. This may indicate that the diversity of those bHLHs found in both species occurred before the generation of duplications.

During our analyses, one sequence “RREIANSNERRRMQSINAGFQSLRSLLPHHEGEKLSKVCIV” (contig number: AEAC01009287.1, coding region from 29454 to 29576) was found to have high similarity with AP4 family. However, the immediately following codon is a stop codon. We consider that it could be a pseudogene or the sequence with a nonsense mutation. But it is also possible that the protein sequence coded by this gene was not accurately predicted.

Based on bootstrap supports provided by the in-group phylogenetic analyses, the identified PabHLH motifs were subdivided into corresponding bHLH families and named according to nomenclature used by DmbHLH sequences. Although a new nomenclature for bHLH proteins has been proposed recently 24 , we adopted nomenclature used in D. melanogaster for facilitating further studies on structural and functional comparison with D. melanogaster. In case that one DmbHLH sequence has two or more H. saltator homologues, we use ‘1’, ‘2’ and ‘3’ etc. to number them. For instance, two homologues of the D. melanogaster Mistr and USF genes were found in H. saltator, respectively. Therefore, these PabHLH genes were named PaMistr1 and PaMistr2, PaUSF1, and PaUSF2, respectively. Names of 57 PabHLHs in accordance with their corresponding D. melanogaster homologues are listed in Table  1.

We named PabHLH genes according to their D. melanogaster homologues. Bootstrap values were obtained from in-group phylogenetic analyses with D. melanogaster or A. mellifera bHLH motif sequences using NJ, MP, and ML algorithms, respectively. OsRa (the rice bHLH motif sequence of R family) was used as the outgroup in each constructed tree. n/m means that a H. saltator bHLH does not form a monophyletic group with any other single bHLH motif sequence. n/m* means that a H. saltator bHLH does not form a monophyletic clade with any specific bHLH motif sequence but forms a monophyletic clade with other bHLH proteins of the same family. * means that orthology of the gene was defined through in-group phylogenetic analyses with bHLH orthologs from A. mellifera.

Orthologous genes in two or more organisms are those that evolved by vertical descent from the same gene in the last common ancestor 25 . Ortholog identification has much uncertainty because of the lack of absolute criterion that can be applied to decide whether two genes are orthologous or not 7 . Nevertheless, in our previous studies 13 14 , in-group phylogenetic analysis was adopted to identify homologues for the unknown sequences that would form a monophyletic clade among themselves. Therefore, a more certain standard based on the criterion used by Ledent et al. was used in this study. That is, we defined bHLH families of orthologs as monophyletic groups which include sequences of a known family and whose monophyly is consistent with the different phylogenetic algorithms and supported by bootstrap values superior to 50 7 9 17 .

We have performed in-group phylogenetic analysis to each of the 57 identified bHLHs, which enabled us to allocate all the identified PabHLHs to defined evolutionary conserved groups of orthology. Figure  3, as an example here, shows distance neighbour-joining (NJ), maximum parsimony (MP), and maximum likelihood (ML) phylogenetic trees constructed with one PabHLH member (PaCrp1) and 10 group B bHLH members from D. melanogaster. PaCrp1 formed monophyletic clade with crp (cropped) sequence of D. melanogaster with bootstrap values of 100, 100 and 94 in NJ, MP and ML phylogenetic trees, respectively. PaCrp1 was therefore considered as an ortholog of fruit fly crp. Similarly, in-group phylogenetic analysis was conducted to each of the identified PabHLH members. All the bootstrap values of constructed NJ, MP and ML trees for each of the identified PabHLH members were listed in Table  1 without displaying the correspondent constructed trees. The majority of these bHLHs could be clearly allocated to the families defined according to bootstrap values of in-group phylogenetic trees. Nevertheless, eight PabHLHs (a significant proportion, about 14%) could not be confidently allocated to the defined families by our phylogenetic analysis with DmbHLHs. They were used to construct trees with Apis mellifera bHLHs (AmbHLH) using the same methods mentioned above. Table  1 showed that orthology of PabHLHs with fruit fly or honey bee bHLHs could be divided into the following categories.

Firstly, among all the 57 PabHLH members, there are 44 bHLH members of which all the bootstrap values ranged from 50 to 100 in constructed NJ, MP and ML trees. Namely, they are PaDa, PaNau, PaTap(bp), PaMistr1, PaMistr2, PaOli, PaAto, PaAmos, PaNet, PaMyoR, PaSage, PaPxs, PaTwi1, PaFer1, PaFer2, PaFer3, PaHand, PaSCL, PaNSCL, PaMnt1, PaMax1, PaMax2, PaDm, PaUSF1, PaUSF2, PaMitf, PaCrp1, PaBmx, PaMLX, PaSREBP, PaTai, PaClk1, PaDys1, PaSs, PaSim, PaTrh, PaSima, PaTgo, PaCyc, PaEmc, PaStich1, PaH1, PaSide and PaKn(col). All these bootstrap values were no lower than the set criterion (50) and thus resulted in assignment of corresponding D. melanogaster homologues for them with sufficient bootstrap support (Table  2).

Secondly, one bHLH member, PaTwi2, had bootstrap value of 43 in MP tree. Nevertheless, it formed monophyletic clade with the same DmbHLH counterpart in NJ and ML tree with bootstrap values of 63 and 73, respectively. Two members, PaAse2 and PaCato, formed monophyletic clade with bootstrap values of 55 and 61 in ML trees, but formed monophyletic clade with weak bootstrap values (33 to 48) in NJ and MP trees. Consequently, we allocated them to defined families of orthologs according to the one or two trees with bootstrap values of over 50.

Thirdly, one bHLH member, PaHey, formed monophyletic clade in NJ and MP trees with bootstrap values of 98 and 45, respectively, but did not form monophyletic group in ML tree. Another PabHLH member, PaH2, formed monophyletic clade with bootstrap value 47 in MP tree, but did not form monophyletic clade in NJ and ML trees (marked with n/m* or n/m in Table  1). Albeit with insufficient statistical support, we tentatively defined orthologs for them because they all have one or two bootstrap support to testify their orthology to the correspondent D. melanogaster ortholog. Obviously, these classifications can be regarded as arbitrary and ought to be modified upon new data available. From a certain perspective, this phylogenetic divergence of bHLH motif sequences between H. saltator and D. melanogaster probably implies that these two insect species have evolved in quite different circumstances.

Finally, the remaining 8 members named PaAse1, PaMnt2, PaCrp2, PaClk2, PaDys2, PaE(spl)1, PaE(spl)2 and PaE(spl)3 did not form monophyletic clade or did not have sufficient bootstrap support in forming monophyletic clade with any single D. melanogaster homologue in all three phylogenetic trees constructed. They were identified through constructing phylogenetic trees with AmbHLH family members accordingly. Six members, namely PaAse1, PaMnt2, PaCrp2, PaClk2, PaE(spl)2 and PaE(spl)3, were identified with sufficient confidence for all the bootstrap values were over 50 in all the constructed trees. The rest two members, PaDys2 and PaE(spl)1, formed monophyletic clade in NJ and MP trees with bootstrap values ranging from 50 to 98. We assigned orthologs for them according to the two trees with bootstrap values over 50, though they did not form monophyletic group in ML trees.

Through prediction by SMART using the full sequence of identified bHLH members whose protein accession number were available (Additional file 3), we found that: a) Among members of group C, there are 4 sequences having one bHLH, one PAC (Motif C-terminal to PAS motifs) 26 and two PAS (PER-ARNT-SIM homology) domains, while EFN88377.1 has one bHLH, and two PAS domains. The remaining one (EFN80844.1) only has bHLH domain. b) For group E, all of the PabHLHs have bHLH and Orange domains, one of which, however, has Orange with scores less significant than the required threshold. And all of them ended with “WRPW” peptide. c) The rest groups were predicted to only have bHLH domains. These results suggest that our analyses are consistent with the previous reports 3 27 28 , and it is conceivable that these domains may cooperate and thereby confer particular functions on the proteins containing them 9 .

Protein sequence accession numbers of the 57 identified PabHLH motifs were listed in Table  1. As we have seen, there are only 46 PabHLH motifs whose protein sequence accession numbers were found in H. saltator genome database (shown as ‘EFN’ plus number). Protein sequences of the other 11 PabHLHs, namely PaAto, PaNet, PaSage, PaPxs, PaMnt2, PaCrp1, PaCrp2, PaClk1, PaDys1, PaDys2, and PaTgo, were not found in current database. The coding regions, intron location and length of 57 PabHLH motifs are listed in Table  2. The intron analysis shows that 22 PabHLH members have introns in the coding regions of their bHLH motifs. It should be noted that: a) coding regions of 18 PabHLH motifs have one intron, among which those of 6 PabHLH motifs have introns in the basic region, 7 have introns in the helix 1 region, 2 have introns in the loop region, and 3 have introns in the helix 2 region. b) Coding regions of 4 PabHLH motifs have two introns, all of which are in the basic and loop regions. Thus, altogether, coding regions of these 22 PabHLH motifs have 26 introns. In addition, 2 PabHLHs named PaCrp1 and PaSide locate on two separate contigs in the genome (Table  2). The longest intron in coding regions of PabHLH motifs is 7,943 bp (base pairs), the shortest one is only 82 bp, and the average length of introns is 1,391 bp. While in pea aphid, fruit fly and honey bee, there are 26, 18 and 9 bHLH members having introns in coding regions of their bHLH motifs, and the total number of introns identified is 34, 20 and 9 with the longest one of 30,718, 11,845 and 4,460 bp, the shortest one of 62, 57 and 72, and the average length of 4,193, 1,082 and 1,326 bp, respectively 14 16 .

In summary, the number of PabHLHs having introns is more than many other insect species, and they are found to have introns with average length only inferior to those of pea aphid. Moreover, PabHLHs have the shortest length of intron only higher than those of fruit fly and honey bee. PabHLH genes are more intron-dense than those of many other insects, indicating that H. saltator either gained introns at a faster rate or lost introns at a slower rate than other insects 29 . Previously hypothesized mechanisms of intron gain mainly involve intron transposition 30 , transposon insertion 31 , tandem genomic duplications 32 , intron transfer 33 , insertion of a Group II intron 30 , intron gain during double strand break repair 34 and intronization 35 36 . Three previously hypothesized mechanisms of intron loss include Reverse Transcriptase-Mediated Intron Loss (RTMIL) 37 , Meiotic recombination 29 and genomic deletions 38 . Notably, the genome of H. saltator contains copies of specific transposable element (TE) families, which differs significantly with other species (i.e. Camponotus floridanus and A. mellifera have very few TEs) 22 39 . We infer that there may be some relationships between the formation of intron in PabHLHs and TEs. Nevertheless, whether the more intron-dense introns are due to growing faster or losing slower needs further investigations.

The existence of EST (expressed sequence tag) sequence corresponding to identified bHLH motifs is an indication of genuine bHLH gene at least at the transcriptional level. However, we were unable to find any EST sequences for our identified PabHLHs due to the unavailability of H. saltator EST database at time of our survey. Therefore, further verification of our identified PabHLH members awaits expansion of H. saltator nucleotide sequence data and experimental cloning of the genuine PabHLH genes. It would thus be of particular interest to study whether the bHLH family members identified in H. saltator are expressed during various developmental processes and how their expression patterns may be related to those of other insect counterparts.

The above searches and analyses enabled us to define orthologs of families for 57 PabHLHs. This figure is comparable with 50, 51, 52, 54 and 59 bHLH members in the red flour beetle, honey bee, domestic silkworm, pea aphid and fruit fly, respectively (Table  3). According to Table  3, all of these five insect species lack genes of families Oligo, MyoRb, and Figα, and many of the families have the same number of genes, such as E12/E47, Ngn, Beta3, Net, Paraxis, Hand, PTFa, SCL, NSCL, SRC, Myc, SREBP, ARNT, Trh, HIF, Emc and COE. The fact that comparable number of bHLH families and similar orthologs were detected strongly suggests that, despite our analysis was made on a draft version of the H. saltator genome sequence, the set of PabHLH we retrieved is likely to be almost complete, and hence gives a highly accurate view of the bHLH repertoire of a ponerine ant. The major obvious difference, other than total genes, is the discrepancy of H/E (spl) family members. D. melanogaster have 11 to 12 H/E(spl) genes while other insects have 5 to 6. H. saltator has fewer genes in families ASCa, and Clock than D. melanogaster. It is conspicuous that H. saltator has one more gene in families Twist, Mnt, Max, USF, AHR and AP4 than most other insect species previously reported. Thirteen PabHLH families have more than one member (accounted for about 29% of the families), while in most other insects, families with more than one member are fewer (mean value 16%; range 13% to 20%). This suggests that some of the H. saltator bHLH genes have been originated through duplications. Moreover, a feature to be noted is that neither of the two families, namely Delilah and MyoRb, was found. Whether H. saltator does have fewer members of these families, or if it was due to incompleteness of the genome sequences remains for further exploitation. Therefore, it can be thought that additional bHLH members may be found after a newer and higher quality version of H. saltator genome sequences is released.

Data of P.a. (ponerine ant) were from this study. Those of A.m. (Apis mellifera) and B.m. (Bombyx mori) were from Wang et al. 2007, 2008 . Those of T.c. (Tribolium castaneum) and D.m. (Drosophila melanogaster) were from Simionato et al. 2007 . Those of A.p. (Acyrthosiphon pisum) were from Dang et al. 2011 .

We also executed several additional alignments consisting of only those bHLH sequences that belong to a particular family from 6 insect species above mentioned (Additional file 4). Based on these alignments, the in-family NJ trees were constructed (see Materials and methods) which has been rooted using a fruit fly bHLH sequence from the related family as outgroup. From our phylogenetic analyses (the trees not shown), 34 PabHLHs and AmbHLHs formed monophyletic clade with high bootstrap values (55 to 100), while the members of other insect species were fewer. So we concluded that the PabHLH genes, to a certain extent, have closer phylogenetic relationships with A. mellifera than with other insect species. Additionally, figure  4 shows a typical phylogenetic tree of a family containing the E(spl) family of the five insect species. The members of fruit fly Esplm3, EsplmBg, EsplmAb and Esplm7 clustering together may reveal that these are the result of species-special duplication.

Candidate genomic sequences encoding bHLH motifs were identified using BLAST, NetGene2 and EditSeq program (version 5.01), prepared and improved with manual checking and phylogenetic analysis. As a starting point, both lists of 59 DmbHLH motifs and the 45 representative bHLH families were retrieved from the additional files of previous reports 17 . Each sequence was used as a query sequence to perform tblastn search against H. saltator genome sequence database ( http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi?organism=insects). The expect value (E) was set at 10 in order to detect all possible bHLH sequences. Redundant sequences were manually identified and subsequently discarded on purpose of keeping only one sequence with the same contig number, reading frame, and coding regions. Next, the sequences were examined again to check whether the obtained amino acids have covered the full bHLH motif or not. If not, the corresponding subject nucleotide sequence was retrieved and translated using EditSeq program (version 5.01) of the DNAStar package to add the missing amino acids on two ends of this motif. In case where a query sequence composed of two or three H. saltator coding regions, intron splice sites were assessed using the online program NetGene2 ( http://www.cbs.dtu.dk/services/NetGene2/) to find location of intron between the separate coding regions.

Moreover, in order to detect whether the retrieved bHLH sequence has corresponding protein sequences deposited in GenBank, each of the retrieved candidate bHLH sequences was used to make similarity searches using blastp algorithm against the NCBI nr database by limiting the query organism to H. saltator. And tblastn search was conducted against the NCBI Expressed Sequence Tag (EST) database to examine whether there were ESTs corresponding to the identified H. saltator bHLHs.

Finally, each obtained sequence was examined for their amino acid residues at the 19 conserved sites 40 by manual checking. Accordingly, a sequence with less than 9 mismatches could be a potential bHLH motif 41 . Therefore, sequences having no less than 10 conserved amino acid residues among the 19 conserved sites were regarded as potential PabHLH (ponerine ant bHLH) members, except for Emc and COE family members which have around 30 and 50 amino acids in their HLH motif respectively and thus the minimum conserved amino acids were adjusted to 5 and 8 respectively. Sequences failing to meet the above requirements were discarded.

To examine sequence features of these PabHLH domains, we performed multiple sequence alignment of all the potential bHLH sequences improved by the aforementioned approaches using Clustal W program (version 5.0) implemented in MEGA 5 42 using the default settings. The aligned PabHLH motifs were highlighted in GeneDoc Multiple Sequence Alignment Editor and Shading Utility (Version 2.6.02) 43 and then copied to rich text file (RTF) for further annotation.

Three different algorithms were employed for phylogenetic reconstruction: distance neighbour-joining (NJ), maximum parsimony (MP), and maximum likelihood (ML). In general, the phylogenetic trees constructed by the different algorithms were congruent and displayed very similar topologies. First, distance trees were constructed with the neighbor-joining (NJ) algorithm 44 using PAUP 4.0 Beta 10 45 relying on the step matrix constructed from Dayhoff PAM 250 distance matrix by R. K. Kuzoff ( http://paup.csit.fsu.edu/nfiles.html). Each PabHLH motif sequence was then used to conduct in-group phylogenetic analysis with DmbHLH motif sequences. The in-group phylogenetic trees were performed with NJ and MP algorithms implemented in PAUP program, as well as with ML algorithm using the program TreePuzzle 5.2 46 . The NJ tree was bootstrapped with 1,000 replicates to provide information about their statistical reliability, while the MP analysis was generated with heuristic search of 100 bootstrap replicates. For ML reconstruction, the parameters were set as follows: quartet-puzzling tree-search procedure, 25,000 puzzling steps and the substitution model set to Jones-Taylor-Thornton 47 . Other parameters were of default values. Additionally, all members from each bHLH family of ponerine ant, red flour beetle, honey bee, domestic silkworm, pea aphid and fruit fly were used to perform phylogenetic analyses, which were termed as in-family phylogenetic analysis because the bHLH motifs for a particular analysis were from the same bHLH family. Each in-family phylogenetic tree was rooted using fruit fly bHLH sequence from a different but related family.

In order to further ascertain the reliability of the retrieved motifs and to examine whether the full-length protein sequences, especially those of Group C, E and F, contain additional characteristic domains, we carried out the predictions of protein domain architectures using Simple modular architecture research tool (SMART, http://smart.embl.de/) available online 48 49 .