Partial to complete sequences of two distinct copies of GH were determined for 14 catostomid species; complete sequences for one of the GH copies were obtained from nine additional species (Table 1). BLAST searches of the coding regions revealed high similarity of the new GH sequences with GH copies of Ictiobus bubalus and other cypriniform fishes. The two GH copies are named GHI and GHII based on their sequence homology with GH copies in I. bubalus [16].

We were able to produce complete coding region data for GHI for most catostomid species using methods described in the experimental section. We were able to produce data for the 5’ end of GHII (Exons 2 and 3) for most catostomid species using GHII specific primers developed in a previous study [16]. However, despite several attempts involving a number of different techniques (also described in the experimental section), thus far we have only been able to produce data for the 3’ end of GHII for species representing tribes Erimyzonini and Catostomini of subfamily Catostominae, in addition to a previously published GHII sequence for I. bubalus of subfamily Ictiobinae [ 16].

The genomic organization of GH in suckers is the same as in other Cypriniformes [26,27]. The complete GH genomic sequence comprises five exons and four introns with a total length of 1,500–2,700 nt depending on lengths of the four introns. Exons of different sucker species are of fixed lengths as follows: 10 (Exon 1), 140 (Exon 2), 117 (Exon 3), 162 (Exon 4), and 204 (Exon 5) nt. Introns vary in size across species, from 155–269 (Intron 1), 154–215 (Intron 2), 311–1,188 (Intron 3), and 102–154 (Intron 4) nt (Table 2). The GHII genomic sequence is shorter than that of GHI, with much of the difference due to the substantially longer 3^rd intron of GHI.

The GH coding region of catostomids is 633 nt in length. The predicted amino acid (aa) sequences of GHI and GHII encode a protein of 210 aa, which is identical to the protein size reported for other cypriniforms [16,22,28].The putative GH signal peptide cleavage site is serine at aa position 23, which gives a predicted mature polypeptide size of 188 aa, consistent with other cypriniform species [28].

The two GH copies are very similar in both nt and aa sequence composition. Mean nt divergence between GHI and GHII is 9.61%. Mean pairwise aa sequence divergence between copies is 8.53%. Mean pairwise nt sequence divergence within paralogs (coding region data only) across catostomid species is 3.33% for GHI and 3.22% for GHII. Mean aa divergence within paralogs is 4.46% for GHI and 2.43% for GHII. The lower percentage in aa divergence for GHII is due to the incomplete data for several of the catostomine species.

An interesting and potentially evolutionarily significant difference in GH copies of suckers involves variation in the number of cysteine residues in the mature peptide. Pairs of cysteine residues form disulfide bonds, important to protein folding and stability [29]. GH in all vertebrates has four cysteine residues in highly conserved positions in the amino acid sequence. Ostariophysan fishes have an unpaired, fifth cysteine in aa position 145. In GHI of catostomids, the extra cysteine is replaced by tyrosine. The functional significance of this disparity has yet to be established.

The consensus trees obtained with MP and ML analyses are identical. Only the MP tree is shown (Figure 1). The MP analysis is based on 230 parsimony informative sites in the combined GHI/GHII data set (300 sites are constant). The MP consensus tree is 726 steps long. Order Cypriniformes (Node 1 in Figure 1) is recovered as a monophyletic group with strong bootstrap support. Gyrinocheilus aymonieri (Family Gyrinocheilidae) is strongly supported as the most basal cypriniform. Thus, GH data does not support a monophyletic Superfamily Cobitoidea inclusive of gyrinocheilids, loaches and catostomids, as supported by morphology [30,31] and analysis of multiple nuclear genes and mitogenome data [24].

Two strongly supported interfamilial groups make up the strongly supported sister group to Gyrinocheilus aymonieri. The first of these groups (Node 2) comprises a strongly supported family Cyprinidae (Node 3), sister to a strongly supported group of GHII sequences for representatives of tribe Catostomini (Node 4). The second group (Node 5) comprises a strongly supported basal group of cobitids and balitorids (Node 6), sister to a strongly supported group of GHI and GHII sequences representing other subfamilies and tribes of family Catostomidae (Node 7).

Family Cyprinidae comprises strongly supported subfamily groups of Cyprinines, and Leuciscines plus Gobionines. Within subfamily Cyprininae, the two copies of GH in tribe Cyprinini form a strongly supported monophyletic group, with sequences for each of the copies forming strongly supported monophyletic sister groups.

Five nt substitutions link cyprinid GH sequences with GHII sequences of suckers representing tribe Catostomini, thus rendering the two copies of GH in suckers, and catostomids as a whole, non monophyletic. In contrast, the GHI portion of the tree is well-resolved, monophyletic, and more or less consistent with hypotheses of catostomid relationships based on other data [32,33]. GHI of Myxocyprinus asiaticus is most basal. This species is sister to a strongly supported group comprising a monophyletic Catostominae GHI plus a strongly supported group of Cycleptus elongatus plus a monophyletic subfamily Ictiobinae GHI, the latter group comprising a monophyletic Carpiodes plus a monophyletic Ictiobus. The catostomid GHI tree is the strongly supported sister group to GHII sequences for the remaining catostomid species. In the latter group, Ictiobus bubalus GHII is basal and sister to a strongly supported group comprising GHII sequences for species representing tribes Erimyzonini, Moxostomatini and Thoburniini of subfamily Catostominae.

The sister group relationship of Catostomini GHII sequences with cyprinids was unexpected. Of the five nt substitutions inferred along this branch, four are not shared with other catostomid GH sequences, and two of these substitutions result in aa changes that are also not shared with other catostomids (valine to methionine at aa position 90 and leucine to methionine at aa position 169). The two aa substitutions are in C-terminal end of the protein, corresponding to Exons 4 and 5. GHII data from this end of the gene is available only for Minytrema melanops among Tribes Erimyzonini, Moxostomatini and Thoburniini. GHII sequences of Tribe Catostomini share nine nt characters with GHI and/or GHII sequences from other catostomids and would likely share more if GHII data were more complete. Two of the nine substitutions result in aa changes that are convergent with aa character states in other catostomid GH sequences (serine to cysteine in aa position 14 [signal peptide] and glycine to aspartic acid in aa position 81). It is possible that missing GHII data from the 3’ end of the gene, especially for other tribes of catostomines, would have supported a different tree topology.

When all catostomid GHII sequences are constrained to be monophyletic, the resulting tree is 11 steps longer than the MP consensus tree. When Catostomini GHII sequences are constrained to be the sister group of catostomid GHI plus the remaining GHII sequences, the resulting tree is only four steps longer than the MP consensus tree. Based on Templeton test results, neither constraint tree is significantly longer than the MP consensus tree (GHII monophyletic: Z = −1.9149, p = 0.0555; Catostomini GHII sister to remaining catostomid GHI and GHII sequences: Z= − 0.8944, p = 0.5034).

We compared coding sequences of the mature GHI and GHII proteins of catostomids to gain insight into the possible evolutionary forces affecting the divergence of the two copies of the hormone. The comparison revealed a lower number of non-synonymous differences per non-synonymous site (d_N) relative to the number of synonymous differences per synonymous site (d_S) (P = 0.003, Z-test of positive selection), indicating a paucity of amino acid replacement changes compared with neutral expectations. Thus, the null hypothesis of strict neutrality (d_N = d_S) can be rejected in favor of the alternative hypothesis of purifying selection (d_N < d_S) for all catostomid species. Purifying selection is also suggested for pairwise comparisons of GHI and GHII of the cyprinids Carassius auratus and Cyprinus carpio. There is no evidence for positive selection among the GH sequences tested. The slow rate of divergence of the GH coding region observed across suckers and other cypriniforms is not surprising considering the protein’s critical role in promoting growth and differentiation at distant target sites [34] as well as its secondary functions in autocrine/paracrine regulation of cellular differentiation during embryonic development [35,36].

Total DNA was extracted from ethanol preserved muscle or fin tissue with the Purelink Genomic DNA Mini Kit (Invitrogen, Carlsbad, CA). PCR amplification was conducted in two steps, long PCR and full-nested short PCR. The long PCR primer pair GH22F (5′-YTGTCKDTGGTSCTGGTYAGT-3′) and GHR (5′-CAGGGTRCAGTTKGAATCSAR-3′) was used in a 15.5-μL reaction mixture containing 9.725 μL sterile water, 1.5 μL Ex Taq buffer, 1.2 μL dNTPs (2.5mM), 1.0 μL each primer (10 μM), 0.075 μL Taq polymerase (Takara Ex Taq, Takara, Japan), and 1.0 μL of template DNA (ca. 50 ng/μL). The thermal cycle protocol was as follows: (1) initial denaturation at 94 °C for 60 s; (2) then 30 cycles of denaturation at 94 °C for 30 s; annealing at 52 °C for 30 s; and extension at 72 °C for 120 s; and (3) final extension at 72 °C for 10 min. The first round PCR produced an amplicon ranging from 1,200 to 2,300 bp (depending on length of introns) that spans half of exon 2 to near the termination codon in exon 5. This product contained both GH copies and was cloned in the pGEM-T Easy Vector (Promega, Madison, WI). Positive colonies (i.e. white colonies) were selected and used as a template for short PCR. Short PCR was conducted using up to three internal primer pairs GH22F and GH264R (5′-GCTYTTYTGBGTTTCATSTTT-3′), GH181F (5′-CAGCTGAGTAAAATCTTYCCT-3′) and GH295R (5′-CTCCCARGAYTCAATGAGGYG-3′), and GH274F (5′-AAGCTBCTTCGYATCTCYTT-3′) and GHR with the same reaction mixtures as above [16,24]. Thermal cycling profile was the same as first round PCR except extension time was reduced to 30 s. Part of the 5’ UTR and exons 1 through 3 were amplified and sequenced for catostomids by pairing GH copy specific primers GHIF (5′-AAAGCCTTCAACTAAGACTAAC-3′) and GHIIF (5′-CAAACCTTCAACTAAGACTTCA-3′), developed for Ictiobus bubalus [24], with primer GH240R (5′-TTCTGGGTTTCATGTTTGTCA-3′). Short PCR products were purified with ExoSAP-IT (USB, Cleveland, OH) and directly sequenced using BigDye Terminator v3.1 Cycle Sequencing Kits (Applied Biosystems, Carlsbad, CA). The resulting products were analyzed on an ABI 3730xl Genetic Analyzer (Applied Biosystems).

Obtaining complete coding region sequences of GHII for all sucker species proved challenging because we were not able to design internal primers specific to the 3’ end of this gene copy. We tried designing primers specific to the different sucker tribes and nesting them with our GHII-specific upstream primer and a non-specific downstream primer. This amplified both GH copies. We varied PCR techniques to increase GH yield by performing re-extensions, reconditioning PCR, and varying primer and DNA volumes. This resulted in non-specific primer binding with multiple bands observed during gel electrophoresis. We extended electrophoresis runs on PCR products, cutting out and gel purifying double bands and cloning both products. This yielded large sequences of GHI, but very small fragments of GHII. Lastly, we diluted the ligation mix during cloning in an effort to decrease plasmid incompatibility, thereby increasing the cloning efficiency of paralogous sequences. This method yielded the 3’ GHII data for species we have completed thus far.

Sequence chromatograms were assembled into contigs and edited with Sequencher 4.6 (Gene Codes, Madison, WI). Inconsistencies in base calls in cloned fragments were infrequent and were resolved by simple majority or left ambiguous. Additional GH sequences were obtained from NCBI by taxonomy and BLASTN searches. Sequences were aligned using CLUSTAL W [37] as implemented in BioEdit [38] and visually inspected for errors and improved manually. Sequence divergence (Tamura-Nei distance), Maximum Likelihood (RAxML [39]) and Maximum Parsimony (PAUP* [40]) analyses were performed on GH coding region data only, using the CIPRES web portal (www.phylo.org). Node support is based on 2,000 bootstrap replicates. The extent of nucleotide sequence divergence was estimated by means of the uncorrected differences (p distance). Sequence variation was examined by plotting pairwise transitional (TS) and transversional (TV) differences against p distance.

Templeton tests, implemented in PAUP* [40], were conducted to test for differences in the lengths of the MP consensus tree and two alternative topologies constrained as follows: 1) All catostomid GHII sequences monophyletic; 2) Catostomini GHII sister to remaining catostomid GHI and GHII sequences.

The number of synonymous substitutions per synonymous site (d_S) and nonsynonymous substitutions per nonsynonymous site (d_N) used in selection tests were estimated using the method of Nei and Gojobori [41] as implemented in MEGA version 4 [42]. Nucleotide and amino acid distances were estimated using a pairwise deletion option for each catostomid species for which complete or partial GHI and GHII sequence were determined. The presence of positive selection was analyzed by testing the null hypothesis that H_o: d_N = d_S, versus the alternative positive selection hypothesis that H1: d_N > d_S using the codon based z-test for selection [43]. The z-statistic and the probability that the null hypothesis is rejected were obtained as indicated by P > 0.05.