The T. acidophilum intein was discovered while sequencing the catalytic subunit of the archaeal ATPase/ATPsynthase from T. acidophilum for systematic purposes 24. More recently, the complete genome sequences of T. acidophilum25 and T. volcanii26 have been reported. In both instances, the catalytic subunit of the ATPsynthase is the only gene in the entire genome for which an intein was reported. Using PSI-BLAST 27 with different inteins as seeds, divergent inteins with less than ten percent sequence identity as compared to the query sequence are recovered (data not shown); however, using this approach we did not discover any additional intein or homing endonuclease encoding genes in T. acidophilum.

Multiple sequence alignments of diverse intein sequences identified eight motifs composed of moderately conserved residues 1828. The A-ATPase A subunit of the Thermoplasma and Pyrococcus intein multiple sequence alignment (with manual modification) is shown in figure 1. The T. acidophilum intein (173 amino acids long) is among the shortest inteins known, and the alignment with other inteins reveals the absence of sequences homologous to the typical endonuclease motifs. Only the motifs characteristic for the self splicing domain are present in the Thermoplasma intein (see figure 1).

The significance of the match between the T. acidophilum and the three pyrococcal ATPase inteins was assessed using PRSS at http://fasta.bioch.virginia.edu/fasta/prss.htm29. The P-value for this match, i.e. the probability of obtaining a match of this quality by chance alone, was calculated to be below 10^-10. This indicates that not only the exteins, but also the inteins themselves are recognizably homologous. In contrast, a sequence alignment of all known inteins shows intein sequences to be much more divergent (not shown). For example, the P-value for the comparison between the T. acidophilum and the Saccharomyces cerevisiae intein was 0.28, i.e. no significant similarity between the yeast and the Thermoplasma inteins was detectable using pairwise alignments only.

The vacuolar and the archaeal ATPases are homologous to the bacterial and organellar F-type ATPases. The structure of the bovine mitochondrial F₁-ATPase has been determined by X-ray crystallography 30. The intein insertion points in the yeasts' V-ATPase and the Thermoplasma and Pyrococcus A-ATPases correspond to the catalytic site where the ATP binds and is hydrolyzed during the catalytic cycle. Figure 2 shows that the inteins are located in the regions of these very conserved proteins 22 that have the lowest substitution rates.

The phylogenies of archaeal ATPase catalytic subunits and small subunit ribosomal RNAs are shown in Figure 3. Both phylogenies were calculated for a similar set of species. While the two phylogenies show some significant differences, neither of them groups the molecules from Thermoplasma with the Pyrococcus homologs. In both phylogenies several other Archaea, i.e., M. jannaschii, M. thermolithotrophicus, Thermococcus sp., Halobacterium sp., Methanosarcina and Methanobacterium thermoautotrophicus branch off between the two groups that carry inteins in their ATPase catalytic subunits. All of these intervening archaeal ATPases do not carry an intein.

Codon usage varies among organisms. The production of tRNAs corresponds to the frequencies with which the different codons are present in their protein coding genes. The exact causes for tRNA regulation and codon usage are not completely understood; however, expression of foreign genes in an organism is often prevented by a different codon usage of the foreign gene 31. Many Archaea have codon usage frequencies and tRNA compositions different from E. coli. The T. acidophilum A-ATPase A subunit has 763 codons. Codon AUA (Ile) is the most frequent (39/1000). In E. coli the same codon (AUA) is a rare codon present at a frequency of only 5.5 per 1000. The other two major differences in codon usage between the T. acidophilum A-ATPase A subunit and E. coli are AGG (Arg) and AGA (Arg) which are present in T. acidophilum A-ATPase A subunit at frequencies of 41 and 16 per 1000 respectively. In E. coli however, these codons are considered rare and occur with frequencies of 1.7 and 2.8 per 1000 respectively.

The gene encoding the T. acidophilum A-ATPase A subunit was cloned into the expression vector pET-11a (Stratagene) and transformed into E. coli Bl21(DE3) and E. coli Bl21-CodonPlus(DE3)-RIL strain for protein expression. When E. coli Bl21(DE3), a strain that did not express additional rare tRNAs, was transformed with the cloned T. acidophilum ATPase A subunit, no additional protein bands were observed in extracts of induced cells (not shown). However, the E. coli Bl21-CodonPlus (DE3)-RIL strain transformed with the same plasmid expresses two additional proteins of 20 and 65 kDalton upon induction (Fig. 4). No additional band at 85 kDa, indicative of an unprocessed intein, was visible after induction. This demonstrates that autocatalytic splicing occurred efficiently in E. coli. Efficient autocatalytic splicing was also observed when the E. coli were grown and induced at 42°C, or when the E. coli were induced at 16°C for 16 hours (Fig. 4A). During preparation for SDS gel electrophoreses the samples are heated to 72°C in the presence of DTT. With respect to temperature these conditions are more similar to the conditions in the T. acidophilum cytoplasm than the conditions in E. coli. Intein excision might occur only during this high temperature treatment. Therefore, proteins from induced E. coli were also separated under non-denaturing conditions with or without addition of DTT. The induced bands were excised from the non-denaturating gel and separated using denaturing SDS gel electrophoresis (Fig. 5). Both of the slower migrating bands visible after induction (A and B in Fig. 5), revealed only one major band corresponding to the spliced and religated A-subunit upon separation in an SDS denaturing gel. Presumably the slower migrating band was a dimer or higher aggregate of the A-subunit monomer. In none of these experiments did we find any indications for unprocessed or incompletely spliced inteins.

The T. acidophilum A-ATPase A subunit encoding gene was amplified from genomic DNA using primers Ta-4 (ATGGATCCTTCTCAACGAAGAGCAGTG) and Ta-5 (GAGGTGAACATATGGGAAAGATAATCAG). These primers match the coding sequence of the Thermoplasma acidophilum A-ATPase A subunit (gene identification number 9369337) and introduce restriction sites useful in subcloning. Initially, the PCR product was cloned into pCR^R 2.1 (TA cloning Vector, Invitrogen). After digestion with NdeI and BamHI the coding sequence was subcloned to the vector pET-11a (Stratagene) for gene expression experiments.

E. coli Bl21(DE3) and Bl21-CodonPlus(DE3)-RIL strain (Stratagene) were used for protein expression. If not stated otherwise, transformed E. coli were grown overnight in a culture wheel at 37°C in Luria-Bertani broth with ampicillin (100 μg/mL). One mL of the broth was used to inoculate 10 mL of LB broth containing ampicillin and incubated at 37°C (200 rpm). After 2 hours, isopropyl thio-β-D-galactoside (IPTG) was added to a final concentration of 1 mM to induce gene expression and the cultures. Cells were harvested 4 hours after induction by centrifugation at 7000 rpm for 10 min and washed with TEP buffer (0.1 M Tris-HCl, pH 7.4, 0.01 M EDTA and 1 mM phenyl methyl sulfonyl fluoride). Cells were resuspended in 500 μL of TEP buffer and sonicated with a Braun-Sonic U with micro-tip for 4 minutes (0.5 duty cycle; power output approximately 120). The disrupted cells were centrifuged at 8000 rpm and the pellet was resuspended in 500 μL of fresh TEP buffer. Both the supernatant and the pellet were diluted by adding 6X sample buffer (10% SDS, 1.2 mg/mL bromphenol blue, 0.6 M DTT, 30% glycerol, .1 M Tris/HCl pH 6.8) and heated to 80°C for 15 minutes to denature the protein. Five to fifty microliters of this preparation were run in a 10% denaturing Tris/Tricine SDS polyacrylamide gel electrophoresis system as described by Schagger and von Jagow 42.

Non-denatured protein was prepared with the same procedure except that SDS was omitted from the buffers and that the samples were not heated. Proteins were electroeluted from the non-denaturing gel as describe here 24. Gels were fixed with fixing solution (50% methanol, 10% acetic acid) for 15 minutes, followed by staining in Coomassie staining solution (20% acetic acid, 0.025% Coomassie blue G-250) for one hour, followed by destaining in destaining solution (5% methanol, 10% acetic acid) 42.

DNA was sequenced using the ABI PRISM BigDye Terminator Cycle Sequencing (PE Applied Biosystems). Sequencing gels were ran and processed in the Biotech Center (University of Connecticut)

The program Codon Usage Tabulated from GenBank (CUTG) at http://www.kazusa.or.jp/codon/43 was used to calculate the codon usage of individual genes and genomes.

The Pyrococcus furiosus A-ATPase sequence was retrieved via blastp from the unfinished genome using the web page http://combdna.umbi.umd.edu/bags.html. The aligned small subunit ribosomal RNA sequences were retrieved from the Ribosomal Database Project II http://rdp.cme.msu.edu44. Francine Perler (New England Biolabs) the curator of the intein database http://www.neb.com/inteins/int_reg.html kindly provided the sequences of all known inteins. All the other sequences were retrieved from the NCBI databank.

Sequences were aligned using CLUSTAL X 1.8 45. The number of substitutions per site in the aligned data set were calculated using a JAVA program written by Olga Zhaxybayeva (University of Connecticut). This program calculates and plots the number of substitutions in a sliding window of 10 aligned positions. The window is moved through the alignment one position at a time. For positions where less than 50% of the sequences have a gap, the average number of substitutions was calculated considering the gap as an additional character. Positions with gaps in more than 50% of the aligned sequences were skipped.

Phylogenies were reconstructed from amino acid sequences aligned using CLUSTAL X 1.8 45. The topologies of the depicted trees were calculated using neighbor joining as implemented in CLUSTAL X with correction for multiple substitutions. Branch lengths and their confidence intervals were calculated with TREE-PUZZLE 5.0 46 using the JTT or the HKY model for substitution respectively, and assuming an among site rate variation described by a gamma distribution. Bootstrap samples were analyzed using parsimony as implemented in PAUP* 4.0 beta 8 47 treating gaps as missing data. Each bootstrapped sample was analyzed using 10 different starting trees built through random addition, tree-branch-reconnection (TBR) branch swapping, and considering gaps as missing data. The following sequences were used for the 16S rRNA phylogeny (the sequences are available under these names from the Ribosomal Database Project II http://rdp.cme.msu.edu/ : Sul.solfa4, Sul.acalda, Ap.pernixl, Mc.thlitho, Mc.janrrnA, Mb.tautot2, Pc.furios2, Pc.abyssi, AP000001, AB016298, Tpl.acidop, Arg.fulgid p, Hf.volcani, Hb.spCh2_2, Hb.salina2, Msr.mazei5, Msr.barke2, Dco.mobili.