This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental material
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Poole, F. L.
Right arrow Articles by Adams, M. W. W.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Poole, F. L., II
Right arrow Articles by Adams, M. W. W.

 Previous Article  |  Next Article 

Journal of Bacteriology, November 2005, p. 7325-7332, Vol. 187, No. 21
0021-9193/05/$08.00+0     doi:10.1128/JB.187.21.7325-7332.2005
Copyright © 2005, American Society for Microbiology. All Rights Reserved.

Defining Genes in the Genome of the Hyperthermophilic Archaeon Pyrococcus furiosus: Implications for All Microbial Genomes{dagger}

Farris L. Poole II, Brian A. Gerwe, Robert C. Hopkins, Gerrit J. Schut, Michael V. Weinberg, Francis E. Jenney Jr., and Michael W. W. Adams*

Department of Biochemistry and Molecular Biology, University of Georgia, Athens, Georgia 30602-7229

Received 16 May 2005/ Accepted 19 August 2005


arrow
ABSTRACT
 
The original genome annotation of the hyperthermophilic archaeon Pyrococcus furiosus contained 2,065 open reading frames (ORFs). The genome was subsequently automatically annotated in two public databases by the Institute for Genomic Research (TIGR) and the National Center for Biotechnology Information (NCBI). Remarkably, more than 500 of the originally annotated ORFs differ in size in the two databases, many very significantly. For example, more than 170 of the predicted proteins differ at their N termini by more than 25 amino acids. Similar discrepancies were observed in the TIGR and NCBI databases with the other archaeal and bacterial genomes examined. In addition, the two databases contain 60 (NCBI) and 221 (TIGR) ORFs not present in the original annotation of P. furiosus. In the present study we have experimentally assessed the validity of 88 previously unannotated ORFs. Transcriptional analyses showed that 11 of 61 ORFs examined were expressed in P. furiosus when grown at either 95 or 72°C. In addition, 7 of 54 ORFs examined yielded heat-stable recombinant proteins when they were expressed in Escherichia coli, although only one of the seven ORFs was expressed in P. furiosus under the growth conditions tested. It is concluded that the P. furiosus genome contains at least 17 ORFs not previously recognized in the original annotation. This study serves to highlight the discrepancies in the public databases and the problems of accurately defining the number and sizes of ORFs within any microbial genome.


arrow
INTRODUCTION
 
The advent of high-throughput DNA sequencing has brought into focus the problems associated with predicting protein-encoding, open reading frames (ORFs) on a genomewide basis. The rapid rate at which genomes are now being sequenced has made the practice of manually identifying and annotating ORFs slow and impractical. Widely used sequence databases such as the Comprehensive Microbial Resource (CMR) (28) of the Institute for Genomic Research (TIGR; http://www.tigr.org/tigr-scripts/CMR2/CMRHomePage.spl) and the Reference Sequence (RefSeq) (29) collection of the National Center for Biotechnology Information (NCBI; http://www.ncbi.nlm.nih.gov/RefSeq/) rely heavily on software, such as Glimmer2, GeneMarkS, and BLAST2, to automatically identify ORFs (3, 6, 12). The majority of ORFs are assigned automatically based on sequence comparisons to ORFs already present in public databases.

It seems reasonable to assume that the different public resources would identify the same set of ORFs in a given genome and that the predicted start site of a given ORF does not depend on the automated analysis that was used. However, as demonstrated herein, these are false assumptions and ones that can have serious consequences. Of course, a genuine ORF in a given genome that shows no sequence similarity to ORFs identified in other genomes tends to be overlooked by automated analyses. This is an important issue because ORFs that are unique to an organism are likely to be at the root of strain- and species-specific differences (19). Conversely, in the absence of anything to compare it with, an ORF that is proposed to be unique to a given genome could also be an artifact of the automated analyses (32). In addition, genome sequencing errors and biological frameshifts could result in the misannotation of ORFs.

Despite these problems, it is imperative that a genome be annotated as accurately as possible, especially given the increasing use of genome-based experimental approaches such as DNA microarrays, protein arrays, and structural genomics. Although the results from the automated analyses in the public databases are frequently assumed to be highly accurate, a plethora of additional bioinformatics tools for ORF prediction continue to be reported, which serve to illustrate that this is not the case as well as highlighting the complexity of the problem (4, 7, 9, 16, 25, 27, 43). Comparative genomics can also play an important role in genome analysis. A prime example is the genome of the eukaryote Saccharomyces cerevisiae, which has been analyzed extensively (8, 38). Remarkably, a recent comparison of its genome with those of three related species led to the revision of almost 15% of the more than 6,000 annotated ORFs, including the elimination of almost 500 of them (24). While an accurate description of the ORFs in a genome must be derived from experimental as well as bioinformatic analyses, there are few examples where this has been achieved. One involves the genome of Mycobacterium tuberculosis, where six previously unannotated ORFs were identified using a proteomics-based approach (21).

Currently, a total of 22 archaeal genomes have been sequenced (www.ncbi.nlm.nih.gov/genomes/lproks.cgi?view = 1) and at least 30 more are in progress (http://www.genomesonline.org). The archaeal group also provides several examples of two or more genome sequences from species of the same genus, which can be used to provide insight into genome annotation. These include three from species of Pyrococcus (10, 22), including P. furiosus (31). P. furiosus grows optimally near 100°C (13) and is one of the best studied of the hyperthermophilic archaea. For example, it has been the subject of several studies using genomewide DNA microarrays (34, 35, 37, 40) and is also the focus of a structural genomics initiative (1).

The genome of P. furiosus is approximately 1.9 Mb in size and the original annotation deposited in GenBank in 2002 contained 2,065 ORFs (PF0001 to PF2065) (31). Subsequently, two automated annotations have been made available by CMR at TIGR and by RefSeq at NCBI. We anticipated that these three annotations would be highly similar, if not identical. However, our analyses show that there are profound differences between them and that this phenomenon is not unique to P. furiosus. In addition, we provide experimental evidence that there are many ORFs that were not identified in the original genome annotation that are functional and/or encode stable proteins.


arrow
MATERIALS AND METHODS
 
Bioinformatic analyses. The original annotation of the P. furiosus genome (2,065 ORFs) deposited in GenBank in 2002 was obtained from NCBI (http://www.ncbi.nlm.nih.gov/entrez/viewer.fcgi?18976372:OLD09:199672). The CMR and RefSeq annotations were obtained in January 2004 from TIGR (http://www.tigr.org/tigr-scripts/CMR2/GenomePage3.spl?database = ntpf01) and NCBI (http://www.ncbi.nlm.nih.gov/entrez/viewer.fcgi?18976372:OLD13:1040042), respectively. All data sets were incorporated into a Microsoft SQL Server 2000 database for indexing and extraction. Ad hoc PERL scripts were used to determine the position and degree of ORF overlap. A Microsoft Visual Basic 6.0 program, based on the Expasy ProtParam Tool (http://www.expasy.org/tools/protparam.html), was written and used to predict the physical properties (molecular weight, isoelectric point, length, extinction coefficient, etc.) of each ORF product (41).

Transmembrane domains and signal sequences were predicted using six web-based programs as previously described (18). InterPro analysis was performed using the InterProScan tool (v. 3.3; http://www.ebi.ac.uk/interpro/), and the PFAM analysis (v. 15; http://www.sanger.ac.uk/Software/Pfam/) was conducted on data from July 2004 as described (5, 42). Ribosome binding sites were predicted using RBSfinder (http://www.tigr.org/software/). Sequence alignments were performed using the BLAST toolkit (v. 2.2.6; ftp://ftp.ncbi.nlm.nih.gov/BLAST/) (3). The TBLASTN and BLASTP programs were run against BLAST databases that were built or downloaded in February 2004 from files on NCBI's FTP site (ftp://ftp.ncbi.nlm.nih.gov). These included nonredundant nucleotide and peptide databases, and the complete genomes of Pyrococcus horikoshii and Pyrococcus abyssi. In addition, each program was run multiple times using different substitution matrices (PAM30, PAM70, and BLOSUM62), word sizes (2, 7, and 3), and filters (low complexity) in order to compensate for the various sequence lengths.

Growth conditions and microarray analyses. The transcriptional analyses of P. furiosus were conducted using PCR-based microarrays. The arrays were generated by spotting full-length PCR products of each gene onto glass slides as previously described by Schut et al. (34). For this study, P. furiosus was grown under six different conditions using previously published methods for growing cells, harvesting RNA and performing the DNA microarray analysis (34, 35, 37, 40). Two conditions involved growing P. furiosus cells in batch culture. These were using either peptides or maltose as the carbon source at 95°C in medium containing elemental sulfur (34, 40).

The other four growth conditions were generated from cold shock kinetic experiments. The P. furiosus cultures were initially grown to mid-log phase with maltose as the carbon source at 95°C in the absence of sulfur. At time zero they were rapidly cooled to 72°C, and samples for RNA analyses were removed at 0, 1, 2, and 5 h postshock (40). The batch experiments were carried out twice with two duplicates for each growth condition and were hybridized to arrays containing two copies of each ORF. This yielded eight data points per ORF per condition. The kinetic experiments were carried out in the same fashion except that the array contained three copies of each ORF, which yielded a total of 12 data points per ORF per condition. Gene expression was assessed using the minimum likely signal intensity (MLSI) value, which was calculated using the equation mean (signal intensity - background intensity) - standard deviation (signal - background).

ORFs were considered to be expressed when the value of the mean (signal intensity minus background intensity) was greater than 2,000 arbitrary fluorescence intensity units (which is twice the detection limit) and an MLSI value above 1,500.

Production of recombinant proteins. For recombinant protein production, all ORFs were cloned and expressed in Escherichia coli. Recombinant cells were grown on a 1-liter scale, expression of the target P. furiosus ORF was induced, and cells were harvested and broken as described previously (1, 20, 39), except that cell extracts were heated at 80°C for 30 min to precipitate E. coli proteins (or unstable recombinant proteins), cooled to 4°C, and then clarified by centrifugation (40,000 x g). Recombinant proteins contained an N-terminal hexa-His tag and each was purified using a column (5 ml) of Histrap Ni affinity resin controlled with an AKTA explorer (GE Healthcare, Piscataway, NJ). After applying the cell extract, the column was washed with 5 column volumes of 20 mM phosphate buffer, pH 7.0, containing 500 mM NaCl, 10 mM imidazole, 5% (vol/vol) glycerol, and 2 mM dithiothreitol. The absorbed protein was eluted with a gradient of 0 to 500 mM imidazole over 20 column volumes.

The major protein peak was collected and concentrated to 10 ml by ultrafiltration (Millipore, Bedford, MA), diluted 15-fold in 20 mM Tris buffer, pH 8.0, containing 5% (vol/vol) glycerol and 2 mM dithiothreitol, and then applied to a column (5 ml) of Q Sepharose (GE Healthcare). The column was washed with 5 column volumes of the same buffer, and the bound protein was eluted with a 0 to 2 M NaCl gradient over 20 column volumes. The major protein was concentrated to 1.5 ml and applied to a 16/60 column of Superdex75 (GE Healthcare) equilibrated with the same Tris buffer. The major protein peak was collected and concentrated to a volume of 1 ml by ultrafiltration. The protein concentration was estimated by the absorption at 280 nm using a calculated extinction coefficient (15). The purity of the recombinant protein was determined by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) (4 to 20% Criterion gels, Bio-Rad, Hercules, CA), and its identity was determined by tryptic digestion of the excised band and analysis by matrix-assisted laser desorption ionization time of flight mass spectrometry (MALDI-TOF-MS). Protein mass was determined by electrospray ionization mass spectrometry at the Department of Chemistry, University of Georgia.


arrow
RESULTS AND DISCUSSION
 
Original annotation of the P. furiosus genome. The original annotation deposited in GenBank in 2002 contained 2,065 ORFs, labeled PF0001 to PF2065. Of these, 1,041 were designated as encoding proteins of known function, meaning that they showed some sequence similarity to proteins (or genes) that had been characterized to some extent in other organisms. In addition, 869 were annotated as conserved hypothetical, showing sequence similarity to ORFs of unknown function in the genomes of other organisms. The remaining 155 ORFs were considered unique to P. furiosus, having no similarity to any other ORFs at the time of deposition, including those in the closely related genomes of P. horikoshii (23) and P. abyssi (10). Current analyses of the 2,065 ORFs using the InterPro and PFAM databases reveal that 473 of the 869 conserved hypothetical ORFs are of "known" function, or at least some general functional information is available for them. Similarly, 25 of the 155 unique ORFs are no longer unique to P. furiosus and can now be reclassified either as of known function (24 ORFs) or as conserved hypothetical (1 ORF). Thus, at present, of the original 2,065 ORFs in the genome, 1,538 (75%) are known, 397 (19%) are conserved hypothetical, and 130 (6%) are unique to P. furiosus.

Prior to the GenBank deposition, a draft of the annotated P. furiosus genome was made publicly available (http://comb5-156.umbi.umd.edu/genemate) (31) that contained an additional 127 ORFs. These ORFs were subsequently removed from the annotation that was deposited in GenBank in 2002 but have been maintained in our own database. The annotation algorithms used at the time of deposition excluded them presumably based on their lack of sequence homologs and/or small size (see below). For the purposes of distinguishing them from ORFs in other databases, they will be referred to as the 127 University of Georgia (UGa) ORFs. Thus, the UGa annotation includes the 2,065 original GenBank ORFs plus another 127 hypothetical (UGa) ORFs, for a total of 2,192 ORFs. All putative ORFs that were not present in the original GenBank annotation of 2,065 ORFs will be referred to as ExGB ORFs (Table 1), and all 127 UGa ORFs fall into this category.


View this table:
[in this window]
[in a new window]
  		TABLE 1. Comparison of three P. furiosus annotationsa

Some properties of the UGa ORFs, including start and stop nucleotides and intergenic distances, are given as supplemental information (Table S1 in the supplemental material). They are numbered with reference to the lowest consecutive ORF in the GenBank annotation (e.g., PF0438) and are designated as PFxxxx0.1 (e.g., PF0438.1, and PF0438.2 if necessary). They are predicted to encode proteins ranging from 14 to 371 amino acids, where the median length is 49 amino acids. Only 13 of the 127 UGa ORFs show sequence similarity (≥80% match length, and E value ≤1 x e–06) to regions in other genomes. Two of them are only found at the nucleotide level, and the remaining eleven are annotated as ORFs in those other genomes. Of those eleven, six are genus specific, as they are found only in the genomes of P. abyssi and P. horikoshii. Thus, comparative genomics provides some evidence that eleven of the 127 hypothetical UGa ORFs are true ORFs. These data are presented in the supplementary information (Table S1 in the supplemental material).

Recent annotations of the P. furiosus genome. After the annotated P. furiosus genome was deposited in GenBank, an automated analysis of the genome sequence was carried out by the CMR of TIGR using Glimmer2 (12) and TIGRFAM (17). This identified a total of 2,261 ORFs, which were numbered sequentially NT01PF0001 to NT01PF2261. Note that there is no correlation between these CMR ORF numbers and those used in the GenBank (PF0001 to PF2065). As shown in Table 1, of the original 2,065 GenBank ORFs, only 2,040 of them were recognized by CMR (such that they have an identical stop site), leaving an additional 221 ExGB ORFs in the CMR annotation. A second automated annotation of the P. furiosus genome has also become available, in addition to that of CMR. This was released (on 13 January 2004) by NCBI under the RefSeq project (29) and contained 2,125 ORFs. Of these, 2,065 are in common with those in GenBank (where they have at least an identical stop site). Hence, there are a total of 60 RefSeq ExGB ORFs.

Comparison of the original ORFs in the three genome annotations. The 2,065 ORFs described in the original annotation in GenBank have been widely used in subsequent studies of P. furiosus. Assuming that all these ORFs are indeed genuine, are they exactly the same in the subsequent automated annotations, and if not, how do they differ? As already noted, the CMR annotation does not recognize 25 of the 2,065 GenBank ORFs, or more precisely, the stop codons for those 25 ORFs are not so designated in CMR. Sixteen of these 25 discarded ORFs overlap with other CMR ORFs (see Table S2 in the supplemental material). RefSeq has not discarded any of the original GenBank ORFs. The stop sites of the remaining 2,040 ORFs in CMR do match those in GenBank and the same is true for all 2,065 GenBank ORFs identified in the RefSeq annotation (see Fig. S1 in the supplemental material). Where they differ, however, is at their start sites. These differences are likely due to the various prediction methods that are used, which range from automated analyses to more subjective manual selection (6, 12, 26, 28, 29).

Of the 2,065 original GenBank ORFs, there are 84 instances where RefSeq and GenBank disagree on the start nucleotides. A similar analysis between the GenBank and CMR annotations reveals that 552 of the 2,065 ORFs in the P. furiosus genome differ in their start nucleotides, and a comparison of the RefSeq to the CMR annotations reveals 589 discrepancies in the start sites. As shown diagrammatically in Fig. 1, in most cases these differences are not a matter of a codon or two. For example, more than 170 of the proteins predicted by CMR and RefSeq differ at the N termini by more than 25 amino acids. It is remarkable that 28% of the 2,065 ORFs differ in their start codons in one or more of the three publicly available annotations of the P. furiosus genome. Moreover, this problem does not appear to be unique to P. furiosus. For example, as shown in Table 2, the CMR and RefSeq annotations of the genomes of the bacterium Clostridium perfringens strain 13 (36) and of the archaeon Pyrobaculum aerophilum strain IM2 (14), both of which were chosen at random, show differences in almost 400 and 1,200 ORFs, respectively. Hence, the discrepancies in P. furiosus genome annotations illustrate a generic problem.



View larger version (24K):
[in this window]
[in a new window]
  		FIG. 1. Database comparisons for the P. furiosus genome. Differences are shown in the positions of start codons for the genes in the annotations in the RefSeq and CMR databases. For clarity, the zero value (where the two annotations agree) is not shown.


View this table:
[in this window]
[in a new window]
  		TABLE 2. Comparison of automated annotations for the genomes of two other organisms

One way to mitigate differences in annotation is for databases to provide versions numbers, dates, and change logs that can be easily seen by the end user. In addition, when alternative ORFs or start sites are possible, some indication of their potential existence and even likelihood could be presented. Obviously, experimental approaches must be used to validate annotations, at least at the present time. In the case of P. furiosus, we chose to experimentally assess the validity of a subset of the ORFs that were not present in original annotation of the genome.

The ORFs that were not part of the 2,065 ORFs in the original GenBank deposition (ExGB ORFs) comprise 127 UGa ORFs, 221 CMR ORFs, and 60 RefSeq ORFs. As summarized in Table 1, there are 35 ORFs in common. Of these, 23 are identical (having the same start and stop sites) and 12 are similar (having only the same stop site). Sixty-three of the 127 UGa ORFs are identical to CMR ORFs and 24 are identical to RefSeq ORFs (see Fig. S2 in the supplemental material). Conversely, 45 of the 127 UGa ORFs are exclusive to the UGa annotation and are not present in the CMR or RefSeq annotations, which contain 127 and 9 exclusive ORFs, respectively (see Fig. S2 in the supplemental material). Note that an exclusive ORF is defined as having start and stop codon that are not recognized as such in the other annotations, regardless of whether the ORF they define overlaps with any other ORFs in those annotations.

The number of exclusive ORFs underscores the differences between the various annotation programs. Nevertheless, one may conclude that there are potentially a maximum of 277 additional ORFs in the P. furiosus genome (Table 1), depending on the annotations that are used. For the purpose of this paper we have chosen to focus our validation efforts on a subset of the UGa ORFs. As will be demonstrated, virtually all of them that appear to be genuine ORFs are also annotated in the RefSeq and CMR versions of the genome.

Experimental evidence for expression of UGa ORFs in P. furiosus. DNA microarray analysis was used to investigate the validity of some of the 127 UGa ORFs, which were not recognized in the original GenBank annotation. A total of 61 of the UGa ORFs did not overlap with any of the GenBank ORFs and full-length PCR products were obtained for each of them. These were added to the DNA microarray containing the 2,065 GenBank ORFs (34), which were used to assess their expression in P. furiosus in cells grown under six different conditions, where each condition yielded between 8 and 12 data points per ORF. These conditions are important because they are known to cause major changes in the expression levels of a large number of genes, which may include the UGa ORFs. Note that the DNA arrays are used here to assess absolute gene expression rather than relative changes in gene expression.

Since the raw fluorescence intensities can be highly variable, a minimum likely signal intensity (MLSI) value was calculated (see Materials and Methods) where ORFs with values above 1,500 were considered to be expressed at a significant level. By this criterion, of the 61 UGa ORFs examined, 11 are expressed in one or more of the six growth conditions. In fact, as shown in Table 3, five of the 11 were expressed under more than one growth condition. Many of the 61 ORFs examined that are apparently not expressed are less than 150 bps in length (see Table S1 in the supplemental material), raising the possibility that they are expressed but that the corresponding cDNA is simply beyond detection under the conditions of the microarray experiment. Nevertheless, it does appear that 11 of the 61 UGa ORFs examined are expressed at sufficient levels to allow detection.


View this table:
[in this window]
[in a new window]
  		TABLE 3. Transcriptional analysis of 11 previously unannotated (UGa) ORFs

Properties of 11 ORFs expressed in P. furiosus. Some sequence properties of the 11 newly recognized ORFs are summarized in Table 4. They vary in size from 35 to 134 amino acids (4.0 to 15.4 kDa). Two of the 11 are predicted to have at least one transmembrane domain and a signal sequence, suggesting that they may be membrane associated or extracellular (data not shown). None of the 11 ORFs have significant sequence similarity to ORFs in other genomes. All of them have a typical start codon (ATG or GTG), and six have a predicted bacterial-type ribosome binding site within 15 nucleotides upstream from the start codon (30). All but one of the 11 ORFs is also present in the CMR annotation (Table 4), eight of which are identical. The other three (PF0031.1, PF1112.1, and PF1735.2) differ in their start nucleotides. In RefSeq, five of the 11 ORFs have identical start and stop codons, and two have the same stop but a different start codon.


View this table:
[in this window]
[in a new window]
  		TABLE 4. Properties of the 11 previously unannotated ORFs expressed in P. furiosus

In E. coli it is known that consecutive ORFs on the same strand with small intergenic distances (~18 nucleotides or less) are likely to be part of an operon (33). Most of the 11 new ORFs are well separated from any neighboring ORFs by at least 50 nucleotides, but there are notable exceptions. One is PF0897.1 which is located only five nucleotides upstream of a possible ATP-binding cassette (ABC) transporter (11). As shown in Fig. 2, this comprises PF0895, the putative transporter, and two hypothetical proteins, PF0896 and PF0897, that are each predicted to contain ten transmembrane domains. PF0897.1 is predicted to have a signal sequence (data not shown). Interestingly, P. furiosus contains paralogues of all four of these genes in another possible ABC transporter (PF1090 to PF1093, see Fig. 2), where PF0897.1 shows high sequence similarity to PF1093.



View larger version (17K):
[in this window]
[in a new window]
  		FIG. 2. ORF PF0897.1 is potentially part of an operon. A potential ABC transporter that is closely associated with a previously unannotated (UGa) ORF (striped) is aligned with part of a known P. furiosus ABC transporter. Both are within 16 nucleotides of the neighboring ORF. Overlapping ORFs are indicated by the overlapping nature of the arrows. PF numbers and functional annotations are given to those ORFs that have been previously annotated by GenBank. Percent similarity is indicated by the numbers between each gene pair.

Recombinant protein production using P. furiosus ORFs. In addition to expression in the native organism, it should be possible to obtain an indication of the validity of an ORF assignment by attempting to obtain a stable recombinant form of the protein that it encodes. It would seem unlikely that an incorrect ORF would generate an artifactual protein that would be resistant to proteolysis by the host organism. It seems even less likely that an artifactual protein would fold into a thermally stable protein of the type found in a hyperthermophilic organism, such as P. furiosus. Therefore, any putative ORF in the P. furiosus genome that encoded a heat-stable recombinant protein was assumed to be genuine. Thus, a total of 54 out of the 127 UGa ORFs were cloned, and attempts were made to express them in E. coli and characterize the recombinant proteins. These included all 11 of the UGa ORFs discussed above that are expressed in P. furiosus, as determined by the microarray analysis (see Table 3). They also included 27 of the 66 UGa ORFs that were not investigated by the microarray approach because they overlapped an original ORF.

Of the 54 UGa ORFs investigated, seven of them yielded detectable amounts of recombinant protein on an SDS-PAGE gel after the Ni affinity purification step (data not shown). The 47 UGa ORFs that did not yield protein products included 10 that were analyzed by the DNA microarray (none of which were expressed in P. furiosus). Of course, one cannot draw any conclusion about the validity of an ORF from the absence of recombinant protein, as this is frequently the case with ORFs that encode well-characterized proteins (1). Conversely, the production of heat-stable recombinant proteins would strongly suggest that the seven new (UGa) ORFs are genuine and that they are expressed by P. furiosus.

All seven recombinant proteins were readily purified from heat-treated E. coli cell extracts by multistep chromatography, although all but one (PF0706.1) yielded multiple bands after SDS-PAGE analysis (data not shown). Each major band was excised, digested with trypsin and analyzed by MALDI-TOF-MS, which confirmed the identity of each band. Moreover, analysis by electrospray ionization mass spectrometry confirmed that all but one of the recombinant proteins (PF0712.1) had the predicted mass (Table 5). PF0712.1 is smaller than the predicted mass of 10,078 Da by 920 Da, indicating that the protein may be proteolytically degraded in E. coli. Hence, all seven recombinant proteins are stable and can be purified from E. coli and in all but one case are of the expected size.


View this table:
[in this window]
[in a new window]
  		TABLE 5. Properties of the seven previously unannotated ORFs that yield stable recombinant proteins

The CMR annotation contained all but one (PF0712.1) of the seven UGA ORFs that yielded stable recombinant protein. However, as shown in Table 5, only four of six ORFs that CMR does contain have identical start and stop sites, while the remaining two have a different start site. The RefSeq annotation contains five of the seven ORFs, four of which have identical start and stop sites. Two of the seven ORFs also have homologs in P. abyssi and P. horikoshii, further confirming the validity of the annotation. All seven are predicted to encode cytoplasmic proteins, and all but one of them are predicted to encode acidic proteins (Table 5). As shown in Fig. 3, one of the seven ORFs, PF0355.1, appears to be part of a three ORF operon because of its close proximity to two previously annotated ORFs. Interestingly, P. furiosus also contains a three-ORF paralog of this putative operon, including an ORF with a high similarity to PF0355.1 (Fig. 3).



View larger version (17K):
[in this window]
[in a new window]
  		FIG. 3. ORF PF0355.1 is potentially part of an operon. A potential oligosaccharide-related operon with a previously unannotated (UGa) ORF (striped) is aligned with another conserved oligosaccharide- related operon in P. furiosus. See the legend to Fig. 2 for more details.

Conclusions. Of the 127 previously unannotated ORFs in the P. furiosus genome, 61 were examined for expression in P. furiosus via DNA microarray analysis, and attempts were made to express 54 of them in E. coli (27 of the 54 were examined by both approaches). A total of 17 of the 88 ORFs examined (61 by array and 27 by expression) are likely to be functional in P. furiosus. Seven of them encode heat-stable proteins that could be obtained in a recombinant form, and 11 are expressed in P. furiosus under one or more growth conditions (which includes only one of the UGa ORFs that also generates a stable recombinant protein even though expression of all them was attempted). The remaining 39 of the 127 UGa ORFs were not studied due to overlapping sequences and lack of resources. Seventeen previously unannotated ORFs can therefore be added to the 2065 ORFs in original annotation of this genome.

Of considerable concern, however, is the fact that more than 25% of these 2,065 ORFs have ambiguous start sites according to the public databases. Primer extension and quantitative PCR experiments will be required to ascertain the true nature of these ORFs in order to provide a more accurate and complete picture of the P. furiosus proteome. Knowledge of correct start sites is essential for understanding gene overlap and regulation, and for practical concerns such as identification of natively purified proteins by N-terminal amino acid sequencing, as well as heterologous protein production. Addition of excess residues to the N terminus may well affect protein folding and cause proteolytic degradation by the host. What cannot be overemphasized is that this phenomenon is not unique to P. furiosus, as shown by cursory analyses of the genomes of both Clostridium perfringens strain 13 (36) and Pyrobaculum aerophilum strain IM2 (14), which indicate similarly large differences in start sites between annotations, as well as different numbers of genes (compare Tables 1 and 2).

There can be no doubt that all genome annotations need to be examined carefully, both for the accuracy of existing ORFs, and for the possibility of currently unannotated ORFs. Confirming any conclusions experimentally represents a major challenge particularly on a genome-wide scale, and particularly for ORFs exclusive to a given genome, many of which are likely to be currently unannotated. Adding to this burden is the constant discovery of novel ORFs, often exclusive to one genome, by the plethora of different automated procedures that are being developed. Moreover, currently unannotated ORFs are potentially of immediate biological significance. For example, we show here that the expression of one of the 17 novel P. furiosus ORFs is dramatically up-regulated by maltose, and its protein product presumably plays a key role in maltose metabolism, a process that would appear to be well understood. Thus, as more and more genomes become sequenced, it is imperative that ORFs such as this not be lost in the ever-expanding world of sequence space, particularly since these ORFs may well be the very essence of species individuality and represent novel biochemistry.


arrow
ACKNOWLEDGMENTS
 
We thank James F. Holden, Matthew S. Eckman, and William A. Lancaster for their assistance throughout the course of this study and Angeli Lal Menon, Frank J. Sugar, Jerry Eichler, and Dawei Lin for their many helpful discussions.

This work was supported by grants from the National Institutes of Health (GM62407), the Department of Energy (DE-FG02-05ER15710), and the National Science Foundation (MCB 0129841 and BES-0317911).


arrow
FOOTNOTES
 
* Corresponding author. Mailing address: Department of Biochemistry and Molecular Biol., Davison Life Sciences Complex, University of Georgia, Athens, GA 30602-7229. Phone: 706 542-2060. Fax: 706 542-0229. E-mail: adams{at}bmb.uga.edu. Back

{dagger} Supplemental material for this article may be found at http://jb.asm.org/. Back


arrow
REFERENCES
 
    1
  1. Adams, M. W. W., H. A. Dailey, L. J. Delucas, M. Luo, J. H. Prestegard, J. P. Rose, and B. C. Wang. 2003. The Southeast Collaboratory for Structural Genomics: A high-throughput gene to structure factory. Accounts Chem. Res. 36:191-198.
    2. 2
  2. Albers, S. V., S. M. Koning, W. N. Konings, and A. J. Driessen. 2004. Insights into ABC transport in archaea. J. Bioenerg. Biomembr. 36:5-15.[CrossRef][Medline]
    3. 3
  3. Altschul, S. F., T. L. Madden, A. A. Schaffer, J. Zhang, Z. Zhang, W. Miller, and D. J. Lipman. 1997. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25:3389-3402.[Abstract/Free Full Text]
    4. 4
  4. Azad, R. K., and M. Borodovsky. 2004. Probabilistic methods of identifying genes in prokaryotic genomes: connections to the HMM theory. Brief Bioinform. 5:118-130.[Abstract/Free Full Text]
    5. 5
  5. Bateman, A., E. Birney, L. Cerruti, R. Durbin, L. Etwiller, S. R. Eddy, S. Griffiths-Jones, K. L. Howe, M. Marshall, and E. L. Sonnhammer. 2002. The Pfam protein families database. Nucleic Acids Res. 30:276-280.[Abstract/Free Full Text]
    6. 6
  6. Besemer, J., A. Lomsadze, and M. Borodovsky. 2001. GeneMarkS: a self-training method for prediction of gene starts in microbial genomes. Implications for finding sequence motifs in regulatory regions. Nucleic Acids Research. 29:2607-2618.[Abstract/Free Full Text]
    7. 7
  7. Bocs, S., S. Cruveiller, D. Vallenet, G. Nuel, and C. Medigue. 2003. AMIGene: annotation of microbial genes. Nucleic Acids Res. 31:3723-3726.[Abstract/Free Full Text]
    8. 8
  8. Brachat, S., F. S. Dietrich, S. Voegeli, Z. H. Zhang, L. Stuart, A. Lerch, K. Gates, T. Gaffney, and P. Philippsen. 2003. Reinvestigation of the Saccharomyces cerevisiae genome annotation by comparison to the genome of a related fungus: Ashbya gossypii. Genome Biol. 4:R45.[CrossRef][Medline]
    9. 9
  9. Chen, L. L., and C. T. Zhang. 2003. Gene recognition from questionable ORFs in bacterial and archaeal genomes. J. Biomol. Struct. Dyn. 21:99-109.[Medline]
    10. 10
  10. Cohen, G. N., V. Barbe, D. Flament, M. Galperin, R. Heilig, O. Lecompte, O. Poch, D. Prieur, J. Querellou, R. Ripp, J. C. Thierry, J. Van der Oost, J. Weissenbach, Y. Zivanovic, and P. Forterre. 2003. An integrated analysis of the genome of the hyperthermophilic archaeon Pyrococcus abyssi. Mol. Microbiol. 47:1495-1512.[CrossRef][Medline]
    11. 11
  11. Davidson, A. L., and J. Chen. 2004. ATP-binding cassette transporters in bacteria. Annu. Rev. Biochem. 73:241-268.[CrossRef][Medline]
    12. 12
  12. Delcher, A. L., D. Harmon, S. Kasif, O. White, and S. L. Salzberg. 1999. Improved microbial gene identification with GLIMMER. Nucleic Acids Res. 27:4636-4641.[Abstract/Free Full Text]
    13. 13
  13. Fiala, G., and K. O. Stetter. 1986. Pyrococcus furiosus sp-nov represents a novel genus of marine heterotrophic archaebacteria growing optimally at 100-degrees C. Arch. Microbiol. 145:56-61.[CrossRef]
    14. 14
  14. Fitz-Gibbon, S. T., H. Ladner, U. J. Kim, K. O. Stetter, M. I. Simon, and J. H. Miller. 2002. Genome sequence of the hyperthermophilic crenarchaeon Pyrobaculum aerophilum. Proc. Natl. Acad. Sci. USA 99:984-989.[Abstract/Free Full Text]
    15. 15
  15. Gill, S. C., and P. H. von Hippel. 1989. Calculation of protein extinction coefficients from amino acid sequence data. Anal. Biochem. 182:319-326.[CrossRef][Medline]
    16. 16
  16. Guo, F. B., H. Y. Ou, and C. T. Zhang. 2003. ZCURVE: a new system for recognizing protein-coding genes in bacterial and archaeal genomes. Nucleic Acids Res. 31:1780-1789.[Abstract/Free Full Text]
    17. 17
  17. Haft, D. H., B. J. Loftus, D. L. Richardson, F. Yang, J. A. Eisen, I. T. Paulsen, and O. White. 2001. TIGRFAMs: a protein family resource for the functional identification of proteins. Nucleic Acids Res. 29:41-43.[Abstract/Free Full Text]
    18. 18
  18. Holden, J. F., F. L. Poole, S. L. Tollaksen, C. S. Giometti, H. Lim, J. R. Yates, and M. W. W. Adams. 2001. Identification of membrane proteins in the hyperthermophilic archaeon Pyrococcus furiosus using proteomics and prediction programs. Comp. Functional Genomics 2:275-288.[CrossRef]
    19. 19
  19. Iliopoulos, I., S. Tsoka, M. Andrade, P. Janssen, B. Audit, A. Tramontano, A. Valencia, C. Leroy, C. Sander, and C. Ouzounis. 2000. Genome sequences and great expectations. Genome Biol. 2:interactions0001.1-interactions0001.3.
    20. 20
  20. Jenney, F. E., Jr., P. S. Brereton, M. Izumi, F. L. Poole, 2nd, C. Shah, F. J. Sugar, H. S. Lee, and M. W. Adams. 2005. High-throughput production of Pyrococcus furiosus proteins: considerations for metalloproteins. J. Synchrotron Radiat. 12:8-12.[Medline]
    21. 21
  21. Jungblut, P. R., E. C. Muller, J. Mattow, and S. H. E. Kaufmann. 2001. Proteomics reveals open reading frames in Mycobacterium tuberculosis H37Rv not predicted by genomics. Infect. Immun. 69:5905-5907.[Abstract/Free Full Text]
    22. 22
  22. Kawarabayasi, Y. 2001. Genome of Pyrococcus horikoshii OT3. Methods Enzymol. 330:124-134.[Medline]
    23. 23
  23. Kawarabayasi, Y., M. Sawada, H. Horikawa, Y. Haikawa, Y. Hino, S. Yamamoto, M. Sekine, S. Baba, H. Kosugi, A. Hosoyama, Y. Nagai, M. Sakai, K. Ogura, R. Otsuka, H. Nakazawa, M. Takamiya, Y. Ohfuku, T. Funahashi, T. Tanaka, Y. Kudoh, J. Yamazaki, N. Kushida, A. Oguchi, K. Aoki, and H. Kikuchi. 1998. Complete sequence and gene organization of the genome of a hyperthermophilic archaebacterium, Pyrococcus horikoshii OT3. DNA Res. 5:55-76.[Abstract]
    24. 24
  24. Kellis, M., N. Patterson, M. Endrizzi, B. Birren, and E. S. Lander. 2003. Sequencing and comparison of yeast species to identify genes and regulatory elements. Nature 423:241-254.[CrossRef][Medline]
    25. 25
  25. Larsen, T. S., and A. Krogh. 2003. EasyGene-a prokaryotic gene finder that ranks ORFs by statistical significance. BMC Bioinformatics 4:21.[CrossRef][Medline]
    26. 26
  26. National Library of Medicine. October 2002. The NCBI handbook. National Library of Medicine, Bethesda, Md.
    27. 27
  27. Ouyang, Z., H. Zhu, J. Wang, and Z. S. She. 2004. Multivariate entropy distance method for prokaryotic gene identification. J. Bioinform. Comput. Biol. 2:353-373.[CrossRef][Medline]
    28. 28
  28. Peterson, J. D., L. A. Umayam, T. Dickinson, E. K. Hickey, and O. White. 2001. The comprehensive microbial resource. Nucleic Acids Res. 29: 123-125.[Abstract/Free Full Text]
    29. 29
  29. Pruitt, K. D., and D. R. Maglott. 2001. RefSeq and LocusLink: NCBI gene-centered resources. Nucleic Acids Res. 29:137-140.[Abstract/Free Full Text]
    30. 30
  30. Ringquist, S., S. Shinedling, D. Barrick, L. Green, J. Binkley, G. D. Stormo, and L. Gold. 1992. Translation initiation in Escherichia-coli-sequences within the ribosome-binding site. Mol. Microbiol. 6:1219-1229.[Medline]
    31. 31
  31. Robb, F. T., D. L. Maeder, J. R. Brown, J. DiRuggiero, M. D. Stump, R. K. Yeh, R. B. Weiss, and D. M. Dunn. 2001. Genomic sequence of hyperthermophile, Pyrococcus furiosus: implications for physiology and enzymology. Methods Enzymol. 330:134-157.[Medline]
    32. 32
  32. Rudd, K. E., I. Humphery-Smith, V. C. Wasinger, and A. Bairoch. 1998. Low molecular weight proteins: A challenge for post-genomic research. Electrophoresis 19:536-544.[CrossRef][Medline]
    33. 33
  33. Salgado, H., G. Moreno-Hagelsieb, T. F. Smith, and J. Collado-Vides. 2000. Operons in Escherichia coli: genomic analyses and predictions. Proc. Natl. Acad. Sci. USA 97:6652-6657.[Abstract/Free Full Text]
    34. 34
  34. Schut, G. J., S. D. Brehm, S. Datta, and M. W. Adams. 2003. Whole-genome DNA microarray analysis of a hyperthermophile and an archaeon: Pyrococcus furiosus grown on carbohydrates or peptides. J. Bacteriol. 185:3935-3947.[Abstract/Free Full Text]
    35. 35
  35. Schut, G. J., J. Zhou, and M. W. Adams. 2001. DNA microarray analysis of the hyperthermophilic archaeon Pyrococcus furiosus: evidence for a new type of sulfur-reducing enzyme complex. J. Bacteriol. 183:7027-7036.[Abstract/Free Full Text]
    36. 36
  36. Shimizu, T., K. Ohtani, H. Hirakawa, K. Ohshima, A. Yamashita, T. Shiba, N. Ogasawara, M. Hattori, S. Kuhara, and H. Hayashi. 2002. Complete genome sequence of Clostridium perfringens, an anaerobic flesh-eater. Proc. Natl. Acad. Sci. USA 99:996-1001.[Abstract/Free Full Text]
    37. 37
  37. Shockley, K. R., D. E. Ward, S. R. Chhabra, S. B. Conners, C. I. Montero, and R. M. Kelly. 2003. Heat shock response by the hyperthermophilic archaeon Pyrococcus furiosus. Appl. Environ. Microbiol. 69:2365-2371.[Abstract/Free Full Text]
    38. 38
  38. Snyder, M., and M. Gerstein. 2003. Genomics. Defining genes in the genomics era. Science 300:258-260.[Abstract/Free Full Text]
    39. 39
  39. Sugar, F. J., F. E. Jenney, Jr., F. L. Poole, 2nd, P. S. Brereton, M. Izumi, C. Shah, and M. W. Adams. 2005. Comparison of small- and large-scale expression of selected Pyrococcus furiosus genes as an aid to high-throughput protein production. J. Struct. Funct. Genomics, in press.
    40. 40
  40. Weinberg, M. V., G. J. Schut, S. Brehm, S. Datta, and M. W. Adams. 2005. Cold shock of a hyperthermophilic archaeon: Pyrococcus furiosus exhibits multiple responses to a suboptimal growth temperature with a key role for membrane-bound glycoproteins. J. Bacteriol. 187:336-348.[Abstract/Free Full Text]
    41. 41
  41. Wilkins, M. R., E. Gasteiger, A. Bairoch, J. C. Sanchez, K. L. Williams, R. D. Appel, and D. F. Hochstrasser. 1999. Protein identification and analysis tools in the ExPASy server. Methods Mol. Biol. 112:531-552.[Medline]
    42. 42
  42. Zdobnov, E. M., and R. Apweiler. 2001. InterProScan-an integration platform for the signature-recognition methods in InterPro. Bioinformatics 17:847-848.[Abstract/Free Full Text]
    43. 43
  43. Zhu, H. Q., G. Q. Hu, Z. Q. Ouyang, J. Wang, and Z. S. She. 2004. Accuracy improvement for identifying translation initiation sites in microbial genomes. Bioinformatics 20:3308-3317.[Abstract/Free Full Text]

Journal of Bacteriology, November 2005, p. 7325-7332, Vol. 187, No. 21
0021-9193/05/$08.00+0     doi:10.1128/JB.187.21.7325-7332.2005
Copyright © 2005, American Society for Microbiology. All Rights Reserved.



This article has been cited by other articles:

  • Davidsen, T., Beck, E., Ganapathy, A., Montgomery, R., Zafar, N., Yang, Q., Madupu, R., Goetz, P., Galinsky, K., White, O., Sutton, G. (2009). The comprehensive microbial resource. Nucleic Acids Res 0: gkp912v1-gkp912 [Abstract] [Full Text]  
  • Menon, A. L., Poole, F. L. II, Cvetkovic, A., Trauger, S. A., Kalisiak, E., Scott, J. W., Shanmukh, S., Praissman, J., Jenney, F. E. Jr., Wikoff, W. R., Apon, J. V., Siuzdak, G., Adams, M. W. W. (2009). Novel Multiprotein Complexes Identified in the Hyperthermophilic Archaeon Pyrococcus furiosus by Non-denaturing Fractionation of the Native Proteome. Mol. Cell. Proteomics 8: 735-751 [Abstract] [Full Text]  
  • Santangelo, T. J., Cubonova, L., Reeve, J. N. (2008). Shuttle Vector Expression in Thermococcus kodakaraensis: Contributions of cis Elements to Protein Synthesis in a Hyperthermophilic Archaeon. Appl. Environ. Microbiol. 74: 3099-3104 [Abstract] [Full Text]  
  • Hu, G.-Q., Zheng, X., Yang, Y.-F., Ortet, P., She, Z.-S., Zhu, H. (2008). ProTISA: a comprehensive resource for translation initiation site annotation in prokaryotic genomes. Nucleic Acids Res 36: D114-D119 [Abstract] [Full Text]  
  • Schut, G. J., Bridger, S. L., Adams, M. W. W. (2007). Insights into the Metabolism of Elemental Sulfur by the Hyperthermophilic Archaeon Pyrococcus furiosus: Characterization of a Coenzyme A- Dependent NAD(P)H Sulfur Oxidoreductase. J. Bacteriol. 189: 4431-4441 [Abstract] [Full Text]  
  • Delcher, A. L., Bratke, K. A., Powers, E. C., Salzberg, S. L. (2007). Identifying bacterial genes and endosymbiont DNA with Glimmer. Bioinformatics 23: 673-679 [Abstract] [Full Text]  
  • Tran, T. T., Dam, P., Su, Z., Poole, F. L. II, Adams, M. W. W., Zhou, G. T., Xu, Y. (2007). Operon prediction in Pyrococcus furiosus. Nucleic Acids Res 35: 11-20 [Abstract] [Full Text]  
  • White, R. H. (2006). The difficult road from sequence to function.. J. Bacteriol. 188: 3431-3432 [Full Text]  

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental material
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Poole, F. L.
Right arrow Articles by Adams, M. W. W.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Poole, F. L., II
Right arrow Articles by Adams, M. W. W.

Copyright © 2009 by the American Society for Microbiology.   For an alternate route to Journals.ASM.org, visit: http://intl-journals.asm.org | More Info»