Angiosperms (flowering plants), including both monocots and dicots, contain small catalase gene families. In the dicot, Arabidopsis thaliana, two catalase (CAT) genes, CAT1 and CAT3, are tightly linked on chromosome 1 and a third, CAT2, which is more similar to CAT1 than to CAT3, is unlinked on chromosome 4. Comparison of positions and numbers of introns among 13 angiosperm catalase genomic sequences indicates that intron positions are conserved, and suggests that an ancestral catalase gene common to monocots and dicots contained seven introns. Arabidopsis CAT2 has seven introns; both CAT1 and CAT3 have six introns in positions conserved with CAT2, but each has lost a different intron. We suggest the following sequence of events during the evolution of the Arabidopsis catalase gene family. An initial duplication of an ancestral catalase gene gave rise to CAT3 and CAT1. CAT1 then served as the template for a second duplication, yielding CAT2. Intron losses from CAT1 and CAT3 followed these duplications. One subclade of monocot catalases has lost all but the 5′-most and 3′-most introns, which is consistent with a mechanism of intron loss by replacement of an ancestral intron-containing gene with a reverse-transcribed DNA copy of a fully spliced mRNA. Following this event of concerted intron loss, the Oryza sativa (rice, a monocot) CAT1 lineage acquired an intron in a novel position, consistent with a mechanism of intron gain at proto-splice sites.
CATALASE (H2O2:H2O2 oxidoreductase; EC 126.96.36.199) dismutates H2O2 into water and oxygen. Together with superoxide dismutase and hydroperoxidase, catalase is part of a defense system for scavenging superoxide radicals and hydroperoxides (Beyer and Fridovich 1987). Catalase is used as a marker for peroxisomes, which are present in almost all eukaryotes (Subramani 1993), and seems to be ubiquitous: no multicellular organism that lacks catalase activity has been found (Scandalios 1987). The active catalase enzyme is a tetrameric iron porphyrin protein. The monomeric catalase subunits are encoded by single genes in most eukaryotes, including mammals (rats, mice, guinea pigs, and humans; Nakashimaet al. 1989; Quanet al. 1986; Shafferet al. 1990; Yaunet al. 1996), Drosophila (Orr et al. 1990, 1996), and several fungi (GenBank accession number Y07763; Didion and Roggenkamp 1992; Fowleret al. 1993; Nakagawaet al. 1995). There are, however, two catalase sequences in the nematode Caenorhabditis elegans (Waterstonet al. 1992; Wilsonet al. 1994), and in the fungi Saccharomyces cerevisiae (Cohenet al. 1988; Spevaket al. 1986), Candida tropicalis (Murray and Rachubinski 1987, 1989; Okadaet al. 1987), and Aspergillus nidulans (GenBank accession number U80672; Navarroet al. 1996).
In contrast to the situation in most animals and fungi, many plants encode catalase as multigene families (Frugoliet al. 1996; Guan and Scandalios 1996; Willekenset al. 1994b), which may reflect the multiple and diverse roles played by plant catalases. Plants use catalase in several pathways in addition to those common to other higher eukaryotes. Many oilseed plants store seed energy as lipids, which, upon germination, are converted to sugars through β-oxidation and the glyoxylate cycle (Beevers 1982). The first step in β-oxidation oxidizes a flavin that is subsequently reduced with the concomitant generation of H2O2, which then must be broken down by catalase (Trelease 1984). The requirements for rapid and massive flux through β-oxidation and the glyoxylate cycle during germination of oilseed plants suggests that catalase activity should be essential. Catalase is also required in photorespiration, the light-dependent evolution of CO2 resulting from the oxygenation, as opposed to the carboxylation, of ribulose-1,5-bisphosphate catalyzed by the bifunctional enzyme ribulose-1,5-bisphosphate carboxylase/oxygenase (Canvin 1990). Catalase is required to dismutate H2O2 produced during the peroxisomal oxidation of glycolate to glyoxylate, an intermediate step in the photorespiratory pathway. Photorespiratory catalase activity is essential, and mutants that lack catalase activity are inviable in conditions under which photorespiration occurs (Kendallet al. 1983; Somerville and Ogren 1982).
In Arabidopsis, there are three catalase (CAT) genes that encode subunits of six to seven detectable tetrameric isozymes (Frugoliet al. 1996). We have taken advantage of the power of Arabidopsis as a model system (Meyerowitz and Somerville 1994) to explore the regulation and function of the three catalase genes (McClung 1997). For example, the three CAT genes show distinct organ-specific patterns of expression; CAT1 and CAT3 mRNAs are most abundant in bolts and leaves, whereas CAT2 mRNA is most abundant in leaves. All three subunit mRNAs, as well as multiple catalase isozymes, however, are detected in each organ examined, from two isozymes in roots to as many as seven in flowers (Frugoliet al. 1996). The three Arabidopsis CAT genes also respond differently to light: when dark-adapted plants are returned to light, CAT1 mRNA is weakly induced and CAT2 mRNA is strongly induced, whereas CAT3 mRNA is not induced (E. L. Connolly, H. H. Zhong, R. M. Learned and C. R. McClung, unpublished observations). Furthermore, the expression of two of the CAT genes is gated by the circadian clock to distinct times of day: mRNA abundance is maximal at dawn for CAT2 and is maximal at dusk for CAT3 (Zhong and McClung 1996; Zhonget al. 1994). The expression of the rhythm in CAT3 mRNA abundance in extended dark is regulated by signaling through both phytochrome and cryptochrome pathways (Zhonget al. 1997), while CAT1 mRNA has no apparent circadian rhythm in abundance (Frugoliet al. 1996).
In this study, we address the evolutionary relationships among these three Arabidopsis catalase genes. Phylogenetic analysis based on the amino acid sequence of catalase has suggested that two major groups of catalases are derived from different prokaryotic ancestors, and that plant catalases arose independently of animal and fungal catalases (von Ossowskiet al. 1993). We focus on angiosperm catalases and provide a phylogenetic analysis that considers more than twice as many catalases as previous analyses have (Guan and Scandalios 1996; Willekenset al. 1994a). We suggest that the evolution of this multigene family entailed a series of gene duplications of an ancestral angiosperm catalase that, before the divergence of monocots and dicots, contained seven introns. In Arabidopsis, analysis of sequence similarity and genetic linkage allows us to postulate a sequential order of duplication events during the evolution of the CAT gene family, as well as a pattern of intron loss consistent with this order of events. The pattern of intron loss seen in the monocot lineage suggests a mechanism of concerted intron loss that is concordant with the replacement of an ancestral intron-containing gene with a reverse-transcribed DNA copy of a fully spliced mRNA, consistent with the mechanism suggested by Baltimore (1985) and Fink (1987). After this event of concerted intron loss, one rice catalase acquired an intron in a novel position. Comparison of sequences surrounding the site of intron addition within this clade supports a mechanism, first proposed by Dibb and Newman (1989), of intron gain at proto-splice sites.
MATERIALS AND METHODS
Plant growth conditions: Arabidopsis thaliana and A. griffithiana plants were grown in constant light (130 μmol·m−2·s−1 photosynthetically active radiation) at 20°, harvested, frozen in liquid nitrogen, and stored at −80°.
Southern analysis: Southern analysis was by standard protocols (Ausubelet al. 1997) using Nytran Plus (Schleicher & Schuell, Keene, NH) or Hybond-N+ (Amersham, Arlington Heights, IL) membranes. Hybridization and washes were as described (Choiet al. 1995). Probes were made by excision of gene-specific DNA fragments (Frugoliet al. 1996) from agarose gels after electrophoresis, purification by Qiaquick gel purification kits (Qiagen, Chatsworth, CA), and random primer labeling with [α-32P] dATP using the Klenow fragment of DNA polymerase I (Feinberg and Vogelstein 1984). Membranes were then wrapped in Saran wrap, and autoradiographs were generated on Bio-Max film (Eastman Kodak, Rochester, NY) in autoradiography cassettes at −80° for 1–3 days, depending on the blot.
Bacterial artificial chromosome (BAC) analysis: Two sets of Texas A&M University BAC filters (Choiet al. 1995) were hybridized with probes for CAT1 and CAT3 (Frugoliet al. 1996), and four clones that hybridized to both genes (T10F14, T5B3, T2C19, and T4J5) were analyzed further. The BAC DNA was isolated and digested with NotI to release the inserts. CHEF gel analysis (Choiet al. 1995) indicated that the inserts ranged from ~40 kb to >150 kb in size. The 40-kb insert of clone T10F14 was chosen for further analysis.
CAT1 genomic sequence: Restriction fragments from TAMU BAC T10F14 digested either with BamHI and/or with XbaI were subcloned (Ausubelet al. 1997) into pBluescript KS (Stratagene, La Jolla, CA). The CAT1- and CAT3-hybridizing fragments were sequenced by dideoxy chain termination using the ABI Dye Terminator Cycle Sequencing kit (Applied Biosystems, Foster City, CA) and various primers designed within the gene and the vector polylinker. The reaction products were run on ABI model 373A and 377 sequenators, and sequences were viewed with ABI Prism View software (Applied Biosystems). The sequenced fragments of clones from both digests were assembled with Geneworks (IntelliGenetics, Mountain View, CA) and various programs of the GCG package (version 8; Genetics Computer Group, Madison, WI), and the resulting sequence was deposited in GenBank under accession number AF021937.
Polymerase Chain Reactions: For confirmation of the CAT3/CAT1 locus structure, 100 ng of DNA from each BAC was used in a 50-μl PCR reaction with 25 nm each of primer 1/3 (5′-ATGGATCCATGATGCTTGAAGAC-3′, corresponding to nt 69–91, using the numbering scheme of GenBank accession number U43340, or nt 4349–4327 according to the numbering scheme of GenBank accession number AF021937) , and 3/1 (5′-AAGGATCCTCACATGTGTTGTGT-3′, corresponding to nt 3806–3828, using the numbering scheme of GenBank accession number AF021937), 100 μM dNTPs, 2.5 units of Taq polymerase, and 3 mm MgCl2 in PCR reaction buffer (Promega, Madison, WI). Reaction steps were 5 min initial denaturation at 94°, followed by 35 cycles, each of 1 min at 94°, 30 sec at 55°, and 1 min at 72°, and a final elongation step of 5 min at 72°.
We wished to determine whether the absence of the last intron in CAT1 was conserved among ecotypes of A. thaliana (collected from North America, Europe, and Africa), as well as A. griffithiana, a related species from Tajikistan (Asia). The ecotypes (geographic origin in parentheses) for which data are presented include Be-0 (Germany), Bu-0 (Germany), Col-2 (United States), Est-0 (Russia), La-0 (Poland), Mh-0 (Poland), Ws-0 (Russia), and Ler-0, a laboratory strain derived from La-0 that contains a mutation at the erecta locus. In addition, we obtained similar results for the following ecotypes: Bu-0 (Germany), Cvi-0 (Cape Verde Islands), Le-0 (Netherlands), Ms-0 (Russia), Nd-0 (Germany), No-0 (Germany), Po-1 (Germany), RLD-1 and Sei-0 (Italy). These ecotypes originated at altitudes of 1–300 meters (Arabidopsis Biological Resource Center Catalog). For each ecotype or species, 50 ng of genomic DNA was amplified using 25 nm each of primers CAT1-11 [5′-GCGATATC-GGTCAATTACTTCCCTTCAAGG-3′, nt 1248–1269, using the numbering scheme of GenBank accession number U43340 (note that the first 8 nt do not correspond to the Arabidopsis genomic sequence and include an EcoRV site added to the primer to facilitate subsequent cloning steps)] and CAT1-12 [5′ GAGATGAATTCATTCAGAAGTTTGGCC-3′, nt 1573–1547, using the numbering scheme of GenBank accession number U43340 (note that mutations A1565T and T1566A have been included in the primer to yield an EcoRI site to facilitate subsequent cloning steps)], 100 μm dNTPs, 2.5 units of Taq polymerase, and PCR reaction buffer (Invitrogen, San Diego, CA). Reaction steps were 5 min initial denaturation at 92°, followed by 35 cycles, each of 30 sec at 92°, 30 sec at 50°, and 3 min at 72°, and a final elongation step of 5 min at 72°. Reaction products were analyzed by agarose gel electrophoresis.
For determination of Hordeum vulgare CAT1 and CAT2 intron positions, a series of primer pairs (Table 1) were designed to amplify, in a set of overlapping products, the complete genomic sequences of the two CAT genes from H. vulgare cv. Harrington genomic DNA (a gift from J. Sherwood). Amplification reactions included 50 ng barley genomic DNA, 100 μm dNTPs, 25 nM of each primer, 2.5 units of Taq polymerase, and PCR reaction buffer (Boehringer-Mannheim, Indianapolis, IN). Reaction steps were 5 min initial denaturation at 94°, followed by 25 cycles, each of 1 min at 94°, 1 min at 55°, and 1 min at 72°, and a final elongation step of 7 min at 72°. PCR products were either subcloned into plasmid pCR2.1 using the TA Cloning Kit (Invitrogen) and then sequenced as described above, or they were directly sequenced, following purification through Centri-Spin 40 columns (Princeton Separations, Adelphia, NJ). We were unable to amplify any CAT-related product in reactions using primers BCAT1-1F or BCAT2-1F, so we did not obtain a sequence that spanned the first or second intron for either H. vulgare CAT gene. Sequences of H. vulgare CAT1 and CAT2 were assembled and analyzed using the GCG programs, and they have been deposited in GenBank under accession numbers AF021938 and AF021939, respectively.
Phylogenetic analysis: We generated a hypothesis for the phylogenetic relationships among the gene sequences from the deduced amino acid sequences of Arabidopsis catalases and of other plant catalases in the GenBank database using cladistic analyses implemented by PAUP version 3.1.1 (Swofford 1993).
Physical characterization of the CAT3-CAT1 loci: As part of a comprehensive analysis of the catalase gene family of Arabidopsis, we wished to characterize the genomic structure of each of the three CAT genes. Low-resolution mapping showed that CAT1 and CAT3 are tightly linked on chromosome 1, whereas CAT2 is on chromosome 4 (Frugoliet al. 1996). To refine the relative chromosomal map positions of CAT1 and CAT3 and to determine the genomic structures of these two loci, we identified four BAC clones that hybridized to both CAT1 and CAT3. CHEF gel analysis determined that clone T10F14 contained an insert of only 40 kb (data not shown), so further analysis focused on this BAC. Restriction endonuclease digestions and subsequent DNA sequence analysis of BAC T10F14 generated the map of the loci shown in Figure 1A. CAT1 and CAT3 are immediately adjacent, with the most distal polyadenylation site of CAT3 (three sites at nt 3945, 3956, and 3959; GenBank accession number AF021937) only 277 bp upstream from the most proximal site of start of transcription of CAT1 (five sites at nt 4236, 4238, 4241, 4244, and 4246; GenBank accession number AF021937), which we determined by primer extension (data not shown). The two genes are transcribed in the same direction, although at this time we remain uncertain as to the orientation of the two genes relative to the telomere and centromere of the upper arm of chromosome 1.
We also sought to confirm that the tight linkage of CAT1 and CAT3 seen in BAC T10F14 did not reflect some rearrangement that occurred during the cloning procedure. First, we performed Southern analysis of Arabidopsis ecotype Columbia genomic DNA using as a hybridization probe; a BamHI fragment subcloned from TAMU BAC T10F14 (BamHI 6.1, 1555 nt, indicated by a hatched line in Figure 1A) that spanned the CAT3-CAT1 intergenic region (Figure 1B). BamHI 6.1 detected a single BamHI fragment of 1.55 kb in the genomic DNA. In addition, two EcoRI fragments of ~6 and 4 kb and two EcoRV fragments of 2.9 and 1.5 kb were detected, consistent with the map in Figure 1A and with the nucleotide sequence reported in GenBank accession number AF021937. Further confirmation was obtained by designing two oligonucleotide primers (Figure 1A and materials and methods), one in a previously sequenced region of CAT3 (Zhong and McClung 1996) and one in a previously sequenced region of the CAT1 cDNA (Frugoliet al. 1996). These primers were used in PCR reactions that used the other three TAMU BAC clones that hybridized to both CAT genes as templates; in each case, we observed the 549-bp product (Figure 1C) predicted by the map (Figure 1A) and by the nucleotide sequence.
Phylogenetic analysis of angiosperm catalases using amino acid sequences: Including the three Arabidopsis catalases, a total of 37 angiosperm catalase sequences, representing 22 species, as well as two Chlamydomonas catalase sequences, have been deposited in GenBank (as of November 8, 1997; Table 2). The amino acid sequences of these 39 catalases were aligned with the PILEUP program (version 8 Program Manual, GCG). We used the heuristic search algorithm (branch swapping with 20 replicates of random sequence additions) of PAUP ver. 3.1.1 (Swofford 1993) to generate a hypothesis for the phylogenetic relationships among these sequences. One tree of shortest length was found by this analysis (Figure 2A).
We also performed a bootstrap analysis (100 replicates) using this same algorithm. This bootstrap analysis strongly supports the phylogenetic affinities of a number of species groups (Figure 2B), but this analysis also indicates that the deep branching patterns identified in the tree of Figure 2A are not robust. This is probably a function of the limited number of genes from different putative catalase lineages that can be included in the current analysis. The topology of the trees in Figure 2, A and B, suggests, however, that the gene duplications giving rise to multiple plant catalases occurred before the divergence of monocots and dicots. The results of our analysis are generally consistent with the previous analyses of smaller sets of plant catalases (Guan and Scandalios 1996; Willekenset al. 1994a). Analysis of the increased number of grass (the only group of monocots represented) sequences suggests, however, that three subclades of grass catalases exist, each including one of the three known Zea mays catalases. Similarly, the three known Nicotiana plumbaginifolia catalases define three distinct subclades. One goal of this analysis was to correlate phylogenetic and functional groupings of catalases to provide testable hypotheses concerning the function of the three Arabidopsis catalases and catalases in general. The three Arabidopsis catalases define distinct lineages, but they do not form large subclades with other catalases. Furthermore, the limited number of sequences from representative taxa across the plant kingdom and the paucity of functional analyses of individual plant catalases undermine present attempts to correlate phylogeny and function.
Intron-exon structure of angiosperm CAT genes: In addition to the genomic sequences of the three Arabidopsis catalases, genomic sequences were available for eight other catalases (Table 2). For a more complete analysis of the grass lineages, we determined partial genomic sequences for the two barley catalase genes (deposited in GenBank under accession numbers AF021938 and AF021939). Among these 13 sequences, introns were observed at a total of eight positions, which we have numbered according to their position within the genes, with intron 1 closest to the 5′ end and intron 8 closest to the 3′ end. Numbers, positions, and sizes of the introns are summarized in Table 3 and Figure 3. All the observed introns interrupt the coding sequences, and the positions of the observed introns, relative to the coding sequence, are conserved among all angiosperm catalases, as was noted previously in an analysis of a smaller data set (Guan and Scandalios 1996). Because the majority of CAT sequences contain seven introns (1–3 and 5–8) in conserved positions, we suggest that an ancestral catalase gene common to monocots and dicots contained seven introns in these positions, and that the presence of intron 4 at a novel position in the Oryza sativa CAT1 sequence represents a derived character (see below). Several of the genes lack one or more introns (Table 3 and Figure 3); we suggest that this occurred by intron loss. Arabidopsis CAT1 lacks intron 8. We confirmed that the absence of this intron in CAT1 was conserved across ecotypes by PCR analysis (Figure 1D). Arabidopsis CAT3 and Ricinus communis CAT2 lack intron 7. Glycine max lacks intron 3. Z. mays Cat2 lacks both introns 3 and 5. Of particular interest is one subclade of grass catalases (Z. mays Cat3, O. sativa CAT1, and H. vulgare CAT2), each of which lacks introns at positions 3, 5, 6, and 7. The Z. mays and O. sativa sequences also lack the intron at position 2, although we were unable to amplify the 5′ regions of either of the H. vulgare catalases, so we have no information about the presence of introns 1 or 2 for either H. vulgare sequence. In addition, the O. sativa CAT1 sequence contains an intron, intron 4, at a novel position after the codon encoding amino acid residue Q272. These data are summarized in Figure 3, which includes only those sequences for which genomic sequence and, hence, intron positions are known.
The majority of studies of plant molecular evolution have focused on the chloroplast genome, and the molecular evolution of plant nuclear genes remains to be comprehensively addressed. Most studies of plant nuclear genes have examined gene families, such as those encoding the small subunit of ribulose-bisphosphate carboxylase/oxygenase (RBCS) or alcohol dehydrogenase (ADH), in which the protein product of the genes plays a defined biochemical role in a limited set of pathways (Clegget al. 1997). Plant actin genes comprise a large and complex gene family in which diversity in function is paralleled by gene family diversity that exceeds that found in other eukaryotes (Meagher 1991; McDowellet al. 1996). Plant catalases play diverse roles in germination, photorespiration, resistance to oxidative stress (McClung 1997), and possibly in mediating signal transduction involving H2O2 as a second messenger (Low and Merida 1996; Mehdyet al. 1996; Ryalset al. 1996; Yanget al. 1997). Therefore, the catalase genes, such as the actin genes (Meagher 1991) and the genes of the flavonoid pathway, notably chalcone synthase (Clegget al. 1997), offer a useful system in which to address how the acquisition of multiple metabolic roles influences the evolution of a gene family.
The evolution of multigene families involves multiple mechanisms (Ohta 1991; Fryxell 1996). Any explanation of the evolution of the Arabidopsis CAT family must take into account three pieces of data: the greater sequence similarity between CAT1 and CAT2 than of either with CAT3 (Frugoliet al. 1996), the tight linkage of CAT1 and CAT3 (Figure 1), and the pattern of intron losses (Table 3 and Figure 3). We propose the following sequence of events to account for the evolution of the Arabidopsis CAT gene family (Figure 4). The ancestral CAT gene, containing seven introns, was one of the two linked genes that now reside on chromosome 1 (Figure 4A), and an initial tandem duplication yielded progenitors of CAT3 and CAT1 (Figure 4B). A second, more recent duplication of the CAT1 sequence was associated with the integration of the duplicated copy on chromosome 4 (Figure 4C). We argue that CAT1 provided the template for this second duplication event because CAT1 and CAT2 are more similar in sequence to each other, at both the nucleotide and amino acid levels, than either is to CAT3 (Frugoliet al. 1996). The loss of intron 7 from CAT3 occurred sometime after the duplication that originally resulted in CAT1 and CAT3, and the loss of intron 8 from CAT1 followed the second duplication (Figure 4D).
Interestingly, a second dicot, Ricinus communis, also has two tightly linked (by ~2 kb) CAT genes that, like Arabidopsis CAT3 and CAT1, are transcribed in the same direction (Suzukiet al. 1994). This suggests that the tandem duplication was an early event that occurred before the separation of the Arabidopsis and Ricinus lineages. The upstream genes from each species, Arabidopsis CAT3 and R. communis CAT2, have lost intron 7; we infer that the loss of intron 7 in this gene lineage predated the divergence of the Arabidopsis and R. communis lineages. The two downstream genes, Arabidopsis CAT1 and R. communis CAT1, are more related to each other than to the other linked catalase (Arabidopsis CAT3 and R. communis CAT2, respectively). R. communis CAT1 retains intron 8, however, whereas Arabidopsis CAT1 has lost intron 8, suggesting that Arabidopsis CAT1 lost intron 8 after the separation of the lineages. One obvious question is why the tight linkage of these two catalase genes should have persisted over evolutionary time. It is possible that CAT1 and CAT3 retain tight linkage to facilitate coregulation, although mRNA expression patterns of these two genes are dissimilar (Frugoliet al. 1996). A better test of this hypothesis would be to determine whether CAT1 and CAT3 monomers assemble into mixed tetramers and thus share common functions, which would suggest common regulatory elements associated with these two genes. We are currently generating the monomer-specific antibodies that will allow us to address this question.
The topology of the trees in Figure 2, with Arabidopsis CAT1 in a subclade with Z. mays Cat2 and Arabidopsis CAT3 in a subclade with Z. mays Cat3, further suggests that the duplication that gave rise to the progenitors of CAT1 and CAT3 predated the divergence of monocots and dicots from a common ancestor. Although this is an intriguing interpretation, it must remain speculative because not all catalases from each of the taxa have been analyzed, and it is equally plausible that this branching pattern represents an artifact caused by the limited gene sampling used in this analysis.
Baltimore (1985) postulated that reverse transcription of a processed cellular mRNA can generate intronless cDNA copies of expressed genes that can be randomly integrated into the genome. The demonstration that the S. cerevisiae Ty elements transpose through an RNA intermediate (Boekeet al. 1985) established that retroelements normally present in the genome can provide a cellular source of reverse transcriptase; experimental demonstration that reverse transcription of a cellular mRNA provides a mechanism for intron loss when the cDNA copy replaces the endogenous genomic copy through homologous recombination has been provided in yeast (Derret al. 1991). Introns located at the gene termini are less likely to be replaced by such a mechanism than are introns in the middle of the gene because little of the cDNA intermediate would extend beyond the intron, providing a less efficient substrate for homologous recombination. Moreover, this is a simple mechanism by which contiguous blocks of introns can be lost in one event. This argument has been invoked to explain the asymmetric location (at the gene termini, usually the 5′ end) of those few introns retained in the S. cerevisiae genome (Fink 1987) and in the red algae (Liaudet al. 1995).
One particularly intriguing example of multiple intron loss apparently occurred in one subclade of grass catalase genes (Z. mays Cat3, O. sativa CAT1, and H. vulgare CAT2). This subclade was defined on the basis of amino acid sequence similarity (Figure 2), and is well supported by the bootstrap analysis. Members of this subclade apparently have lost introns 2, 3, 5, 6, and 7, but they retain the 5′-most and 3′-most introns (Figure 5). This pattern of intron loss is consistent with reverse transcription of a cellular RNA, followed by gene replacement by a homologous recombination event in which the recombination break points lie downstream of the 5′-most intron and upstream of the 3′-most intron. Such a mechanism of concerted loss of adjacent introns provides the most parsimonious explanation for the loss of multiple contiguous introns. Plants are rich in retroelements, which presumably could provide a cellular source of reverse transcriptase. For example, more than 60% of the DNA from a 280-kb region surrounding the maize Adh1-F locus represented retroelements (SanMiguelet al. 1996). Arabidopsis has more than 20 characterized retroelements, and there is evidence that at least some of these retroelements have been active since the founding of the Arabidopsis lineage and the divergence of the various ecotypes (Koniecznyet al. 1991; Voytas and Ausubel 1988; Wrightet al. 1996).
This mechanism of homologous recombination of a reverse-transcribed cDNA copy of a processed mRNA could also explain the loss of the adjacent introns 3 and 5 in the Z. mays Cat2 sequence (Figure 3). Although the related Z. mays Cat1 sequence also has lost intron 5, we infer that Z. mays Cat1 and Cat2 suffered independent losses of intron 5 (Figure 5). The most parsimonious explanation, based on the model of concerted intron loss, is that Z. mays Cat2 lost introns 3 and 5 simultaneously and that Z. mays Cat1 subsequently lost intron 5 (note that H. vulgare CAT1 retains intron 5). Within the angiosperm catalases, there are no examples of loss of two nonadjacent introns without the loss of intervening introns. This mechanism of reverse transcription followed by homologous recombination could also explain the loss of single introns from Arabidopsis CAT1 and CAT3, and has been invoked to explain the loss of single introns in potato and tomato actin genes (Drouin and Moniz de Sá 1997). Other examples of the loss of single introns from plant genes have been noted in a number of plant gene families (Huanget al. 1990; Kumar and Trick 1993; Hägeret al. 1996). Other mechanisms, however, may also have been responsible for these losses of single introns. We suggest that the examples of concerted loss of multiple, contiguous introns in the grass catalases provide stronger evidence for gene replacement with a reverse-transcribed cDNA as a mechanism of intron loss.
Within the subclade of grass catalases in which introns 2, 3, 5, 6, and 7 have been lost, one sequence, O. sativa CAT1, contains an intron in a novel position (intron 4; Figures 3 and 5). The simplest explanation is that the O. sativa CAT1 sequence gained an intron after the separation of the O. sativa lineage from the other grasses. The opposite explanation, the loss of intron 4 from all other plant catalases, would require multiple loss events in the individual lineages. Intron insertion has been postulated to have occurred in families of G protein genes (Dietmaier and Fabry 1994), in the triose-phosphate isomerase gene family (Kwiatkowskiet al. 1995; Logsdonet al. 1995), and in the RBCS gene family in the Solanaceae (Fritzet al. 1993). On the basis of an analysis of the actin and tubulin gene families, Dibb and Newman (1989) postulated that intron gain occurs between the G and R of the intron-flanking consensus sequence, C/AAGR, which they term a protosplice site. This consensus corresponds to the consensus exon/intron 5′ splice junction and, therefore, is observed at the site of existing introns. Such sequences, however, also may originate and persist in the absence of introns because of coding constraints (Dibb and Newman 1989). In each of the grass catalases, the sequence surrounding the potential intron 4 site is CAGR, which corresponds exactly to the consensus site, yet only the O. sativa CAT1 sequence has an intron at this position. We argue that this represents an example of intron gain at a proto-splice site.
We suggest that a mechanism of concerted intron loss in which reverse transcription of a cellular mRNA followed by partial gene replacement of the endogenous gene copy by an intronless cDNA is a relatively common event during genomic evolution. The evolution of the plant catalase gene family provides evidence for at least two independent events of intron loss by this mechanism. The gene duplications to create three genes occurred early in the evolution of angiosperms and preceded the split between monocots and dicots that occurred between 200 and 100 mya (Stewart and Rothwell 1993). Subsequently, one event of concerted intron loss of introns 2–7 occurred in one grass subclade that includes Z. mays Cat1, O. sativa CAT1, and H. vulgare CAT2 (Figure 5). After this event, the gain of an intron in position 4 occurred in the O. sativa CAT1 sequence. A second event of concerted intron loss occurred in Z. mays Cat2, and numerous other single-intron losses have occurred within the angiosperm catalases. In plants, examples of critical regulatory elements residing within intron sequences have accumulated in recent years (Calliset al. 1987; Fu et al. 1995a,b; Kaoet al. 1996; Sieburth and Meyerowitz 1997). Intron loss or gain, as well as the modification of sequences residing within introns, may therefore provide critical steps in the divergence of expression patterns of individual gene family members.
We acknowledge the Arabidopsis Biological Resource Center (Ohio State University, Columbus, OH); Rod Wing and Bob Creelman for the BAC filters and clones; Tom McKnight for helpful discussions of the data; and Harry D. Kurtz, Jr., Mary Lou Guerinot, Rich Meagher, an anonymous reviewer, and the members of our laboratories for critical reading of the manuscript. This work was supported by a National Science Foundation predoctoral fellowship to J.A.F., by a grant from the National Science Foundation to M.A.M., by grants from the U.S. Department of Agriculture National Research Initiative Competitive Grants Program to T.L.T. and to C.R.M., and by an institutional grant from the American Cancer Society to the Norris Cotton Cancer Center at Dartmouth.
Communicating editor: J. Chory
- Received September 3, 1997.
- Accepted January 26, 1998.
- Copyright © 1998 by the Genetics Society of America