Of the Nine Cytidine Deaminase-Like Genes in Arabidopsis, Eight Are Pseudogenes and Only One Is Required to Maintain Pyrimidine Homeostasis in Vivo

Bioinformatic, biochemical, and reverse genetics analyses suggest that only one of the nine annotated Arabidopsis cytidine deaminase genes encodes a fully functional enzyme involved in pyrimidine catabolism.

Abstract

CYTIDINE DEAMINASE (CDA) catalyzes the deamination of cytidine to uridine and ammonia in the catabolic route of C nucleotides. The Arabidopsis (Arabidopsis thaliana) CDA gene family comprises nine members, one of which (AtCDA) was shown previously in vitro to encode an active CDA. A possible role in C-to-U RNA editing or in antiviral defense has been discussed for other members. A comprehensive bioinformatic analysis of plant CDA sequences, combined with biochemical functionality tests, strongly suggests that all Arabidopsis CDA family members except AtCDA are pseudogenes and that most plants only require a single CDA gene. Soybean (Glycine max) possesses three CDA genes, but only two encode functional enzymes and just one has very high catalytic efficiency. AtCDA and soybean CDAs are located in the cytosol. The functionality of AtCDA in vivo was demonstrated with loss-of-function mutants accumulating high amounts of cytidine but also CMP, cytosine, and some uridine in seeds. Cytidine hydrolysis in cda mutants is likely caused by NUCLEOSIDE HYDROLASE1 (NSH1) because cytosine accumulation is strongly reduced in a cda nsh1 double mutant. Altered responses of the cda mutants to fluorocytidine and fluorouridine indicate that a dual specific nucleoside kinase is involved in cytidine as well as uridine salvage. CDA mutants display a reduction in rosette size and have fewer leaves compared with the wild type, which is probably not caused by defective pyrimidine catabolism but by the accumulation of pyrimidine catabolism intermediates reaching toxic concentrations.

The catabolism of nucleotides is part of the plant metabolic network for nutrient remobilization (Zrenner et al., 2009; Werner et al., 2010), which is particularly important for nitrogen (Xu et al., 2012). Pyrimidine nucleotides contain two nitrogen atoms in the ring, and in the case of cytosine, a third amino group nitrogen is linked to the C4 carbon of the heterocycle. The hydrolysis of this amino group occurs at the nucleoside level by the deamination of cytidine to uridine mediated by CYTIDINE DEAMINASE (CDA; Fig. 1). By contrast, adenosine is deaminated as nucleoside monophosphate (AMP to IMP by AMP deaminase; Xu et al., 2005), whereas G, similar to C, is deaminated at the level of the nucleoside (guanosine to xanthosine by guanosine deaminase; Dahncke and Witte, 2013). Cytidine deamination is required to feed C bases into pyrimidine ring catabolism, which is initiated from uracil (Fig. 1).

Figure 1.

Model of cytosolic pyrimidine catabolism and salvage in Arabidopsis. kinase, Nucleoside kinase; PPase, phosphatase acting on 5′ monophosphate nucleotides (5′ nucleotidase). Dashed connections are only relevant in cda mutants.

There are two classes of CDAs: (1) a tetrameric type of about 15 kD per subunit represented by the enzyme from Bacillus subtilis (Johansson et al., 2002) and also found in mammals; and (2) a dimeric type of about 32 kD per subunit represented by the enzyme from Escherichia coli (Betts et al., 1994), which is also present in plants. The dimeric CDAs probably arose from the tetrameric class by gene duplication and subsequent fusion (Faivre-Nitschke et al., 1999). Although the structural fold of each CDA domain in the dimeric CDAs has been conserved, the sequences of these domains have diverged. Only the N-terminal domain has catalytic activity and is able to bind a zinc cofactor (Johansson et al., 2002).

Cytosine in RNA can be subject to deamination as well, a process called C-to-U editing. In plants, this process occurs mainly in chloroplasts and mitochondria involving pentatricopeptide proteins for target recognition. Which factor catalyzes the deamination is still not clear, but certain subtypes of pentatricopeptide proteins containing a C-terminal DYW domain bind zinc and may contribute directly to the deaminating activity (Hayes et al., 2013; Shikanai, 2015). Recently, C-to-U editing also was discovered in nuclear plant tRNA-Ser(AGA) and tRNA-Ser(GCT), but the biological function of this modification and the responsible deaminase are still unknown (Zhou et al., 2014).

A CDA of Arabidopsis (Arabidopsis thaliana; At2g19570) has been expressed in E. coli and purified for biochemical characterization (Faivre-Nitschke et al., 1999; Vincenzetti et al., 1999). The dimeric enzyme deaminates cytidine (K_m of 150–250 µm, V_max of 58–60 units mg⁻¹) and deoxycytidine (K_m of 75–120 µm, V_max of 38–49 units mg⁻¹), binds one zinc ion per subunit, and was suggested to locate to the cytosol because of the lack of an apparent subcellular targeting sequence. Faivre-Nitschke et al. (1999) postulated the presence of several CDA copies in Arabidopsis based on Southern-blot data. Sequencing of the Arabidopsis genome revealed the presence of a CDA gene family comprising nine members. These form a phylogenetic group distinct from other nucleoside or RNA-editing deaminases of this plant (Zhou et al., 2014). Eight members of the Arabidopsis CDA gene family are tightly clustered on chromosome 4, beginning at locus At4g29570 and ending at locus At4g29650 with one unrelated gene (At4g29590) interspersed.

Here, the CDA gene family of Arabidopsis and soybean (Glycine max) was analyzed, predicting and testing the functionality of different family members. The analysis was extended to predict the functionality of CDA orthologs from other sequenced plants. Using mutants, we queried whether AtCDA is required and sufficient to maintain pyrimidine homeostasis in vivo and describe the effects of CDA mutation on growth, the pyrimidine metabolite profile, and the resistance to toxic pyrimidine nucleoside analogs.

RESULTS

Bioinformatic Analyses

To determine consensus sequences of highly conserved amino acids in dimeric CDAs, a multiple alignment of genuine CDA sequences was generated. Using the amino acid sequence of one biochemically characterized CDA from Arabidopsis (locus At2g19570; Faivre-Nitschke et al., 1999; Vincenzetti et al., 1999) as a query in BLAST searches of the Phytozome database version 9.1 accessible on the Internet, CDA sequences from those plants were recovered that possess only a single CDA gene. Assuming that CDA is an indispensable component of plant primary metabolism, this strategy likely identified only truly functional CDAs. Additionally, the Concise Protein Database at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/genomes/prokhits.cgi), which contains data from fully sequenced genomes, was searched using the same query as above but selecting only bacterial sequences corresponding to CDA of the dimeric type. From the alignment (Supplemental Fig. S1), two consensus sequences were derived: (1) the overall consensus of absolutely conserved amino acids in dimeric CDAs (marked red in the alignment); and (2) the consensus of absolutely conserved amino acids in dimeric CDAs of vascular plants (marked with green triangles above the alignment).

Next, all nine members of the CDA family from Arabidopsis as well as members of the CDA families of the closely related species Arabidopsis lyrata and Capsella rubella were aligned, and the two consensus sequences derived from Supplemental Figure S1 were annotated in this alignment (Supplemental Fig. S2; deviations from the overall consensus are marked in yellow). Similarly, CDA sequences of all other plants that contain more than one CDA copy in their respective genomes were aligned (Supplemental Fig. S3; here, deviations from both consensus sequences are marked in yellow). Deviations from the respective consensus and insertions/deletions were counted in each alignment and are summarized in Supplemental Tables S1 and S2, respectively.

For the CDA gene family of Arabidopsis, these data revealed that only the CDA encoded at locus At2g19570 matches exactly to the general and the plant CDA consensuses. The other eight proteins (for locus identifiers, see Supplemental Table S1) deviate in at least one and up to 12 positions from the general consensus and in at least nine and up to 24 positions from the plant consensus. Additionally, these proteins have at least two insertions or deletions not found in other plant CDAs encoded by single-copy genes. All genes with sequence alterations were expressed weakly or not at all in comparison with the CDA encoding the protein that matches the consensus (Supplemental Table S3). Similar data were obtained for the CDA families from A. lyrata and C. rubella. This suggested that each of these three species possesses only one functional copy of CDA, whereas the other genes had accumulated mutations that likely resulted in the loss of CDA activity.

The eight CDA family members with deviations from the CDA consensuses were called CDA-LIKE1 (AtCDAL1) to AtCDAL8 (or Ath-L1 to Ath-L8 in the Supplemental Data), and the one CDA matching the consensuses was called AtCDA without numbering. This nomenclature was analogously applied for the CDA families from A. lyrata and C. rubella.

The sequence analysis was extended to the nine other plant species in Phytozome version 9.1, which contained more than one CDA gene (Supplemental Fig. S3; Supplemental Table S2). By the same criteria outlined above, probable nonfunctional CDAs were identified. Judged by sequence, in only five species out of 32 vascular plants analyzed were more than one fully functional CDA identified.

If AtCDAL1 to AtCDAL8 lost function, one would postulate that the number of mutations resulting in a different amino acid sequence (nonsilent) observed for the corresponding genes would be far higher than for AtCDA, which is subject to conserving selection. Using the sequence data of 812 Arabidopsis accessions, the quantity of independent mutations was compared for these genes and expressed as silent and nonsilent mutations on the protein level. Incidents of 1-bp deletions resulting in frameshift mutations in each gene were counted as well (Table I). Indeed, mutational rates were strongly increased for AtCDAL1 to AtCDAL8 in comparison with AtCDA, reinforcing the notion that these are pseudogenes.

Table I. Analysis of mutational frequency in AtCDA and AtCDAL genes of different Arabidopsis accessions.

For each CDA isoform, the number of independent single-nucleotide polymorphisms (SNPs) detected in 812 Arabidopsis ecotypes from the 1001 Genome Project resulting in either the same amino acid as in the reference (syn/100aa) or a different amino acid/stop codon (nonsyn/100aa) is shown normalized to the different protein lengths (SNPs per 100 amino acids). The number of deletions resulting in frameshifts is given in addition to the amino acid position(s) affected by the deletion. SNPs supported by the sequence of only one ecotype are not included.

Gene	Encoded Protein	syn/100aa	nonsyn/100aa	No. of Frameshifts
At2g19570	AtCDA	4.7	2.7	–
At4g29570	AtCDAL1	4.4	16.0	1 (Asp-175)
At4g29580	AtCDAL2	2.9	10.6	2 (Trp-227 and Glu-420)
At4g29600	AtCDAL3	7.2	17.6	–
At4g29610	AtCDAL4	6.8	5.5	–
At4g29620	AtCDAL5	4.2	17.8	2 (Glu-221 and Leu-317)
At4g29630	AtCDAL6	6.3	9.4	1 (Leu-68)
At4g29640	AtCDAL7	9.2	20.5	–
At4g29650	AtCDAL8	4.0	7.2	–

Enzymatic Activity

To validate the results from the bioinformatic analyses, AtCDA, AtCDAL3, and AtCDAL4 were cloned, the corresponding proteins were transiently expressed in Nicotiana benthamiana as C-terminally tagged StrepII variants and affinity purified, and the specific activity for cytidine and deoxycytidine was assessed. AtCDAL3 and AtCDAL4 were selected because they show comparatively few deviations from the general and plant consensuses (Supplemental Table S1). We failed to amplify AtCDAL3 and AtCDAL4 from leaf complementary DNA (cDNA) but were able to obtain clones from genomic DNA. Transient expression and purification of the corresponding proteins resulted in highly pure preparations for each enzyme (Fig. 2; Supplemental Fig. S4) used for the determination of kinetic constants. The catalytic efficiencies of AtCDA were 48.3 and 85.9 mm⁻¹ s⁻¹ for cytidine and deoxycytidine, respectively. For the same substrates, AtCDAL4 showed catalytic efficiencies of 0.02 and 1.7 mm⁻¹ s⁻¹, respectively (Table II; Supplemental Fig. S5), whereas for AtCDAL3, no activity could be detected. Surprisingly, AtCDAL4, with one deviation from the general consensus and 11 deviations from the plant CDA consensus (Supplemental Table S1), had not completely lost activity. Nonetheless, the kinetic parameters of AtCDAL4 (Table II) are very likely insufficient for a function as cytidine/deoxycytidine deaminase in vivo.

Figure 2.

Purification of C-terminally StrepII-tagged AtCDA from leaf extracts of N. benthamiana with Streptactin affinity chromatography. Samples (10 µL) were taken at different stages of the purification and visualized by SDS gel electrophoresis followed by colloidal Coomassie Blue staining (A), silver nitrate staining (only elution fraction; B), and western blotting with Streptactin-alkaline phosphatase (AP) conjugate detection (C). Lane 1, Extract of soluble proteins; lane 2, proteins not bound after incubation with Streptactin affinity matrix; lane 3, protein in the last wash fraction; lane 4, pool of eluted protein; lane 5, protein left on the matrix after elution, released by boiling in SDS buffer.

Table II. Kinetic parameters for substrates of recombinant CDAs.

k_cat, Turnover number.

Substrate	AtCDA			AtCDAL4			GmCDA1			GmCDA2
	Km	kcat	kcat/Km	Km	kcat	kcat/Km	Km	kcat	kcat/Km	Km	kcat	kcat/Km
	mm	s−1	mm−1 s−1	mm	s−1	mm−1 s−1	mm	s−1	mm−1 s−1	mm	s−1	mm−1 s−1
Cytidine	0.35 ± 0.07	17.01 ± 0.87	48.3	28.93 ± 1.76	0.50 ± 0.01	0.02	1.37 ± 0.17	139.01 ± 4.44	101.5	3.02 ± 0.29	118.97 ± 2.97	39.4
Deoxycytidine	0.11 ± 0.02	9.70 ± 0.34	85.9	3.74 ± 0.35	6.38 ± 0.17	1.7	0.24 ± 0.01	64.85 ± 0.94	269.1	0.33 ± 0.02	58.84 ± 0.94	178.3

The soybean genome contains three full-length CDA genes potentially encoding the proteins GmCDA1 to GmCDA3. GmCDA3 deviates in two positions from the general and in seven positions from the plant CDA consensus and has a longer sequence deletion affecting amino acids absolutely conserved in dimeric CDAs (Supplemental Fig. S3; Supplemental Table S2). Therefore, it is highly unlikely that GmCDA3 represents a functional CDA. GmCDA1 and GmCDA2 both deviate from the plant consensus in several positions. Most strikingly, GmCDA1 translated from the soybean genome sequence contains a Y at position 164 that is conserved as N in all other plant sequences, and GmCDA2 contains F at position 65 where normally L is conserved in plants. These nonconservative exchanges might compromise the activity of the respective proteins, which would reduce the capability of soybean to catalyze cytidine deamination. To test this hypothesis, the coding sequences of both proteins were cloned from soybean leaf cDNA. To our surprise, we found an N and not a Y codon in the cDNA coding for amino acid 164 of the GmCDA1 protein (codon UAC instead of AAC, as found in the genome sequence). Because N is absolutely conserved in plant CDAs, we concluded that the soybean genome sequence must be wrong at this position. Nonetheless, GmCDA1 and GmCDA2 were transiently expressed and purified (Supplemental Fig. S4) and the activity was assessed. GmCDA1 had catalytic efficiencies of 101.5 and 269.09 mm⁻¹ s⁻¹ for cytidine and deoxycytidine, respectively. The catalytic efficiencies of GmCDA2 were 39.4 and 178.3 mm⁻¹ s⁻¹ for the same substrates (Table II). It is possible that the nonconservative L-to-F exchange of amino acid 65 is mainly responsible for the comparatively low activity of GmCDA2. The equivalent residue lines the active site pocket in the enzymes of E. coli (Betts et al., 1994) and Vibrio cholera (Protein Data Bank accession no. 4EG2) and is conserved as L, I, V, or M in CDAs of plants and bacteria (Supplemental Fig. S1). For soybean, GmCDA1, lacking this nonconservative exchange, is the most active CDA isoform. The related species Phaseolus vulgaris and Medicago truncatula also possess a CDA isoform with this L-to-F exchange (PvCDA2 and MtCDA2; Supplemental Fig. S3), indicating that, in these plants, the other respective CDA isoforms, PvCDA1 and MtCDA1, which agree with the consensus at this position, are the most active enzymes.

Subcellular Localization

To clarify the subcellular location of AtCDA and the two soybean CDAs, fusion proteins comprising a C-terminal yellow fluorescent protein (YFP) tag (AtCDA-YFP, GmCDA1-YFP, and GmCDA2-YFP) were transiently expressed in leaves of N. benthamiana. As a control, cytosolic β-ureidopropionase fused to cyan fluorescent protein (CFP) from Arabidopsis (β-UP-CFP; PYD3-CFP) was coexpressed (Zrenner et al., 2009). Additionally, transgenic Arabidopsis plants expressing AtCDA-YFP were generated and leaf protoplasts were prepared. Confocal microscopy revealed that AtCDA is located in the cytosol after transient expression (Fig. 3, A–C) as well as in leaf protoplasts from the transgenic line (Fig. 3, D and E). The stability of the transgene was demonstrated by an immunoblot (Fig. 3F). A cytosolic location also was found for the two soybean CDAs (Supplemental Figs. S6 and S7).

Figure 3.

Subcellular localization of AtCDA. A to C, Confocal fluorescence microscopy images of cells at the lower leaf epidermis of N. benthamiana transiently coexpressing C-terminally AtCDA-YFP and β-UP-CFP fusion proteins. YFP (A), CFP (B), and YFP and CFP (C) detection are shown. Bars = 30 µm. D and E, Mesophyll cell protoplasts of cda-1 plants transformed with AtCDA-YFP. YFP (D) and YFP and chlorophyll (E) detection are shown. Bars = 10 µm. F, Stability test of the AtCDA-YFP fusion protein from transgenic Arabidopsis plants analyzed by an immunoblot developed with a GFP-specific antibody.

Deamination of Cytidine in Vivo

The bioinformatic analyses (Table I; Supplemental Fig. S2; Supplemental Table S1) and the biochemical assessment of two members of the Arabidopsis CDA family (Table II) suggested that AtCDA is the only cytidine-deaminating enzyme in this plant. Two independent plant lines with homozygous transfer DNA (T-DNA) insertions in the AtCDA gene were isolated from the mutant collection of the German Plant Genomics Research Program (cda-1 [GK645H07]; Kleinboelting et al., 2012) and the Salk Institute Genomic Analysis Laboratory (cda-2 [SALK_036597]; Alonso et al., 2003).

The insertion positions were mapped to 3′ of base 96 for cda-1 and to 3′ of base 150 for cda-2 when counted from the start codon (Fig. 4A). For both mutant lines, the absence of intact mRNA was confirmed by reverse transcription-PCR (Fig. 4B).

Figure 4.

Genetic characterization and seed pyrimidine metabolite profiles of two independent cda mutant lines. A, Genomic organization of the At2g19570 locus and positions of the T-DNA insertions (triangles) in cda-1 (GK645H07) and cda-2 (SALK_036597C). The box represents the coding sequence, which does not contain any introns. Approximate primer positions for N261, 448, N61, and N262 are indicated. B, PCR and reverse transcription-PCR analyses of homozygous cda-1 and cda-2 lines and the wild type Columbia-0 (Col-0). C, Analysis of seed metabolite extracts using HPLC with photometric detection of the wild type, the two cda mutants, and a cda-1 line complemented with an AtCDA-YFP transgene. mAU, Milliabsorption units. D, Quantification of metabolites accumulating in mutant seeds. Error bars indicate sd (n = 4). Different letters mark significant differences at P < 0.05. fw, Fresh weight; n.d., not detectable.

In wild-type seeds of Arabidopsis, only a small amount of cytidine was detected by HPLC. By contrast, in seeds of both mutant lines, an accumulation of cytidine (60 times higher than the concentration in the wild type), cytosine, CMP, and also some uridine was observed (Fig. 4, C and D). The metabolites were identified by retention time and UV spectra of the corresponding standards. The introduction of an AtCDA-YFP transgene driven by the 35S promoter into the cda-1 background suppressed the accumulation of these metabolites (Fig. 4C, bottom).

In leaves, cytidine and cytosine accumulated in the mutant lines in an age-dependent manner but to lower total concentrations than in seeds (CMP and uridine were not detected). Such an accumulation was not observed in the wild type or the complementation line (Fig. 5). The higher metabolite accumulation in seeds in comparison with leaves corresponded well to a significantly higher transcript amount of CDA in reproductive organs, especially siliques, compared with leaves and other tissues (Supplemental Fig. S8).

Figure 5.

Cytidine and cytosine contents in leaves during plant development. Quantification of cytidine (top) and cytosine (bottom) is shown in rosette leaves of Col-0, cda-1, cda-2, and the complementation line from 15 to 55 d after germination (dag). Error bars indicate sd (n = 4). Different letters mark significant differences at P < 0.05. fw, Fresh weight; n.d., not detectable.

These data demonstrate that AtCDA is the central cytidine-deaminating enzyme in vivo, despite the presence of another eight CDAL genes in Arabidopsis. The accumulation of deoxycytidine was not observed. Either DNA degradation was very limited or an alternative deaminase, such as AtCDAL4, might have compensated for the loss of AtCDA with respect to deoxycytidine deamination in the mutant lines. We favor the first explanation because (1) AtCDAL4 has a very low catalytic efficiency (Table II) and (2) an accumulation of deoxyguanosine was not observed in guanosine deaminase knockout lines despite strong accumulation of guanosine (Dahncke and Witte, 2013), indicating that no significant DNA degradation occurs.

The Origin of Cytosine in the cda Mutant

The cda-1 mutant was crossed to the mutant of NUCLEOSIDE HYDROLASE1 (NSH1; mutant allele nsh1-1 [SALK_083120]; Jung et al., 2011). NSH1 catalyzes the hydrolysis of uridine (Fig. 1) or xanthosine to ribose and the respective nucleobases uracil or xanthine. We hypothesized that NSH1 also might degrade cytidine to cytosine and ribose in the cda background, explaining the accumulation of cytosine in this mutant. It has been reported that cytidine is not a substrate for NSH1 (Jung et al., 2009). However, at the high cytidine concentration in the cda mutant and during the long exposure times of the enzyme to the potential substrate in vivo, even catalysis at a very low rate might be enough to partially destabilize cytidine. In the cda-1 nsh1-1 double mutant, cytosine accumulation was reduced by 78% compared with the concentration observed in the cda-1 single mutant (Fig. 6B), consistent with the idea that NSH1 can hydrolyze cytidine in vivo if this metabolite accumulates. An alternative explanation is that the high uridine concentration in the nsh1 background prevents an unknown nucleoside hydrolase from using cytidine as a substrate. Interestingly, the decrease of cytosine did not lead to a corresponding increase in cytidine concentration in the double mutant, possibly because the cytidine pool also can be tapped by a kinase generating CMP (Fig. 1).

Figure 6.

Pyrimidine metabolite analysis of seeds of the cda-1 nsh1-1 double mutant in comparison with the respective single mutants and the wild type. A, HPLC spectrophotometric traces of nsh1-1 and cda1 nsh1-1 seed extracts. mAU, Milliabsorption units. B, Quantification of pyrimidine metabolites in seed extracts of Col-0, cda-1, nsh1-1, and the double mutant. Error bars indicate sd (n = 4). Different letters mark significant differences at P < 0.05. n.d., Not detectable.

The slight increase of uridine in the cda mutant compared with the wild type (Fig. 4, C and D), which at first seems counterintuitive because uridine is the product of the CDA reaction (Fig. 1), might be explained by an occupation of NSH1 with the abundant cytidine, thereby partially preventing NSH1-catalyzed uridine hydrolysis.

The nsh1 mutant accumulated uridine and also some UMP in the seed (Fig. 6, A and B; Dahncke and Witte, 2013), as was shown previously for nonseed tissues (Jung et al., 2009; Riegler et al., 2011). The uridine concentration is about half in the double mutant in comparison with the nsh1 mutant (Fig. 6B), indicating that cytidine deamination contributes 50% to the uridine pool in seeds of the nsh1 mutant. The UMP accumulation observed in the nsh1 mutant background is decreased and the CMP accumulation observed in the cda mutant background is slightly increased in the double mutant (Fig. 6).

The Susceptibility to Toxic Nucleoside Analogs Is Altered in cda Plants

Nucleoside analogs are frequently used in cancer therapy, because they are cytotoxic after incorporation into the cellular nucleotide pool that is used for DNA replication (Jordheim et al., 2013). Cytidine deaminase is a well-recognized factor in altering the effects of nucleoside analogs on human cells (Serdjebi et al., 2015).

To investigate the impact of CDA in plants on the toxicity of the nucleoside analogs 5-fluorocytidine (5-FC) and 5-fluorouridine (5-FD), cda-1, cda-2, Col-0, and the complementation line (cda-1 + AtCDA-YFP) were grown on one-half-strength Murashige and Skoog medium supplemented with these compounds. The growth of cda-1 and cda-2 plants was blocked shortly after germination by 5 µm 5-FC, whereas Col-0 and the complementation line grew only slightly slower than the control (Fig. 7). The situation was reverse on medium supplemented with 50 µm 5-FD. The mutant lines grew better than Col-0 and the complementation line (Fig. 7). Arabidopsis shows a lower tolerance to 5-FC than to 5-FD. The mutation of CDA decreases the 5-FC tolerance even further but increases the 5-FD tolerance. It appears that 5-FC is easily salvaged by a nucleoside kinase and that the extent of salvage is increased in the cda mutant, because the catabolic route is blocked in this genetic background. The increased CMP pool in the cda mutant (Fig. 6B) also suggests that cytidine salvage is boosted. 5-FD is salvaged as well, and the stronger tolerance of the cda mutant to this compound might be explained by a high occupancy of the nucleoside kinase with the excess cytidine in this mutant, which leaves less kinase capacity for 5-FD salvage. Interestingly, these data indicate that pyrimidine nucleosides are probably salvaged by kinases with dual specificity for uridine and cytidine.

Figure 7.

Growth responses to 5-FD and 5-FC. A, Scheme of seed placement on agar plates with four partitions. B, Col-0, cda-1, cda-2, and the complementation line after 10 d of growth on standard medium under long-day conditions (16 h of light). C, Like B but in the presence of 5 µm 5-FC, recorded on day 15. D, Like B but in the presence of 50 µm 5-FD, recorded on day 27.

The Mutation of CDA Reduces Plant Performance

Both cda mutant lines displayed decreased growth and produced fewer leaves compared with the wild type or the complementation line (Fig. 8). The effect becomes apparent at about 26 d after germination under long-day conditions (16 h of light) when quantified by measuring the rosette diameter. Partially compromised pyrimidine catabolism or the interference of the accumulating intermediates with metabolism might be the reasons for the reduced plant performance. Because such a drastic growth depression is not observed in other pyrimidine catabolism mutants, such as the nsh1 mutant (Jung et al., 2011; Riegler et al., 2011) or the pyd mutants (Zrenner et al., 2009), the interference of accumulating pyrimidine compounds with metabolism appears to be the most plausible explanation for the growth depression observed in the cda mutants. However, in reproductive organs and during embryo development, where the highest accumulation of pyrimidine catabolism intermediates (Fig. 5) and the highest expression of CDA (Supplemental Fig. S8) were detected, no phenotypic differences between the wild type, the mutants, and the complementation line were observed (Supplemental Fig. S9).

Figure 8.

Growth phenotypes of cda lines. A, Rosette diameters of Col-0, cda-1, cda-2, and the complementation line (n = 10 for each) grown 45 d under long-day conditions (16 h of light). B, Rosettes and rosette leaves of the different genotypes at 45 d after germination (dag). For documentation, the inflorescences were removed. Bars = 5 cm.

DISCUSSION

Although Arabidopsis possesses a family of nine genes potentially encoding cytidine deaminases, only one of these genes gives rise to a fully functional enzyme: AtCDA. What is the function of the other eight CDAL genes? It has been speculated that members of the CDA family might be involved in a putative cellular machinery that introduces hypermutations into viral genomes as a means of antiviral defense (Lin et al., 2009). Alternatively, CDAL proteins might have a function in C-to-U editing in the mitochondria and plastids or in the recently discovered C-to-U editing of two nuclear plant tRNAs (Zhou et al., 2014). However, there are several arguments against a functional role of CDAL proteins in these processes. (1) The expansion of the CDA family is limited to Arabidopsis and a few other closely related Brassicaceae members. Depending on the Brassica spp., the number of CDA family members varies, arguing that there are at least some nonessential genes (Supplemental Fig. S2). In line with this, there are some members that are not expressed in any tissue of Arabidopsis according to publicly available expression data (Supplemental Fig. S3), and in some Arabidopsis accessions, certain CDAL family members have suffered frameshift mutations (Table I). (2) Most plants have only one CDA gene and lack CDAL genes. Therefore, it is unlikely that C-to-U editing, which is found everywhere in the plant kingdom, is performed by CDAL proteins only in the Brassicaceae. (3) Most editing occurs in mitochondria and plastids, but AtCDA is cytosolic, and there are no predicted targeting peptides in the CDAL proteins. It also has been shown that AtCDA does not bind RNA (Faivre-Nitschke et al., 1999). (4) Only AtCDAL4 retains weak deoxycytidine-deaminating activity. AtCDAL3 is not active, and the others are so highly mutated that they will not have maintained activity, especially because in every protein except AtCDAL3 and AtCDAL4, absolutely conserved residues are affected, which were shown to be directly involved in zinc or substrate binding in the CDA from E. coli (Supplemental Fig. S2; Betts et al., 1994). The inability of the CDAL proteins to compensate for the loss of CDA in the cda mutant is in line with these observations. (5) The high number of SNPs in CDAL genes of different Arabidopsis ecotypes resulting in amino acid changes of the corresponding proteins indicates that they are subject to low selective pressure, which is typical for nonfunctional genes. The CDAL genes are located in a cluster on chromosome 4. The equivalent region in Brassica rapa still contains a CDA gene, which is very likely functional (BrCDA2; Supplemental Table S2), in addition to the BrCDA1 gene. Such a situation also is found in Aethionema arabicum, belonging to the most basal group of the Brassicaceae (Kagale et al., 2014). This indicates that a duplication of CDA early in Brassicaceae evolution created two genomic CDA loci. In some crucifers, both loci were functionally maintained, maybe stabilized by differential expression patterns, and in others, one locus lost functionality, creating CDAL genes. In summary, it can be concluded that AtCDAL1 to AtCDAL8 are most likely pseudogenes, at least with respect to a C-deaminating function.

In soybean, many genes are present in multiple copies that, in other plants, are only required in a single copy (Polacco et al., 2011; Werner et al., 2013). Soybean is a paleopolyploid plant whose genome underwent polyploidization at least twice, the last time relatively recently (Kim et al., 2009). This explains the increased number of gene copies that also affected CDA. However, GmCDA1 is the most active isoform, whereas GmCDA2 already has suffered critical mutations reducing its catalytic efficiency (Table II). It is possible that GmCDA2 fulfills a tissue-specific role, or will be lost over time, or will acquire a new function.

Cytidine accumulation in seeds of the cda mutant has secondary effects increasing CMP, cytosine, and uridine pool sizes. Cytosine is likely generated by cytidine hydrolysis via NSH1, which in the wild type would not occur because cytidine is of low abundance and not a good substrate for NSH1 (Jung et al., 2009). The CMP pool size is probably influenced by the respective rates of (1) cytidine phosphorylation, likely increased by high substrate availability, and (2) CMP dephosphorylation, which might be inhibited by high cytidine concentrations. Even more CMP accumulates in seeds of the cda nsh1 double mutant, which in addition to cytidine also accumulates uridine (Fig. 6B). A possible explanation is a regulatory role of uridine on the kinase/phosphatase module (Fig. 1), leading to the promotion of mononucleotide formation. A central regulatory role of the uridine-degrading enzyme NSH1 has been discussed (Jung et al., 2009), which might be mediated by uridine abundance. By contrast, UMP concentrations are significantly lower in the double mutant than in the nsh1 single mutant. This finding might be explained by decreased substrate availability to the kinase and/or reduced inhibition of the phosphatase, because the uridine pool size also is reduced by 50% in the double mutant compared with nsh1. Additionally, if uridine has a regulatory function stimulating mononucleotide formation, as hypothesized above, a reduction in uridine concentration would lead to decreased mononucleotide levels. In this model, uridine rather has a positive effect on the activity of the nucleoside kinase(s) than a negative effect on the nucleotide phosphatase(s), because there is no CMP accumulation in the nsh1 mutant. A further explanation for the reduced UMP concentrations in the double mutant could be that the nucleoside kinase is more occupied with cytidine phosphorylation and not fully available to uridine in this mutant. That cytidine can successfully compete for the nucleoside kinase is indicated by the weak phenotype of the cda mutants in comparison with the wild type in the presence of fluorouridine (Fig. 7). Together, these data suggest that the cytosolic nucleoside kinase(s) have dual specificity for uridine and cytidine and may be stimulated by uridine. Two plastidic uridine kinases (UKL1 and UKL2) from Arabidopsis have already been characterized (Chen and Thelen, 2011), but whether cytidine is a substrate has not been assessed. An additional three genes with high similarity to uridine kinases (UKL3, UKL4, and UKL5) are found in the Arabidopsis genome (Mainguet et al., 2009), but they have not been functionally characterized yet. However, a biochemically characterized uridine kinase of maize (Zea mays) was shown to have dual specificity for uridine and cytidine (Deng and Ives, 1975).

It has been claimed that CDA is involved in the age-related resistance response of Arabidopsis (Carviel et al., 2009), because plants compromised in CDA showed reduced resistance to the virulent bacterium Pseudomonas syringae pv tomato and to the downy mildew fungus Hyaloperonospora parasitica. Additionally, CDA was transcriptionally induced in leaves by Pseudomonas spp. infection. However, it is shown here that cda mutants are compromised in their growth, likely due to the accumulation of metabolites, either cytidine directly or its derivatives, at toxic concentrations. This metabolic stress might well explain why the cda mutants are more susceptible to pathogens, and a specific role of CDA in age-related resistance does not need to be postulated. The transcriptional up-regulation of CDA in this context might be caused by senescence induction due to pathogen infection. Public expression data as well as our transcriptional data (Supplemental Fig. S8) show that CDA transcript amounts increase in senescence.

MATERIALS AND METHODS

Plant Material and Cultivation

Arabidopsis (Arabidopsis thaliana) ecotype Col-0 was chosen as the wild type. T-DNA insertion mutants of Arabidopsis from the Gabi-Kat collection (GK645H07, cda-1; Kleinboelting et al., 2012) and the SALK collection (SALK_036597, cda-2; and SALK_083120, nsh1-1; Alonso et al., 2003) were ordered from the European Arabidopsis Stock Centre. Arabidopsis and Nicotiana benthamiana plants were cultivated under the same condition described before (Witte et al., 2004, 2005). Transient expression in N. benthamiana was performed as described by Werner et al. (2008).

To determine alterations in rosette diameter and leaf number, plants of the different genotypes were grown randomly distributed on growth trays and analyzed for morphological characteristics and rosette diameter. For each genotype, 10 replicates were grown. Embryo preparation and microscopy were performed as described by Hauck et al. (2014).

Agar plates were prepared with one-half-strength Murashige and Skoog nutrients supplemented with the toxic nucleoside analog 5-FC or 5-FD where indicated. Plates were incubated in a controlled growth chamber under long-day conditions (16 h of light at 150 µmol m⁻² s⁻¹, 20°C day and 18°C night, 16% relative humidity).

Cloning, Reverse Transcription, and Real-Time PCR

RNA from plants was prepared using TRI reagent (Sigma) and treated with DNase I (Sigma) following the manufacturer’s instructions. Reverse transcription using 1 µg of total RNA was performed with Moloney murine leukemia virus reverse transcriptase (Invitrogen) and a poly(T) primer. PCR employed the following primers: for AtCDA, N354 + N355; for AtCDAL4, N466 + N467; for AtCDAL3, P1 + P2; for GmCDA1, N363 + N364; and for GmCDA2, N365 + N366. Real-Time PCR employed the following primers: for ACTIN2, 2531 + 2532; for glyceraldehyde 3-phosphate dehydrogenase (GAPDH) 3′, 2527 + 2528; for GAPDH 5′, 2529 + 2530; and for AtCDA, N515 + N516.

Protein Purification and Enzymatic Assays

StrepII-tagged AtCDA was affinity purified after transient expression in N. benthamiana as described by Werner et al. (2008). Purified protein was quantified using the Bradford reagent from Serva using bovine serum albumin as a standard. For the standard enzymatic assays, recombinant CDA solutions were adjusted to the following concentrations using elution buffer of the purification procedure: AtCDA at 5 µg mL⁻¹, GmCDA1 and GmCDA2 at 1 µg mL⁻¹, and AtCDAL4 at 20 µg mL⁻¹. For individual reactions, 75 µL of the substrate solution was incubated at 30°C for 5 min. The reaction was started by adding 25 µL of enzyme solution. In a time course, 20-µL aliquots were withdrawn and added to 80 µL of water, followed by the addition of 25 µL of phenol nitroprusside reagent and 50 µL of hypochlorite reagent for colorimetric ammonia quantification (Witte and Medina-Escobar, 2001). The absorbance was determined by a photometric measurement at 636 nm. Ammonium standard curves were generated by adding elution buffer instead of enzyme solution to the reaction mix and placing 20-µL aliquots in 80-µL NH₄Cl solutions of different concentrations prior to detection.

The kinetic constants were determined at 0.1, 0.25, 0.5, 1, 3, and 5 mm cytidine or 0.05, 0.1, 0.25, 0.5, 1, and 2 mm deoxycytidine for AtCDA; 5, 10, 25, 50, 100, 200, and 500 mm cytidine or 1, 2.5, 5, 10, 20, and 50 mm deoxycytidine for AtCDAL4; 0.5, 1, 2.5, 5, 10, and 20 mm cytidine or 0.1, 0.25, 0.5, 1, 2.5, and 5 mm deoxycytidine for GmCDA1; and 1, 3, 5, 10, 20, and 40 mm cytidine or 0.1, 0.25, 0.5, 1, 2.5, and 5 mm deoxycytidine for GmCDA2. Kinetic curves were recorded in three to four independent repeats, and kinetic constants were determined by fitting the data to the Michaelis-Menten equation using Graph Pad Prism software.

Mutant Characterization

Homozygous mutant lines were screened out of a segregating population by PCR using primers 448 + N262 and N261 + N262 for cda-1, primers N61 + N261 and N261 + N262 for cda-2, and primers 488 + 2616 and 1905 + 2616 for nsh1-1 (Supplemental Table S4). Double mutants were obtained by crossing nsh1-1 (male) and cda-1 (female). The PCR products from the mutants were cloned and sequenced to map the exact positions of the insertions.

To determine the amount of gene-specific mRNA in the mutants, cDNA from seedlings was prepared as described above. The PCR employed primers N261 + N262, giving rise to a product of 578 bp from the wild-type allele. In both mutant lines, the T-DNA insertions are flanked by these primers. For the amplification of ACTIN2 as a control, primers 2531 + 2532 were used. The expression level of NSH1 in the nsh1-1 mutant was determined before by Jung et al. (2011).

Metabolite Analyses

Seedlings were ground in liquid nitrogen using a mortar, 150 mg was passed frozen into a 1.5-mL micocentrifuge tube, and 360 µL of cold 0.5 m HClO₄ was added followed by grinding with a rotating pestle. Samples were incubated on ice for 10 min and centrifuged (15 min, 20,000g, and 4°C), and the supernatant was mixed with 20 µL of alkaline potassium carbonate solution (5 m KOH and 2 m K₂CO₃) to precipitate the perchlorate. After incubation for 5 min on ice, samples were centrifuged (15 min, 20,000g, and 4°C), and supernatants were frozen in liquid nitrogen, thawed, centrifuged again as before, and the new supernatants were transferred to HPLC sample vials. Seed extraction was performed by grinding 10 mg of material in 150 µL of HClO₄ with a rotating pestle. The supernatant was removed after centrifuging (15 min, 20,000g, and 4°C), and the extraction was repeated. Both supernatants were pooled and incubated on ice for 10 min and then mixed with 15 µL of alkaline potassium carbonate solution. All subsequent steps were performed as described above. Cytosine, cytidine, CMP, uridine, and UMP (50, 100, 250, 500, 1,000, and 2,000 µm solutions) were treated as the samples and used as standards. The HPLC analysis procedure followed was according to the description by Dahncke and Witte (2013).

Protoplasting and Subcellular Localization

For protoplasting, two 12-mm leaf discs were cut from N. benthamiana plants after 3 d of transient protein expression. The discs were washed using sterile water and wounded slightly at the lower surface with a sharp blade. After incubation with 500 mm mannitol for 1 h, the discs were transferred into 1 mL of enzyme solution (500 mm mannitol, 10 mm CaCl₂, 5 mm MES/KOH, pH 5.5, 3% Cellulase Onozuka, and 0.75% Macerozyme [Yakult Pharmaceutical]) and vacuum infiltrated at less than 25 mbar for 5 min. Protoplasts were isolated by slow shaking in darkness for 3 h.

For subcellular localization, AtCDA-YFP, GmCDA1-YFP, GmCDA2-YFP, and β-UP-CFP were coexpressed and analyzed by confocal microscopy as described by Dahncke and Witte (2013).

Mutational Rate and Gene Expression Analyses

Data for SNPs from 812 ecotypes from all available collections were obtained from http://1001genomes.org. Whenever possible, the data set with the more stringent SNP-calling parameters was chosen, and data files were reformatted using in-house scripting for annotation by SNPeff (Cingolani et al., 2012) using The Arabidopsis Information Resource 10 Arabidopsis gene model as a reference. Results were filtered for CDA isoforms, separated into synonymous and nonsynonymous SNPs, and counted using in-house scripting.

Short read data sets were downloaded from the National Center for Biotechnology Information short read archive available at http://www.ncbi.nlm.nih.gov/sra and mapped against The Arabidopsis Information Resource 10 Arabidopsis gene model using the CLC genome bench (Qiagen). Only reads mapping uniquely to one location were considered and normalized against the total number of mapped reads and the length of the respective CDA transcript.

Statistical Analyses

ANOVA followed by Tukey’s honestly significant difference test were performed for statistical evaluation. Significance levels of P < 0.05, P < 0.01, and P < 0.001 are indicated in the figures by single, double, and triple asterisks, respectively. Different letters represent significant differences at the P < 0.05 significance level.

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. Multiple alignment of CDA sequences from plants, algae, and bacteria.

• Supplemental Figure S2. Multiple alignment of sequences of the CDA families from Arabidopsis and its close relatives A. lyrata and C. rubella from the Camelineae tribe.

• Supplemental Figure S3. Multiple alignment of CDA from Arabidopsis with CDA protein sequences from several plants possessing more than one CDA gene copy.

• Supplemental Figure S4. Affinity-purified AtCDAL3, AtCDAL4, GmCDA1, and GmCDA2.

• Supplemental Figure S5. Determination of kinetic constants for AtCDA, AtCDAL4, GmCDA1, and GmCDA2.

• Supplemental Figure S6. Subcellular localization of GmCDA1.

• Supplemental Figure S7. Subcellular localization of GmCDA2.

• Supplemental Figure S8. Tissue-specific expression level of AtCDA.

• Supplemental Figure S9. Comparison of the phenotypes of reproductive organs and embryos of the wild type, cda-1, cda-2, and the complementation line.

• Supplemental Table S1. Quantitative evaluation of the deviation from the consensus sequences derived from the protein alignment in Supplemental Figure S2.

• Supplemental Table S2. Quantitative evaluation of the deviation from the consensus sequences derived from the protein alignment in Supplemental Figure S3.

• Supplemental Table S3. Expression analysis of CDA isoforms based on RNA sequencing data analysis.

• Supplemental Table S4. Primers used in this study.

Glossary

cDNA

complementary DNA

T-DNA

transfer DNA

5-FC

5-fluorocytidine

5-FD

5-fluorouridine

Col-0

Columbia-0

SNP

single-nucleotide polymorphism