Differentiation of 1-aminocyclopropane-1-carboxylate (acc) Deaminase from Its Homologs Is the Key for Identifying Bacteria Containing Acc Deaminase

One sentence summary: We developed an efficient molecular method for identifying bacteria containing ACC deaminase and clarified previous misunderstandings regarding identification of bacteria containing ACC deaminase and horizontal transfer of acdS genes. ABSTRACT 1-Aminocyclopropane-1-carboxylate (ACC) deaminase-mediated reduction of ethylene generation in plants under abiotic stresses is a key mechanism by which bacteria can promote plant growth. Misidentification of ACC deaminase and the ACC deaminase structure gene (acdS) can lead to overestimation of the number of bacteria containing ACC deaminase and their function in ecosystems. Previous non-specific amplification of acdS homologs has led to an overestimation of the horizontal transfer of acdS genes. Here, we designed consensus-degenerate hybrid oligonucleotide primers (acdSf3, acdSr3 and acdSr4) based on differentiating the key residues in ACC deaminases from those of homologs for specific amplification of partial acdS genes. PCR amplification, sequencing and phylogenetic analysis identified acdS genes from a wide range of proteobacteria and actinobacteria. PCR amplification and a genomic search did not find the acdS gene in bacteria belonging to Pseudomonas stutzeri or in the genera Enterobacter, Klebsiella or Bacillus. We showed that differentiating the acdS gene and ACC deaminase from their homologs was crucial for the molecular identification of bacteria containing ACC deaminase and for understanding the evolution of the acdS gene. We provide an effective method for screening and identifying bacteria containing ACC deaminase.


INTRODUCTION
1-Aminocyclopropane-1-carboxylate (ACC) is the immediate biosynthetic precursor to the plant hormone ethylene that regulates plant development and responses to environmental stress (reviewed by Bleecker and Kende 2000).ACC deaminase (ACCD) catalyzes the cleavage of ACC to α-ketobutyrate and ammonia (Honma and Shimomura 1978) and belongs to the tryptophan synthase β-subunit family (TRPSβ) of pyridoxal 5 -phosphate (PLP)-dependent enzymes (Todorovic and Glick 2008).ACCD and other TRPSβ members, such as D-cysteine desulfhydrase (DCyD), share the initial catalysis step, the formation of external aldimine between the substrate and PLP cofactor (Todorovic and Glick 2008).Unlike other TRPSβ members, ACCD catalyzes an unusual cleavage of the Cα-Cβ bond of an amino acid with a cyclopropane ring (Thibodeaux and Liu 2011).The Lys51, Ser78 and Tyr294 residues (position numbers in the ACCD of the model strain Pseudomonas sp.UW4), essential for ACCD activity, are also conserved in ACCD homologs, whereas the Glu295 and Leu322 residues are unique to ACCD (Fujino et al. 2004;Todorovic and Glick 2008).Notably, the double E295S/L322T mutation converted the ACCD of the strain UW4 into a DCyD; the double S358E/T386L mutation converted the tomato DCyD into an ACCD (Todorovic and Glick 2008).
ACCD exists widely in bacteria, fungi and stramenopiles (Bruto et al. 2014;Nascimento et al. 2014).ACCD is localized in the bacterial cytoplasm.ACCD-producing bacteria associated with plants can act as a sink for plant ACC by taking up and cleaving ACC released from the plant cells and thus can reduce ethylene generation in plants; in this way they can reduce the extent of growth inhibition caused by high-level ethylene, particularly under abiotic stresses (Glick, Penrose and Li 1998).ACCD-mediated plant growth promotion has stimulated tremendous interest in research into the isolation and application of ACCD-producing bacteria (reviewed by Glick et al. 2007Glick et al. , 2014)).
ACCD-producing bacteria are usually isolated on minimal media containing ACC as the sole nitrogen source and identified via detection of ACCD activity in bacteria grown on the ACC media (Glick, Karaturovíc and Newell 1995;Penrose and Glick 2003).However, growth on an ACC medium and low detectable ACCD activity are not confirmative of bacteria containing ACCD.First, nitrogen-fixing bacteria without ACCD are able to grow on the ACC medium (Li et al. 2011).Second, trace nitrogen in medium components and in agar may support the growth of some bacteria without ACCD.Third, a PLP-dependent deaminase that is not an ACCD may show low detectable non-specific ACCD activity (McDonnell et al. 2009;Nascimento et al. 2014).Fourth, some bacteria containing ACCD may not show ACCD activity in freeliving states (Ma et al. 2003;Blaha et al. 2006;Nascimento et al. 2012).Therefore, unambiguous detection of the ACC deaminase structure gene (acdS) is important for predicting ACCD activity and for identifying ACCD-producing bacteria.
Amplification of the acdS gene by PCR with degenerate primers has been widely used for molecular identification of ACCD-producing bacteria (Shah et al. 1998;Ma et al. 2003;Hontzeas et al. 2005;Blaha et al. 2006;Caballero-Mellado et al. 2007;Govindasamy et al. 2008;Onofre-Lemus et al. 2009;Nikolic, Schwab and Sessitsch 2011).However, these degenerate primers are either non-specific or only effective for amplification of acdS genes from a narrow range of bacteria.
An extensive horizontal transfer of acdS genes among bacteria was proposed because the acdS phylogeny of some strains did not agree with what would be predicted from a vertical transmission of the acdS gene (Hontzeas et al. 2005;Blaha et al. 2006;Glick et al. 2007).In contrast, Nascimento et al. (2014) recently did a comprehensive comparison of acdS and 16S rRNA genes from bacteria belonging to various classes and suggested that the acdS gene evolved mainly through vertical transmission, with occasional horizontal gene transfer.
Misidentification of ACCD and the acdS gene causes misunderstandings of microbial function and evolution, microbial community composition and microbial adaptation to ecosystems.The importance of differentiating ACCD and the acdS gene from their homologs has not been fully recognized in the identification of ACCD-producing bacteria and in the determination of the evolution of the acdS gene.The objective of this study was to establish an effective method for screening and identifying ACCD-producing bacteria.We designed primers based on the differentiation between ACCDs and homologs for specific amplification of partial acdS genes from a wide range of bacteria.We also aimed to clarify previous misunderstandings regarding the identification of ACCD-producing bacteria and the horizontal transfer of acdS genes.

Bacteria
A total of 374 bacterial isolates that were able to grow on minimal media containing ACC as the sole nitrogen source were tested for PCR amplification of the acdS gene and the assay of ACC consumption.The bacteria consisted of 44 known ACCDproducing isolates (18 Burkholderia, 10 Herbaspirillum and 16 Pseudomonas) (Li et al. 2011), the nitrogen-fixing Pseudomonas stutzeri strain A1501 (Yan et al. 2008), Bradyrhizobium sp.strain DGR24 isolated from a nodule of soybean, Bradyrhizobium sp.strain LR51 isolated from a nodule of mung bean and 327 other isolates obtained by culturing material from sugarcane roots and rhizosphere soils, soybean nodules, mung bean nodules and black locust nodules on a nitrogen-deficient DF medium (Li et al. 2011) and a modified Ashby medium (Lin et al. 2012)

Primer design
Consensus-degenerate hybrid oligonucleotide primers (CODE-HOP) (Rose et al. 1998;Rose, Henikoff and Henikoff 2003) were designed for specific amplification of partial acdS genes.Each CODEHOP consists of a short 3 degenerate core region (11-12 nucleotides) based on 3-4 highly conserved amino acid residues and a longer 5 consensus clamp region based on the most probable nucleotide predicted at each position to stabilize the 3 core during annealing to the template.Correctly amplified products are produced initially by precise matching of the primer to the template at the 3 core region and later by precise matching of the primer to the product at the 5 clamp region (Rose et al. 1998).

Gene amplification and sequencing
The acdS and 16S rRNA genes were amplified from the 374 bacterial isolates.Bacterial suspensions (10 μl), prepared from colonies (Lin et al. 2012) or washed liquid cultures, were heated in a P7021TP-6 microwave oven (Galanz, Foshan, China) on full power for 2 min.Bacterial lysates (1 μl) were used as the templates for PCR.
Bacterial 16S rRNA gene sequences were amplified using the universal 27F/1492R primers (Lane 1991).A template-free mixture without bacterial lysate was used as a negative control.The PCR products were sequenced after cloning (Lin et al. 2012) or sequenced directly using the 27F/1492R primers.
Partial acdS genes were amplified using the CODEHOP acdSf3/acdSr3 or acdSf3/acdSr4 primer pairs.Each primer was used at 0.4 μM.A ready-to-use 2× concentrated Taq PCR Master-Mix containing 0.1 U μl −1 of Taq polymerase, 0.5 mM each dNTP, 3 mM MgCl 2 , 100 mM KCl, 20 mM Tris-HCl (pH 8.3) and other additives (Tiangen Biotech Co. Ltd, Beijing, China) was used for reactions in 25 μl or 50 μl of the mixtures.A template-free mixture without bacterial lysate and a mixture with lysate from Es. coli strain DH5α were used as negative controls.PCR was started with a 4-min denaturation step at 94 • C, followed by 35 cycles of denaturation at 94 • C for 45 s, annealing at 53 • C for 45 s and extension at 72 • C for 1 min, and completed with a final extension at 72 • C for 10 min.PCR was done with an S1000 thermal cycler (Bio-Rad Laboratories Inc. Hercules, CA, USA).PCR products were detected using agarose gel electrophoresis, purified using a TIANgel Midi Purification Kit (Tiangen Biotech), and se-quenced after cloning (Lin et al. 2012) or sequenced directly with the acdSr3 and acdSr4 primers.When non-specific amplification occurred with the acdSf3/acdSr4 primers, a higher annealing temperature (55-59 • C) and the touchdown strategy were used to improve amplification specificity.The touchdown annealing temperature was decreased by 0.5 • C every cycle from 65 • C to the final annealing temperature (55-59 • C), and 20 additional cycles were run at the final annealing temperature.

Genomic search of ACCD and the acdS gene
'1-Aminocyclopropane-1-carboxylate deaminase' was searched from the annotated genomes of bacteria belonging to P. stutzeri and the genera Enterobacter, Klebsiella and Bacillus deposited in the NCBI database as at 14 April 2015.A genomic BLAST search (www.ncbi.nlm.nih.gov/sutils/genomtable.cgi) of the acdS gene was done with these bacterial genomes using the acdS sequence of Pseudomonas sp.UW4.

Phylogenetic analysis
Phylogenetic trees were constructed using the MEGA 5.2 program (Tamura et al. 2011) and tested by bootstrap with 1000 replicates.Sequences were aligned with the MUSCLE program and analyzed using the neighbor-joining algorithm and Kimura-2parameter model integrated in the MEGA program.

Determination of ACC consumption
Consumption of ACC was tested for the 374 bacterial isolates grown in the minimal DF-ACC medium (Li et al. 2011) or the M9-ACC medium (Duan et al. 2009) using a modified colorimetric ninhydrin assay of ACC with chimney-top 96-well PCR plates (Li et al. 2011).Approximately 80 μl of the 10-fold-diluted supernatant from the bacterial culture and 160 μl of the ninhydrin reagent were used for the ninhydrin reaction.The chimney-top 96-well PCR plate was supported by the matching 96-hole holder (Fig. S1, Supporting Information) and heated on boiling water.

Measurement of ACC deaminase activity
ACCD activity of acdS-positive isolates was determined after induction for 8-48 h in the DF-ACC medium or the M9-ACC medium and by the colorimetric 2,4-dinitrophenylhydrazine assay of α-ketobutyrate, as described by Penrose and Glick (2003) for non-rhizobia and Duan et al. (2009) for rhizobia.
All genomes of P. stutzeri strains deposited in the NCBI database contained putative acdS genes.An alignment of the deduced amino acid sequences showed that the putative ACCDs

Phylogenetic differentiation between acdS genes and homologs
The acdS sequences of H. seropedicae SmR1 and P. psychrotolerans L19 were at the border of the acdS clade and were distant from those of other betaproteobacteria and gammaproteobacteria (Fig. 2).The acdS sequence of the unreliably classified strain ACP grouped with that of B. phytofirmans PsJN.The acdS sequence of the unreliably classified strain CAL2 grouped with that of P. fluorescens F113.The acdS sequence of Ba. cereus AcdSPB4 grouped with those of Se. rubidaea AcdSPB1, Klebsiella pneumonia AcdSPB2, Pseudomonas sp.UW4 and misclassified E. cloacae UW4 (Fig. 2).
Another phylogenetic analysis of the putative acdS genes from the genomes of P. stutzeri strains and the reference acdS genes and acdS homologs showed that these putative acdS genes were clearly different from the acdS genes (Fig. S3, Supporting Information).

CODEHOP for amplification of the partial acdS genes
The CODEHOP program generated primers based on the conserved blocks containing the K51, S78, Y294, E295 or L322 residues and their neighboring residues.The alignment and comparison of nucleotide sequences of the output CODEHOP and corresponding regions of acdS genes and homologs showed that one forward primer, acdSf3, and two reverse primers,  acdSr3 and acdSr4, were likely to be useful for differentiating acdS genes from homologs (Table 1 and Fig. 3).
The primer acdSf3 covers a region containing the codon of the essential S78 residue and an average GC content of ∼60%; its 3 degenerate core generally differentiated the test acdS and homolog sequences at the 3 end (Fig. 3).The primer acdSr3 covers a region containing the codons of the essential Y294 and E295 residues and an average GC content of ∼54%; its 3 degenerate core weakly differentiated the test acdS and homolog sequences (Fig. 3).The primer acdSr4 covers a region containing the codon of the essential L322 residue and an average GC content of ∼72%; its 3 degenerate core generally differentiated the test acdS and homolog sequences (Fig. 3), thus is likely to be specific to acdS genes but requires stringent annealing conditions.

CODEHOP-based amplification of the partial acdS genes
The CODEHOP pairs of acdSf3/acdSr3 or acdSf3/acdSr4 were first tested on the 44 known ACCD-producing isolates (18 Burkholderia, 10 Herbaspirillum and 16 Pseudomonas), P. stutzeri A1501 containing putative acdS genes, and the negative control Es. coli DH5α.Predicted amplification products (∼680 bp with acdSf3/acdSr3 or ∼760 bp with acdSf3/acdSr4) were obtained from the 44 known ACCD-producing isolates but not from P. stutzeri A1501 or Es. coli DH5α (Fig. S4, Supporting Information).Pseudomonas stutzeri A1501 and Es. coli DH5α were also negative for the assays of ACC consumption and ACCD activity.
The CODEHOP pairs of acdSf3/acdSr3 or acdSf3/acdSr4 were then used to screen acdS-positive bacteria from the two Bradyrhizobium strains and the other 327 isolates.Predicted amplification products were obtained from eight proteobacterial isolates belonging to five genera (Bradyrhizobium, Achromobacter, Ralstonia, Variovorax and Pseudomonas) and five actinobacterial isolates belonging to three genera (Marmoricola, Microbacterium and Mycobacterium) (Table S1, Supporting Information).These 13 isolates were positive for the assays of ACC consumption and ACCD activity.The other 316 isolates were negative for the amplification of the acdS gene and the assay of ACC consumption.Notably, the 57 Enterobacter isolates, 51 Klebsiella isolates and 52 Bacillus isolates were negative for the amplification of the acdS gene and the assay of ACC consumption.Moreover, no acdS was identified in the genomes of 321 Enterobacter strains, 580 Klebsiella strains or 700 Bacillus strains deposited in the NCBI database as at 14 April 2015.
A phylogenetic analysis of the sequences from the 57 amplification-positive isolates (Table S1, Supporting Information) and the reference acdS genes and acdS homologs (Fig. S3, Supporting Information) showed that the amplified sequences grouped with the acdS genes (data not shown), indicating that the amplified products were partial acdS genes.The amplified acdS sequences from the known ACCD-producing Herbaspirillum and Pseudomonas isolates (Table S1, Supporting Information) grouped with the acdS gene of H. seropedicae SmR1 and the acdS gene of P. psychrotolerans L19, respectively (data not shown), and formed a unique group that was distant from the acdS genes of other betaproteobacteria and gammaproteobacteria (Fig. 2).
PCR with the acdSf3/acdSr3 primers and an annealing temperature of 53 • C generally resulted in specific amplification of acdS fragments of ∼680 bp (Fig. S4, Supporting Information), whereas PCR with the acdSf3/acdSr4 primers and an annealing temperature of 53 • C resulted in the predicted 760 bp fragments and frequent non-specific fragments (data not shown).In line with the digital evaluation, PCR with the acdSf3/acdSr4 primers required a stringent annealing condition.Using higher annealing temperatures (variable for different isolates) and the touchdown strategy eliminated the non-specific products close to the predicted amplification products (Fig. S4, Supporting Information).

CODEHOP-based PCR amplification of partial acdS genes is effective for screening and identifying ACCD-producing bacteria
E295 and L322 residues are the key positions for differentiating ACCDs from homologs and for designing degenerate primers for specific amplification of the acdS gene.We used the consensusdegenerate hybrid strategy to broaden the primer spectrum.The CODEHOP-based PCR amplified partial acdS genes from a wide range of proteobacteria and actinobacteria.Notably, the CODE-HOP also worked for amplification of partial acdS sequences from the fungus Fusarium graminearum PH-1 (data not shown).However, the extent and power of the CODEHOP was not adequately tested, due to our limited collection of bacteria and fungi.
The specificity of the CODEHOP was supported by the amplification results.First, all the predicted amplification products from the test isolates are partial acdS genes.Second, no nonspecific product close to the predicted amplification products was amplified from the test isolates under optimized amplification conditions.Third, no acdS homolog was amplified from the bacteria containing acdS homologs, such as P. stutzeri A1501.DCyD is a characteristic of bacteria belonging to the family Enterobacteriaceae (Nagasawa et al. 1985).However, no dcyd was amplified from the major test isolates belonging to the genera Enterobacter and Klebsiella, such as E. sacchari (now Kosakonia sacchari) SP1 and K. variicola DX120E containing the dcyd gene (Chen et al. 2014;Lin et al. 2015).
The CODEHOP-based amplification of partial acdS genes did not produce false positives or false negatives.All the 57 acdSpositive test isolates were positive for the assay of ACC consumption, whereas the acdS-negative isolates were negative.The colorimetric ninhydrin assay of ACC with chimney-top 96well PCR plates to determine ACC consumption is a simple and effective method for screening ACCD-producing bacteria  (Li et al. 2011), but is not able to identify ACCD-producing bacteria that do not show ACCD activity in free-living states (Ma et al. 2003;Blaha et al. 2006;Nascimento et al. 2012).PCR amplification of the acdS gene is independent of the bacterial ACCD activity.Notably, the CODEHOP-based PCR amplification obtained partial acdS genes (accession no.KP117011 and KP117012) from mesorhizobial strains (CCBAU 61382 and CCBAU 61385 provided by CCBAU, Beijing, China) that did not show ACC consumption or ACCD activity in free-living states (data not shown).
The CODEHOP pair of acdSf3/acdSr3 is likely superior to previous degenerate primers for specific amplification of the acdS gene from a wide range of bacteria.Shah et al. (1998) designed P1/P2 primers and amplified acdS sequences from two of eight strains displaying ACCD activity.Govindasamy et al. (2008) used the F1936/F1938 primers designed by Blaha et al. (2006) and amplified acdS sequences from three of 14 strains displaying ACCD activity.Hontzeas et al. (2005) designed the DegACC5 /DegACC3 primers with very high degeneracy degrees of 3456 and 1728.In the DegACC3 primer, the three degenerate nucleotides, HKY corresponding to the codon of Glu295, match to Lys, Asn, Thr, Glu, Asp and Ala.The high degeneracy reduced the amplification specificity and resulted in the non-specific amplification of acdS homologs from A. xylosoxidans BM1, Achromobacter sp.CM1, E. aerogenes CAL3, P. marginalis DP3, P. syringae GR12-2 and Se.proteamaculans SUD165.These six strains, which displayed ACCD activity (Hontzeas et al. 2005), may contain both ACCD and ACCD homologs as is the case for A. xylosoxidans NBRC 15126, P. psychrotolerans L19, P. syringae DC3000 and Serratia sp.M24T3 (Figs 1 and 2), or may not contain ACCD as is the case for E. aerogenes KCTC 2190 and Se.proteamaculans 568 (Figs 1 and 2), but show non-specific ACCD activity by ACCD homologs.

ACCD-producing P. stutzeri, Enterobacter, Klebsiella and Bacillus have been overestimated
Pseudomonas stutzeri A1501 was negative for the amplification of the acdS gene and the assay of ACC consumption.Its putative acdS gene was found by genome sequencing (Yan et al. 2008), and all P. stutzeri genomes deposited in the NCBI database contained putative acdS genes.Here, we show that these putative acdS genes do not encode ACCD, but instead ACCD homologs (Figs S2 and S3,Supporting Information).
All the test isolates belonging to the genera Enterobacter, Klebsiella and Bacillus were negative for the amplification the acdS gene and the assay of ACC consumption.Moreover, no acdS gene was identified in the genomes of Enterobacter, Klebsiella and Bacillus strains deposited in the NCBI database.In contrast, many strains classified into the genera Enterobacter, Klebsiella and Bacillus were described as growing with ACC as the sole nitrogen source (e.g.Timmusk et al. 2011;Marasco et al. 2012;Kadyan et al. 2013) or showing ACCD activity (e.g.Ghosh et al. 2003;Babu, Kim and Oh 2013;Barnawal et al. 2013;Gupta et al. 2014;Kim et al. 2014;Xu et al. 2014).However, these findings fall short of direct molecular evidence.
Putative acdS sequences have been amplified from three Enterobacter strains [UW4, CAL2 (Shah et al. 1998) and CAL3 (Hontzeas et al. 2005)], two Klebsiella strains [K.pneumonia Acd-SPB2 (Chen et al. 2013) and K. oxytoca Rs-5 (accession no.FJ357241)] and two Bacillus strains [B.cereus AcdSPB4 (Chen et al. 2013) and B. subtilis MBPSB207 (Kumar, Kumar and Pratush 2014)].The putative acdS sequence of CAL3 (Hontzeas et al. 2005) is a non-specifically amplified acdS homolog (Fig. 2); CAL3 may not contain ACCD.UW4 and CAL2 had been likely misclassified as E. cloacae, based on their fatty acid profiles (Shah et al. 1997), but strain UW4 was recently reclassified into Pseudomonas (Duan et al. 2013).CAL2 may also belong to Pseudomonas because it was initially designated as Pseudomonas, based on its growth on selective media and fluorescent siderophore production (Glick, Karaturovíc and Newell 1995).Chen et al. (2013) amplified the acdS sequences of K. pneumonia AcdSPB2, B. cereus AcdSPB4 and Se.rubidaea AcdSPB1 and used Pseudomonas sp.UW4 as the positive control.Coincidentally, we found that the acdS sequences of AcdSPB2, AcdSPB4 and AcdSPB1 had only two nucleotides, four nucleotides and one nucleotide different from the acdS sequence in the UW4 genome (CP003880), respectively, and were phylogenetically grouped with the acdS gene of UW4 (Fig. 2).A possibility is that the acdS genes of these Bacillus, Serratia and Klebsiella strains were acquired from Pseudomonas through horizontal gene transfer.Another possibility is that the almost identical acdS sequences of the three strains were amplified after contamination of the amplification products from the positive control strain UW4; the nucleotide difference between the UW4 acdS and the UW4-derived acdS may have been generated by amplification and sequencing.This possibility weakens the molecular evidence for ACCD-producing bacteria belonging to the genera Klebsiella and Bacillus.
We suggest that the molecular detection of the acdS gene, the phylogenetic differentiation between the acdS gene and homologs, and the structural differentiation between ACCD and homologs are necessary for the identification of ACCDproducing bacteria.When putative acdS sequences are obtained by PCR amplification or from genome sequences, phylogenetic analysis of the putative acdS sequences and identified acdS and acdS homolog sequences (Fig. S3, Supporting Information) differentiates acdS from acdS homologs.When putative acdS sequences include the coding region of the protein C-terminal, alignment of the deduced amino acid sequences at the key 295 and 322 positions differentiates ACCD from homologs.For example, the putative acdS gene found in Rhizobium sp.TAL1145 (Tittabutr et al. 2008) phylogenetically diverged from the identified acdS sequences and grouped with the acdS homolog of Py. horikoshii (Fig. 2); its deduced amino acid sequence contains two Thr residues but not Glu and Leu at the aligned 295 and 322 positions (Fig. 1).Therefore, it can be assumed that the strain TAL1145 does not contain ACCD.This assumption explains the contradictory results that acdS expression in TAL1145 was not induced by ACC and that bacterial growth was inhibited by ACC (Tittabutr et al. 2008).

Horizontal transfer of acdS genes among bacteria has been overestimated
The early suggestion of the extensive horizontal transfer of acdS genes among bacteria was based on phylogenetic analyses of acdS sequences including non-specifically amplified acdS homologs (Hontzeas et al. 2005; as discussed above) and unreliably classified bacteria (Blaha et al. 2006).The unreliably classified bacteria are the strains UW4 and CAL2 discussed above and the strain ACP.ACP, which was designated as Pseudomonas by limited phenotypic tests (Honma and Shimonura 1978), may belong to Burkholderia (Nascimento et al. 2014).
Misannotation of acdS homologs as acdS genes led to the overestimation of ACCD-producing bacteria, such as the P. stutzeri strains whose genomes have been made available, and may also lead to an overestimation of the horizontal transfer of acdS genes (Nikolic, Schwab and Sessitsch 2011).
Our doubt about the contaminated amplification of acdS sequences discussed above also weakens the phylogenetic evidence (Fig. 2) for the horizontal transfer of acdS genes from Pseudomonas into Bacillus, Serratia and Klebsiella.Notably, when we did PCR amplification on the ACCD-producing Burkholderia, Herbaspirillum and Pseudomonas isolates, we found crosscontamination of amplification products and later took a great care to avoid contamination among samples.We thus call attention to the intergeneric transfer of the acdS gene among bacteria tested in the same study, which may be artifacts of crosscontamination of amplification products.
The acdS genes of H. seropedicae SmR1 and P. psychrotolerans L19 formed a unique phylogenetic group distinct from the acdS genes of other betaproteobacteria and gammaproteobacteria (Fig. 2).Most acdS genes in proteobacteria have an acdR gene encoding the ACCD regulatory protein nearby (Prigent-Combaret et al. 2008;Nascimento et al. 2014), whereas most acdS homologs do not have an acdR homolog nearby (Nikolic, Schwab and Sessitsch 2011).Herbaspirillum seropedicae SmR1 and P. psychrotolerans L19 have an acdR gene nearby the acdS gene and may have acquired both the acdS and the acdR genes simultaneously from a different class of bacteria yet to be determined through horizontal gene transfer (Nascimento et al. 2014).We propose that the acdS sequences of H. seropedicae SmR1 and P. psychrotolerans L19 can be used as the reference borderline of the acdS clade for phylogenetic analysis of the test putative acdS sequences.

CONCLUSION
ACCD-producing bacteria in ecosystems have been overestimated due to simply identifying them by growth on minimal media containing ACC as the sole nitrogen source, detection of non-specific ACCD activity, non-specific amplification of acdS homlogs and misannotation of acdS genes in genomes.Horizontal transfer of the acdS gene has been overestimated due to non-specific amplification of acdS homologs, misclassification of ACCD-producing bacteria and misannotation of acdS genes in genomes.Phylogenetic and structural differentiating the acdS gene and ACCD from homologs is crucial for the molecular identification of ACCD-producing bacteria and for understanding the evolution of the acdS gene.We propose an effective method for screening and identifying ACCD-producing bacteria: amplification of the partial acdS genes from bacterial cultures without DNA extraction using the CODEHOP acdSf3/acdSr3, and identification of amplified sequences by phylogenetic analysis with reference acdS and acdS homolog sequences.

Figure 1 .
Figure 1.Alignment of amino acid sequences showing the key difference between ACCDs and homologs at the 295 and 322 residues (position numbers in the ACCD of the model strain Pseudomonas sp.UW4).Note that E295 and L322 are unique to ACCDs.The ACCDs consist of identified ACCDs (filled circles) from various bacteria and fungi and ACCDs (filled squares) annotated in genomes of various bacteria.ACCD homologs consist of a d-cysteine desulfhydrase (DCyD) from Es. coli (open circle), an ACCD homolog from Py. horikoshii (open circle), a putative ACCD from Rhizobium sp.TAL1145 (open square) and putative ACCDs and DCyDs (open sqaure) annotated in genomes of various bacteria.Accession numbers of the proteins in the NCBI databases are given in parentheses.The alignment was created by the MUSCLE program integrated in the MEGA 5.2 program with default parameters.contained two Thr residues at the aligned 295 and 322 positions (Fig. S2, Supporting Information).

Figure 2 .
Figure 2. Phylogenetic tree constructed for selected acdS and acdS homolog sequences.The acdS sequences consist of acdS genes encoding identified ACC deaminases from various bacteria and fungi and acdS genes annotated in genomes of various bacteria.AcdS homologs consist of a d-cysteine desulfhydrase gene (dcyd) from Es. coli (open circle), an acdS homolog from Py. horikoshii (open circle) and a putative acdS gene from Rhizobium sp.TAL1145 (filled square) and putative acdS genes and dcyd genes (open square) annotated in genomes of various bacteria.Unreliably classified strains (filled circle), non-specifically amplified acdS homologs (inverted filled triangle) and probably cross-contaminated acdS sequences (filled triangle) are marked.Note that the acdS sequences of H. seropedicae SmR1 (filled diamond) and P. psychrotolerans L19 (filled diamond) are at the border of the acdS clade.Accession numbers of the nucleotide sequences in the NCBI databases are given in parentheses.The phylogenetic tree was constructed by the MEGA 5.2 program and the neighbor-joining algorithm.Bootstrap values (≥50%) based on 1000 replications are shown at branch nodes.Bars present 0.1 substitutions per nucleotide position.

Figure 3 .
Figure 3. Alignment of nucleic acid sequences for the acdSf3, acdSr3 and acdSr4 primers and corresponding sequences in acdS sequences (above the primer sequences) and acdS homolog sequences (below the primer sequences).

Table 1 .
Primers used in this study.
a Nucleotide position in the acdS sequence of Pseudomonas sp.UW4.b 3 degenerate core regions are underlined.