Abundance and distribution of the highly iterated palindrome 1 (HIP1) among prokaryotes

Abstract

We have studied the abundance and phylogenetic distribution of the Highly Iterated Palindrome 1 (HIP1) among sequenced prokaryotic genomes. We show that an overrepresentation of HIP1 is exclusive of some lineages of cyanobacteria, and that this abundance was gained only once during evolution and was subsequently lost in the lineage leading to marine pico-cyanobacteria. We show that among cyanobacterial protein sequences with annotated Pfam domains, only OpcA (glucose 6-phosphate dehydrogenase assembly protein) has a phylogenetic distribution fully matching HIP1 abundance, suggesting a functional relationship; we also show that DAM methylase (an enzyme that has the four central nucleotides of HIP1 as is site of action) is present in all cyanobacterial genomes (independently of their HIP1 content) with the exception of marine pico-cyanobacteria whom might have lost this enzyme during the process of genome reduction. Our analyses also show that in some prokaryotic lineages (particularly in those species with large genomes), HIP1 is unevenly distributed between coding and non-coding DNA (being more common in coding regions; with the exception of Cyanobacteria Yellowstone B’ and Synechococcus elongates where the reverse pattern is true). Finally, we explore the hypothesis that the HIP1 can be used as a molecular “water-mark” to identify horizontally transferred genes from cyanobacteria to other species.

Introduction

Highly Iterated Palindrome 1 (HIP1) is a short sequence of eight nucleotides (5′-GCGATCGC-3′) first described in Synechococcus PCC 6301 flanking an adaptive gene-deletion event.1 Initial sequence analysis of GenBank/EMBL/DDBJ DNA Nucleotide Sequence Data Libraries indicated that this motif is abundant only in cyanobacterial species and that it is present in coding as well as in non-coding regions.1 A further in-depth study has confirmed previous observations, but also showed that HIP1 is not common among all lineages of cyanobacteria.2

However, several questions remain regarding the evolution and the adaptive significance (if any) of this sequence. For instance, previous studies were not able to resolve the polyphyletic or monophyletic nature of individual HIP1 motifs among homologous sequences, nor the ancestral or derived state of this character among cyanobacterial lineages. Regarding the origin and mobility of this sequence, it was suggested (based on the lack of gaps surrounding HIP1 motifs into aligned homologous sequences), that HIP1 propagates in cyanobacteria via nucleotide substitutions.3 The rationale behind previous conclusion was as follow, if HIP1 would propagate through insertion events in coding sequences, downstream-frameshifts generated by this octanucleotide would have to be accommodate via additional insertion or deletions events. The analysis of 68 genes from Synechococcus PCC 7942 aligned to their corresponding homologs from Escherichia coli showed a lack of such additional deletion or insertions, therefore supporting the hypothesis of propagation via nucleotide substitutions.3

It is not known whether HIP1 confers some kind of advantage to its host, or if it is a molecular parasite (or the footprint left by a selfish genetic element). If HIP1 propagates through nucleotide substitutions (i.e., point mutations originating an HIP1 sequence are selected for), then a legitimate question is, what is the function encoded by this sequence? It has been speculated that one possibility is that HIP1 could provide desiccation tolerance by promoting intramolecular illegitimate recombination, much in the same way as it happens in Deinococcus radiodurans.3 Previous analysis have shown that HIP1 can indeed promote recombination.1,3 However there is no evidence of any putative recombinase which recognizes HIP1, and attempts to identify a protein complex associated with HIP1 have remained unsuccessful.3 As of today, the only enzyme proposed to interact with HIP1 is D12 class N6 adenine-specific DNA methyltransferases (DAM methylases).3,4 This is because the central 4 nt of HIP1 (5′-GATC-3′) is the site of this enzyme.3,4

More recent studies showed that one kind of repetitive element named Small Dispersed Repeat 5 (SDR5) is found interrupting HIP1 motifs between their third and fourth nucleotide in Nostoc punctiforme (i.e., 5′-GCG|ATCGC-3′, where the vertical bar indicates the place of insertion of SDR5).5 This repeat (SDR5) consist of 21 nucleotides that seems to fold into an stable secondary structure.5 The relationship between HIP1 and SDR5 is not understood, but it has been hypothesized that SDR5 could integrate into HIP1 through target-primed reverse transcription.5

In this study we address the phylogenetic distribution of HIP1 among completely sequenced prokaryotic genomes, its frequency of occurrence among coding and non-coding regions, and its co-occurrence with DAM methylases and other proteins among cyanobacteria. Finally, we test the hypothesis that HIP1 could serve as a water-mark of horizontally transferred genes from cyanobacteria to other prokaryotic lineages.

Results

Occurrence of HIP1 among cyanobacteria and other sequenced prokaryotic genomes.

As shown in Figure 1, some lineages of cyanobacteria have, by far, the largest ratio of observed vs. expected number of copies of HIP1 in their genome. This excess of copies of HIP1, is not related to GC-content content (Fig. 2), nor genome size (Fig. 3).

Figure 1 Abundance of HIP1 on sequenced prokaryotic genomes. Each dot represents a replicome. The figure shows the observed over the expected Log₁₀(obs/exp) number of copies of HIP1, vs. the squared difference of observed minus expected Log₁₀[(obs - exp)²/exp] number of copies of HIP1 per replicome. Expected numbers of copies of HIP1 are calculated directly from nucleotide frequencies.

Figure 2 Number of copies of HIP1 vs. GC-content content per replicome.

Figure 3 Number of copies of HIP1 vs. genome size per replicome.

The specie with the largest excess of HIP1 in its central replicome is the marine diazotrophic cyanobacteria Trichodesmium erythraeum, followed by some heterocyst forming cyanobacteria (Nostoc punctiforme, Anabaena sp. PCC7120 and Anabaena variabilis) and the obligate autotrophic fresh water cyanobacteria Synechococcus elongatus. On the other hand, the group of cyanobacteria with fewest copies of HIP1 comprises the early diverging species (Gloeobacter violaceus, Cyanobacteria Yellowstone A-Prime and B-Prime) and the marine pico-cyanobacteria (in particular those from Prochlorococcus genus) (Table S1).

To have an evolutionary perspective on the abundance/distribution pattern of this palindromic sequence, we have reconstructed a 16SrRNA phylogenetic tree and mapped the occurrence of HIP1 among lineages (Fig. 4). Contrary to what has been previously reported,3 HIP1's abundance do seem to follow a simple phylogenetic distribution. An excess of HIP1 starts at the branch separating early diverging cyanobacteria (G. violaceus, Cyanobacteria Yellowstone A-Prime and B-Prime) from the rest of species, and terminates at the branch leading to marine picocyanobacteria (Fig. 4).

Figure 4 16SrRNA Neighbor-Joining tree from cyanobacterias whose genomes have been completely sequenced. Branches are colored according to the observed over expected (obs/exp) number of copies of HIP1 (see the color code at the bottom of the figure). The actual value of (obs/exp) for each replicome is shown in parenthesis. A brown dot indicates those genomes with 10 times more copies of HIP1 than expected by chance, and a value larger than 1000 for the sum of squared differences (i.e., the rightmost cyanobacteria from Figure 1). A gray dot indicates the particular case of Cyanobacterium UCYN-A which has a value smaller than 1000 for the squared differences while having 18 times more copies of HIP1 than expected by chance. Arrows 1 and 2 indicate the branch where the excess of HIP1 was gained and lost, respectively. For reference, the branch leading to heterocyst forming cyanobacteria is indicated. Bootstrap values shown are for Neighbor-Joining/Maximum-Likelihood trees. When the node has a value smaller than 50 (or is inexistent in the Maximum-Likelihood tree) a dash is shown instead.

Phylogenetic distribution of DAM methylase.

As mentioned above, it has been noted that the central 4 nt of HIP1 (5′-GATC-3′) is the site of action of DAM methylase.3,4 To see if there is an association between DAM methylases and the abundance of HIP1, we have studied the phylogenetic distribution of this enzyme [Pfam family: MethyltransfD12 (PF02086)] among cyanobacterial genomes. As shown in Table 1 (and Fig. S1), DAM methylases are present in all genomes with the exception of marine pico-cyanobacteria.

Proteins associated to high and low HIP1 copy number replicomes.

In order to identify candidate proteins that might be responsible of generating high copy numbers of HIP1, we have calculated Pfam domains for all cyanobacterial proteins, and then, asked for those proteins (according to their domain structure) shared among all genomes with high HIP1 copy numbers, and absent in low HIP1 copy number genomes. To check if the excess of HIP1 is due to a secondary loss (instead of an acquisition of a new protein), we have also asked for the inverse pattern, this is, those proteins shared among all low HIP1 copy numbers, and absent in all high HIP1 copy number genomes.

The percentage of proteins annotated with at least one Pfam domain in each genome, varied from 53% in Microcystis aeruginosa to 79% in Cyanobacterium UCYN-A, with a mean value of 67% (Table S2). When we asked for single protein domains found among all high HIP1 copy number genomes with the exclusion of low HIP1 copy number genomes (and vice versa), we found none. However, if we ask for protein-domain architectures instead of single protein domains (where an architecture is composed of one or more Pfam domains in a specific order), we do find a candidate family (OpcA) with the desired phylogenetic distribution.

Interestingly, the OpcA protein is composed of two domains among high HIP1 copy number genomes, and of a single domain among low HIP1 copy number genomes, therefore satisfying both phylogenetic conditions at the same time (Table 1 and Fig. S2). The domain shared by all cyanobacteria (having high and low HIP1 copy numbers genomes) is named OpcA_G6PD_assem after Glucose-6-phosphate dehydrogenase subunit (PF10128); and the domain found only among high HIP1 copy number genomes is named PG_binding_1 from Putative peptidoglycan binding domain (PF01471). According to Pfam database, proteins containing PG_binding_1 (peptidoglycan binding domains) are mostly involved in bacterial cell wall degradation.6

A phylogenetic analysis of the OpcA_G6PD_assem domain, shows that marine pico-cyanobacteria acquired their OpcA protein directly from early diverging cyanobacteria (Fig. 5), very likely through horizontal gene transfer. To discard long branch attraction as an explanation for the branching pattern of marine pico-cyanobacteria, we have eliminated poorly aligned positions from the multiple alignment by using Gblocks7 with default parameters, and we have obtained the same branching pattern for marine pico-cyanobacteria (data not shown).

Figure 5 Maximum-Likelihood tree of cyanobacterial OpcA_G6PD_assem protein domain. As shown, marine pico-cyanobacteria branches from early diverging cyanobacteria. Domain structure of OpcA proteins are also shown. The star indicates the likely position of the event of PG_binding_1 fusion to OpcA_G6PD_assem domain. Only bootstrap values larger than 50 are shown.

Distribution of HIP1 among coding and non-coding regions.

We have analyzed whether or not HIP1 is distributed randomly between coding and non-coding regions. The result of a χ² test for each of the sequenced prokaryotic genomes is shown in Figure 6. Accordingly, there are a number of replicomes on which there is a bias in the distribution of HIP1 between coding and non-coding DNA. Among cyanobacteria (and in most cases), HIP1 is found more often in coding regions, with the exception of S. elongatus and Cyanobacteria Yellowstone B-Prime, where the opposite pattern is true (Table S3). Although the result of an heterogeneity χ² test indicate that replicomes with and without bias in their HIP1 distribution belong to two different populations, this result has to be taken cautiously because there seems to be a relationship between genome size and rejection of the null hypothesis of no difference in the abundance of HIP1 in coding and no-coding regions (i.e., the mean genome size of replicomes where the difference is significant is larger than otherwise; Fig. S3).

Figure 6 χ² statistics to test the null hypothesis of no difference in the abundance of HIP1 between coding and non-coding regions. Each dot represents a replicome. The horizontal black line indicates the critical χ² value for 1df and α of 0.05.

Presence of HIP1 in DNA Sequences from Species Other than Cyanobacteria

To test the idea that genes with abundant copies of HIP1 in non-cyanobacterial species arrived by horizontal gene transfer from cyanobacteria with large number of copies of HIP1, we have scanned GenBank nt database for non-cyanobacterial genes with copies of HIP1. First, we studied the abundance of HIP1 per-gene in genome sequences from cyanobacteria and other prokaryotes (Fig. 7 and Table S4). In an attempt to have a balance between sensibility and specificity, we decided to use as a cutoff value, four copies of HIP1 per gene. This is, because the fraction of genes having four or more copies of HIP1 is ∼0.004 for non-cyanobacterial genes and ∼0.09 for cyanobacterial genes, (i.e., there are approximately 20 cyanobacterial genes for each non cyanobacterial genes). Since we were interested in detecting genes (or small groups of contiguous genes) that could have been transferred from cyanobacteria to other species, only sequences with less than 100,000 bp in GenBank nt database were considered for this analysis. Results are shown in Figure 8. Only 99 DNA sequences from species other than cyanobacteria have less than 100,000 bp and four or more copies of HIP1 in GenBank nt database. Noteworthy, 53 of them belong to the phylum Chlorophyta (species: Ostreococcus lucimarinus, O. tauri, Micromonas sp RCC299 and M. pusilla). These species belong to the Prasinophyceae which are ancient members of the green lineage; the lineage giving rise to higher plants.8 Whether or not sequences from these species have a cyanobacterial origin requires more in depth phylogenetic analyses. Other species having 4 or more copies of HIP1, belong to different phyla from Bacteria, Eukarya and some phages (Table S5).

Figure 7 Distribution of HIP1 abundance per gene among cyanobacteria and other prokaryotes. The fraction of genes in each category is calculated as [(observed number of genes having x_i number of copies of HIP1) / (total number of genes having copies of HIP1)]. Where x_i = 1, 2, … 20. The arrow indicates the cutoff value to search for candidate xenologous cyanobacterial genes in non-cyanobacterial species.

Figure 8 Number of copies of HIP1 vs. sequence length from GenBank entries. Only sequences having 4 or more copies of HIP1 and less than 100,000 bp are shown. Different circle sizes are only for visualization purposes.

Discussion

Although there are some prokaryotes other than cyanobacteria showing relative large copy numbers of HIP1 in their genomes (like the Actinobacteria Conexibacter woesei, the Alphaproteobacteria Sphingomonas wittichii, and the Deltaproteobacteria Sorangium cellulosum; Table S1) all of them have relative high GC-content content values (0.72, 0.68 and 0.71 respectively), a factor that affects the abundance of HIP1 (Fig. 2). Therefore, large copy numbers of HIP1, not related to GC-content content, is clearly a phenomenon restricted to some of the cyanobacterial lineages (Figs. 1 and 4).

It is not yet known which molecule(s) is responsible for the origin and proliferation of HIP1 among cyanobacteria; or the adaptive value (if any) of this palindromic sequence. At present times, only DAM methylases have been proposed to interact with HIP1.

Lack of DAM methylases (following its Pfam9 domain definition) and low copy numbers of HIP1 among marine pico-cyanobacteria are clearly derived characters of this lineage of bacteria (i.e., the ancestral state was abundance of HIP1 and presence of DAM methylase, as suggested by the rooting of the 16SrRNA tree from Fig. 4). The low copy number of HIP1 might be related to the process of genome reduction suffered by this lineage of bacteria.10 For instance, a common trend among prokaryotes with reduced genomes is a low GC-content content.11 This in turn, influences the abundance of HIP1. During the course of evolution of genome reduction, the number of copies of HIP1 diminished as the GC-content content dropped. This process seems to have been particularly strong for Prochlorococcus spp where the average GC-content content is ∼0.36 and the lowest values of observed and expected numbers of HIP1 are found among cyanobacteria (Table S1). On the other hand, the enzyme DAM methylase seems to have been lost early during the course of genome reduction of marine pico-cyanobacteria. Its deletion may not be related to the low number of copies of HIP1 among these marine bacteria, since other bacteria, like the early diverging Cyanobacteria Yellowstone A' and B' have lower numbers of copies of HIP1 than marine Synechococcus while coding for DAM methylase.

Among early diverging cyanobacteria, the rarity of HIP1 is not related to GC-content content, which is relatively high (∼0.6). These bacteria have an HIP1 content which is not particularly different from that of other prokaryotes (Table S1). They represent, the “ancestral state” prior to the proliferation of HIP1 among newcomer cyanobacterial lineages. Some genes absent in the genomes of early diverging cyanobacteria and marine pico-cyanobacteria must be the responsible for the large proliferation of HIP1 among the rest of cyanobacterial lineages studied here.

The concordance between the evolution of OpcA and HIP1 abundance suggest a functional relationship. The branching of marine pico-cyanobacterial OpcA among early diverging cyanobacteria suggests that marine Synechococcus and Prochlorococcus acquired their OpcA through horizontal gene transfer from low HIP1 copy number cyanobacterias. This adds to the hypothesis that HIP1 abundance is somehow related to the kind of OpcA coded in the genome (i.e., high HIP1 copy number cyanobacteria have an OpcA protein with a PG_binding_1 domain added). However, from the function of OpcA it is not clear how this protein can be functionally related to the proliferation of HIP1. OpcA has been shown to acts as an allosteric activator of Glucose-6-Phosphate dehydrogenase (G6PD) in Nostoc punctiforme, which is an essential enzyme for nitrogen fixation and dark heterotrophic growth.12 Being OpcA an allosteric activator, it could be hypothesized that OpcA also activates the molecule responsible of proliferating HIP1. However, a spurious association between OpcA and HIP1 abundance can't be ruled out. After all, approximately 33% of all proteins from cyanobacterial species studied here do not match any Pfam doman. It wouldn't be surprising that the molecular machinery to produce HIP1 is encoded among them.

An intriguing observation is the almost obligatory cooccurrence of SDR5 with HIP1 along the genome of the multicellular cyanobacterium Nostoc punctiforme.5 However, it is not clear if both sequences originate by the same molecular machinery or not. For instance, the phylogenetic distribution of HIP1 is larger than that of SRD5. SDR5 is confined only to heterocystous cyanobacteria5 while HIP1 has a much wider distribution. This observation suggests that HIP1 predates SRD5, but, it could be also possible that SDR5 has a different sequence outside heterocystous cyanobacteria and has been overlooked by previous analysis.5 This point clearly deserves further attention.

Here we show that in some prokaryotic genomes there is a statistically significant difference in the number of copies of HIP1 between coding and non-coding DNA. This result has to be taken cautiously because the propensity to reject the null hypothesis (of no difference) seems to be related to genome size. This is, the mean size of replicomes rejecting the null hypothesis is much larger (4,119,783 bp) than that of those replicomes not rejecting the null hypothesis (1,608,597 bp) (Fig. S3C). It is not known if this tendency is due simply by a sample-size effect, or if there are other parameters affecting the result. The data also show some other general tendencies (Fig. S3A and B). For instance, in most cases when the null hypothesis is rejected, is due to an excess of copies of HIP1 on coding regions than in non-coding regions, and only in a few cases the reverse pattern is true. Another general tendency is that, on average, statistically significant differences are due to an underrepresentation of copies of HIP1 in non-coding regions (rather than an overrepresentation of copies of HIP1 in coding regions) when compared with replicomes on which the difference is not statistically significant (Figure S3D). More in depth studies are required to understand the basis of the differences and tendencies outlined above. Nevertheless, it is likely that there is not a particular functional reason to have more copies of HIP1 in coding regions, because the pattern found in cyanobacteria is extensible to other prokaryotes where the difference of observed vs. expected number of copies of HIP1 is not as pronounced.

Being a distinctive feature of several cyanobacterial genomes, it could be possible to use HIP1 as a molecular “water-mark” to identify genes horizontally transferred into non cyanobacterial species. However, due to the amelioration process suffered by of horizontally transferred genes, only recent cases of HGT could be detected by this approach.13 From the analysis reported here, it is not possible to confirm that sequences from the Prasinophyceae protists originated in cyanobacteria. Spurious similarity it is also a perfectly possible explanation. Further analysis could consist in taking longer sequences than 100,000 bp and split them up into smaller regions in order to expand the search.

Regarding the molecular evolution of HIP1 several questions remain. Which molecules are responsible of originating HIP1? What is the relationship between SRD5 and HIP1? Does rich HIP1 sequences from Ostreococcus spp, and Micromonas spp truly originated in cyanobacteria? How does HIP1 affect the amino acid composition and the tertiary structure of proteins? On what kind of genes does HIP1 is found more often? We hope this work stimulates new lines of research on this very peculiar feature of some cyanobacterial lineages.

Materials and Methods

Genome databases and statistical analyses.

Complete sequenced prokaryotic genomes where downloaded from KEGG database (www.genome.jp/kegg/). HIP1 abundance, GC-content content, genome size and χ² were all estimated and calculated using Perl scripts available upon request. The expected numbers of copies of HIP1 per replicome were estimated as the product of the frequencies of individual bases conforming HIP1 multiplied by the total number of bases in each one of the replicomes. The critical value for the heterogeneity χ² was calculated using the formula provided by.14

Phylogenetic analyses.

The molecules of 16srRNA of each of the cyanobacterial genomes were downloaded, and aligned using MUSCLE software using default parameters.15 A Neighbor-Joining (500 bootstrap replications, Maximum Composite Likelihood distance estimation and uniform rates among sites) and Maximum-Likelihood (model Kimura-2P plus Gamma distribution with invariant sites, selected according to the Bayesian Information Criterion, 500 bootstrap replications) phylogenetic trees were reconstructed using MEGA5 software.16

Domain search.

Protein sequences from cyanobacterial genomes were searched with HMMER software using Pfam profile MethyltransfD12 (PF02086).17

GenBank analyses. The GenBank nt database18 was scanned for non-cyanobacterial DNA sequences having at least four copies of HIP1 and a maximum of 100,000 bp. The searh was implemented using a Perl script available upon request.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Abbreviations

HIP1	=	highly iterated palindrome 1
HGT	=	horizontal gene transfer
SDR	=	small dispersed repeat
OpcA	=	glucose 6-phosphate dehydrogenase assembly protein

Figures and Tables

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Table 1 Phylogenetic distribution of DAM methylase and OpcA proteins among cyanobacterial genomes

Species	DAM methylase	OpcA	PG-OpcA§	Log(obs/exp)¥
Gloeobacter violaceus	*	*		0.329240852
Cyanobacteria Yellowstone A-Prime	*	*		−0.108423649
Cyanobacteria Yellowstone B-Prime	*	*		−0.152610163
Thermosynechococcus elongatus	*		*	1.849962474
Acaryochloris marina	*		*	1.439737442
Cyanothece sp PCC 7425	*		*	1.591576294
Microcystis aeruginosa	*		*	1.62684149
Cyanothece sp PCC 7822	*		*	1.25460178
Cyanothece sp PCC 7424	*		*	1.905410355
Synechocystis sp PCC6803	*		*	1.846474569
Cyanobacterium UCYN-A	*		*	1.267171728
Cyanothece sp ATCC 51142	*		*	2.056541881
Cyanothece sp PCC 8802	*		*	2.059242916
Cyanothece sp PCC 8801	*		*	2.056171617
Synechococcus sp PCC7002	*		*	2.062752865
Trichodesmium erythraeum	*		*	2.162564407
Nostoc punctiforme	*		*	2.114004089
Anabaena azollae 0708	*		*	1.645422269
Anabaena sp PCC7120	*		*	2.087517289
Anabaena variabilis	*		*	2.085779943
Synechococcus elongatus PCC7942	*		*	2.09119783
Synechococcus elongatus PCC6301	*		*	2.08849047
Synechococcus sp WH7803		*		0.569390875
Synechococcus sp CC9311		*		0.545386086
Synechococcus sp RCC307		*		0.450387001
Synechococcus sp CC9902		*		0.49560466
Synechococcus sp WH8102		*		0.36339074
Synechococcus sp CC9605		*		0.333878907
Prochlorococcus marinus MIT 9303		*		0.414973348
Prochlorococcus marinus MIT 9313		*		0.393344257
Prochlorococcus marinus MIT 9211		*		−0.84509804
Prochlorococcus marinus MIT 9215		*		0.301029996
Prochlorococcus marinus SS120		*		−0.301029996
Prochlorococcus marinus NATL1A		*		−0.22184875
Prochlorococcus marinus MED4		*		−0.301029996
Prochlorococcus marinus NATL2A		*		−0.096910013
Prochlorococcus marinus MIT9312		*		0
Prochlorococcus marinus AS9601		*		0
Prochlorococcus marinus MIT 9515		*		0
Prochlorococcus marinus MIT 9301		*		0

§ OpcA proteins associated with PG_binding_1 domains.

¥ For reference, the Log₁₀(observed/expected) number of copies of HIP1 per replicome is shown.

Acknowledgments

L.D. wishes to thank CINVESTAV Unidad Irapuato for all facilities provided. Financial support was provided by grant PROMETEO 2009/092 from Conselleria d'Educació, Generalitat Valenciana, Spain, to A.M.