Abstract
Cyanobacteria thrive in very diverse environments. In Earth history however, delayed oxygenation has raised questions of growth limitation in ancient environmental conditions. As a single genus, the Thermosynechococcus are known to be cosmopolitan and live in chemically diverse habitats. To understand the genetic basis for this, we compared the protein coding component of Thermosynechococcus genomes. Supplementing the known genetic diversity of Thermosynechococcus, we report draft metagenome-assembled genomes of two Thermosynechococcus recovered from ferrous carbonate hot springs in Japan. We find that as a genus, Thermosynechococcus is genomically conserved, having a small pan-genome with few accessory genes per individual strain and only 18 protein clusters appearing in all Thermosynechococcus but not in any other cyanobacteria in our analysis. Furthermore, by comparing orthologous protein groups, including an analysis of genes encoding proteins with an iron related function (uptake, storage or utilization), no clear differences in genetic content, or adaptive mechanisms could be detected between genus members, despite the range of environments they inhabit. Overall, our results highlight a seemingly innate ability for Thermosynechococcus to inhabit diverse habitats without having undergone substantial genomic adaptation to accommodate this. The finding of Thermosynechococcus in both hot and high iron environments without adaptation recognizable from the perspective of protein coding genes has implications for understanding the basis of thermophily within this clade, and also suggests that ferrous iron in ancient oceans may not have inhibited the proliferation of Cyanobacteria on Earth. The conserved core genome may be indicative of an allopatric lifestyle – or reduced genetic complexity of hot spring habitats relative to other environments.
Introduction
Water oxidizing cyanobacteria fundamentally altered the distribution of carbon and electrons on Earth (Canfield, 1998; Raven, 2009; Ward and Shih, 2019). A marker of this reorganization is in the Great Oxygenation Event (GOE), which marks a major transition in the evolution of life on Earth ~2.3 billion years ago (Lyons et al., 2014; Fischer et al., 2016). While the GOE is widely accepted to have been driven by the production of O2 by oxygenic photosynthesis performed by members of the Cyanobacteria, the timing and proximal trigger for the GOE is debated. At least six hypotheses for the timing of the GOE can be considered i) the evolution of oxygenic photosynthesis by Cyanobacteria just before the GOE (Fischer et al., 2016; Shih et al., 2017), ii) an earlier evolution of Cyanobacteria with O2 accumulation delayed due to the transition of cyanobacteria from small-scale freshwater to large-scale marine environments (Sánchez-Baracaldo, 2015), iii) the transition from unicellular to multicellular organisms for increased evolutionary success (Schirrmeister et al., 2015), iv) the inhibition of early cyanobacteria due to high iron concentrations (Swanner, Mloszewska, et al., 2015; Swanner, Wu, et al., 2015), v) a possible nitrogen throttle on cyanobacterial growth (Fennel et al., 2005; Shi and Falkowski, 2008; Allen et al., 2019), vi) or depressed Archaean productivity due to phosphate availability (Hao et al. 2020). Although cyanobacterial taxonomy continues to be refined (Knoll 2006; Tomitani et al., 2006; Shih et al., 2013; Soo et al., 2019; Parks et al., 2020), understanding the ecological distribution of cyanobacteria may help us generate, or reject hypotheses about their evolutionary trajectories, and the factors which led to the observed timing of the GOE.
Cyanobacteria are found in a wide range of environments – the knowledge of which continues to expand (Whitton, 1992; Puente-Sánchez et al., 2018; Callieri et al., 2019). As a single phylogenetic group, members of the Thermosynechococcus genus have been documented to inhabit a range of chemical environments, some of which might be similar to places and times on the early Earth. Those Thermosynechococcus with genomes available are from hot springs which vary in temperature from 44 – 94 °C, pH ranging from 5.4 – 9.3, sulfate concentrations between 0.06 mM and 17.4 mM and iron concentrations between 0.0004 mM and 0.261 mM (Table 1) at their source. It is noteworthy that for two of the genomes - those from Okuoku-hachikurou and Jinata hot springs – the ferrous iron concentrations at the source by far exceed those at the other springs with 114 μM at OHK and 261 μM at Jinata (Ward et al., 2019; Cheng et al., 2020).
Seminal work by Papke et al. on phylotype:geographical relationships of Thermosynechococcus posed questions as to how these organisms could be so widely distributed: in their analysis, the distribution of Thermosynechococcus could not be explained by measured geochemical parameters. Thermosynechococcus thus appear to be cosmopolitan, but the basis for this remains unresolved. Motivated by the finding of Thermosynechococcus members in ferrous iron carbonate hot springs that we have been studying (Ward et al., 2017, 2019), and in an attempt to find a genetic basis for the geochemically wide distribution of the organisms, we took a comparative genomics approach to search for resolvable traits which may underlie environmental adaptations of Thermosynechococcus members.
Methods
Genome recovery
The OHK43 genome was recovered from genome-resolved metagenomic sequencing of samples from Okuoku-hachikurou Onsen (OHK) in Akita Prefecture, Japan, following methods described previously (Ward et al., 2019, 2020) and described briefly here. Samples of thin biofilms in the outflow of the hot spring were sampled for metagenomic sequencing in September 2016. DNA was preserved in the field with a Zymo Terralyzer BashingBead Matrix and Xpedition Lysis Buffer (Zymo Research, Irvine, CA) after disruption of cells in polyethylene sample tubes via attachment to and operation of a cordless reciprocating saw (Makita JR101DZ). Microbial DNA was extracted and purified after return to the lab with a Zymo Soil/Fecal DNA extraction kit (Zymo Research, Irvine, CA Quantification of DNA was performed with a Qubit 3.0 fluorimeter (Life Technologies, Carlsbad, CA). DNA was submitted to SeqMatic LLC (Fremont, CA) for library preparation using an Illumina Nextera XT DNA Library Preparation Kit prior to 2×100bp paired-end sequencing via Illumina HiSeq 4000 technology. Raw sequence reads were quality controlled with BBTools (Bushnell 2014) and assembled with MegaHit v. 1.02 (Li et al., 2016). The OHK43 genome bin was recovered via differential coverage binning with MetaBAT (Kang et al., 2015). Completeness and contamination/redundancy were determined with CheckM v1.1.2 (Parks et al., 2015). The genome was uploaded to RAST v2.0 for annotation and characterization (Aziz et al., 2008). Presence or absence of metabolic pathways of interest was predicted using MetaPOAP v1.0 (Ward et al., 2018). Taxonomic assignment was determined with GTDB-Tk v1.2 (Parks et al., 2018, 2020; Chaumeil et al., 2019). Genome assembly and quality statistics for OHK43 and other Thermosynechoccus genomes are reported in Supplemental Table 4.
Organismal Phylogenies
Concatenated ribosomal phylogenies were constructed following methods from Hug et al. (2016). Members of the Thermosynechococcaceae and outgroups were identified using GTDB (Chaumeil et al., 2019) and their genomes downloaded from the NCBI WGS and Genbank databases. Ribosomal protein sequences were extracted using the tblastn function of BLAST+ (Camacho et al., 2009) and aligned with MUSCLE (Edgar, 2004). Trees were built with RAxML v.8.2.12 (Stamatakis, 2014) on the Cipres science gateway (Miller et al., 2010). Transfer bootstrap support values were determined with BOOSTER (Lemoine et al., 2018). Visualization of trees was performed with the Interactive Tree of Life Viewer (Letunic and Bork, 2016).
Genome comparison
We compared the core- and pangenomes of Thermosynechococcus at the genus level, family level and with a sub-sample of Cyanobacteria across the GTDB defined class Cyanobacteriia. ProteinOrtho (Lechner et al., 2011) was used for the identification of Conserved Likely Orthologous Groups (CLOG) and analysis.
At the genus level 7 Thermosynechococcus strains from varying hot spring environments (Table 1 and 2) were compared. The genomic data from five available sequences of Thermosynechococcus strains T. sp. CL1/1-2178 (CL1), T. elongatus BP1/1-2178 (elongatus BP1), T. vulcanus NIES2134/1-2178 (vulcanus) T. sp. NK55a/1-2022 (NK55a), T. elongatus PKUAC-SCTE542 (elongatus PKUAC) and two unnamed strains from Jinata Onsen (Jinata) and Oku-Okuhachikurou Onsen (OHK) in Japan were compared using ProteinOrtho (Lechner et al., 2011), BLAST (Altschul et al., 1990), and FeGenie (Garber et al., 2019). Phylogenetic relationships between these strains were established using concatenated ribosomal protein phylogenies following Hug et al. (2016), taxonomic classifications with GTDB-tk (Chaumeil et al., 2019) and average nucleotide identities (Rodriguez-R and Konstantinidis, 2017).
For the analysis at the family level, we included 6 more species that appear as representative strains at the Thermosynochoccaceae family level in the Genome Taxonomy Data Base (GTDB). The representative GTDB strains include Acaryochloris marina MBIC11017, Cyanothece sp. PCC 7425, Acaryochloris sp. CCMEE 5410, Synechococcus sp. PCC 6312, Synechococcus lividus PCC 6715 and Acaryochloris sp. RCC1774. This resulted in a total of 13 strains for family level analysis.
At the class level we used the 7 genus level strains and additionally 16 species, which do not cover all of the family level species, 15 after (Beck et al., 2012) and one more Gloeobacter species (G. kilaueensis JS-1). Here the goal was to test the coherence of our analysis parameters when compared to previous studies of cyanobacterial core- and pangenomes (Beck et al., 2012, 2018). We acknowledge that the distribution of species within the families of the class is not even, however this comparison was used to verify our methods and the stability of the class level core. This resulted in a total of 23 genomes for the class level analysis.
It is important to note that although the spring source water properties differ, the source water is not necessarily where the DNA originated. Thus there is some uncertainty in the precise geochemistry:genome relationships discussed here. We also acknowledge that not all genomes in this study reach 100 % completeness. We might therefore be missing some gene clusters and some genomes might cluster differently if they were complete. We compared genomic data as of August 4th, 2020.
To compare the appearance of genes related to iron uptake and regulation we used FeGenie (Garber et al., 2019) with standard parameters. We used results from ProteinOrtho (Lechner et al., 2011) for further analysis of the core, shared, unique, TS core and TS shared clusters. Proteinortho was applied such that the output included also singleton clusters (only containing a single protein) and with an algebraic connectivity of 0 as a measure for the structure of the orthologous clusters. We did not obtain many differences when running ProteinOrtho on our data using a value of 0 or 0.1 (default) for the algebraic connectivity, however, a value of 0 resulted in slightly larger clusters for the core and a few less singletons. A comparison of results from FeGenie and ProteinOrtho resulted in similar gene clusters for iron related genes such that our results here are based on both program outputs.
Results and Discussion
Phylogeny of the Thermosynechococcus and the species proposal T. nakabusensis
The Thermosynechococcus genus is phylogenetically coherent within the Cyanobacteria (Figure 1) and the genome sizes of genus members are similar to one another (Table 2). Based on similarity observed with ANI and GTDB-tk (Table 2, supplementary Table 4), 4 species are present within the Thermosynechococcus genus: T. elongatus BP1 and T. vulcanus belonging to one species, T. NK55a, the Jinata and OHK genomes belonging to a second species, and T. CL1 and T. elongatus PKUAC-SCTE542 as one species each. For the species including the T. NK55a, Jinata, and OHK genomes we propose the name Thermosynechococcus nakabusensis as the first and so far only isolated organism which originates from Nakabusa hot spring in Nagano Prefecture, Japan.
Genus and family level comparison of Thermosynechococcus and Thermosynochococcaceae
Comparing the conserved likely orthologous groups (CLOGS), we analyzed i) the core-genomes: those CLOGS shared by all genomes in an analysis, ii) the shared CLOGs: those shared by at least 2 but not all of the genomes in the analysis, and iii) unique CLOGs: those CLOGs that are unique to a single genome (Table 3). The Thermosynechococcus genus specific core (core TS) comprises CLOGs shared by all 7 genus level genomes that are not present in any other species, and the Thermosynechococcus genus-specific shared CLOGs (shared TS) corresponds to CLOGs that are shared by at least 2 and at most 6 genus level genomes.
Comparing the genomes of the seven genus members, the protein core is made up of 1878 CLOGs and contains 80 to 89 % of the putative protein coding genes in a genome. This is higher than in other comparisons of organisms, for example in Reno et al. (2009) 69 – 79 % of genes from S. islandicus strains made up the core. Wu et al. (2018) found that the core genes of species belonging to the genus Comamonas account for 18 – 33 % of all genes, and Barajas et al. (2019) investigated the core genome within the genus Streptococcus which ranges in size from 9.6 % to 24 %. The high proportion of CLOGs which make up the core leaves few unique CLOGs for each genome and between genus members, and only slight variations in genome content between genome are observed: for example 27 CLOGs are specific to the 5 genomes that do not include Jinata and OHK and 28 CLOGs are specific to Jinata and OHK (Figure 2, supplementary Figure 1).
Golicz et al. (2020) suggested that the size of a pangenome is related to organismal lifestyles, with sympatric organisms having open pangenomes with many accessory genes and allopatric (isolated from other organisms) organisms having more closed and conserved pangenomes. The Thermosynechococcus genus could thus be considered as allopatric as they have a comparatively large core and few shared and unique genes. Chen et al. (2020) also noted that differences in horizontal gene transfer (HGT) are related to genome size with smaller genomes showing less HGT and bigger genomes having a larger probability that HGT occurred. They also found that hot spring cyanobacteria specifically have smaller genome sizes and less HGT into the genome. Excluding gene loss, we suggest that the conserved core might be indicative of a more ancient gene repertoire in hot spring cyanobacteria, with other cyanobacteria gaining more functionality through HGT over evolutionary timescales. Furthermore thermotolerance is phylogenetically scattered across the cyanobacterial tree of life, mostly in organisms comprising smaller genomes. Since all Thermosynechococcus in this study are found in hot springs – which typically have reduced microbial diversity in comparison to other environments such as soils (Ward et al., 1998) – this finding provides support for the still tentative hypothesis that hot spring environments may provide more limited opportunity for lateral gene transfer, which in turn could lead to less opportunity for lateral gene transfer and smaller genomes.
In contrast to the genus level where genome size varies from 2.25 MB to 2.64 MB, at the family level it varies to 8.36 MB (Acaryocholoris marina MBIC11017, Table 2). Running proteinortho analyses with the seven genus level Thermosynechococcus and the six family level sequences, the overall core is reduced by just over 30%, from 1878 to 1283 CLOGs. This number of ~1900 core CLOGs within the genus is different from a recent analysis (Cheng et al., 2020), with the higher number observed here being due to the inclusion of the newly available and revised T. elongatus PKUAC-SCTE542 genome. Shared genes, not core genes, comprise a larger percentage of the genomes for smaller genomes, while unique genes are abundant in larger genomes (Figure 2a).
Including the added diversity at the family level, the number of CLOGs shared between the 5 Thermosynechococcus genus members that do not include Jinata and OHK is reduced from 27 to 3, showing that these CLOGs are found in other closely related members at the family level. However, the number of CLOGs found only in Jinata and OHK changes from 28 to 25 CLOGs, highlighting that these are unique. Also observed when comparing the Thermosynechococcus genus across the Thermosynochococcaceae family, the specific CLOGs that appear in only, and all of the 7 genus level genomes is made up of 18 CLOGs, showing a low number of conserved Thermosynechococcus genus only CLOGs. From the CLOG perspective, the seven members share much in common with their closest relatives, but they appear to lack major distinguishing qualities.
Genomic differentiation of the Thermosynechococcus genus from other cyanobacteria
At the class level the number of Thermosynechococcus genus level core CLOGs increases when compared to the family level due to the exclusion of some family level species at this level (see Methods and Table 3). The overall core decreases by almost 60% when class level representatives are added (1878 core CLOGs at the genus level and 742 core CLOGs at the class level). This is expected as Beck et al. (2018) suggested that the clustering of CLOGs depends on the variability in genome size as well as phylogenetic distance between the analyzed genomes as well as the total amount of genomes analyzed. Moreover this is also in line with the larger analyses of microbial genomes which have shown that the continued addition of taxonomic diversity in an analysis in an analysis leads to increasingly smaller cores (Charlebois and Doolittle, 2004; Lapierre and Gogarten, 2009).
Overall these results conform with Beck et al. (2012), with variations attributed to the different methods and parameters used for each analysis. Additionally, we confirm that the class level core is stable when adding a higher diversity of organisms, similar to Beck et al. (2018), as the number of core family CLOGs only slightly decreases when adding more species at the class level.
At the class level, we looked at specific cases among the shared, but not unique CLOGs, to investigate gene content which may differentiate the Thermosynechococcus genus from other cyanobacteria. In the case of CLOGs which appear in all Thermosynechococcus but not in any other cyanobacteria in our analysis, 79 CLOGs were found, 47 of which are hypothetical proteins of unknown functions and could be of interest to further biochemical studies to understand the adaptations of the genus. CLOGs unique to the genus appear to be of known functionalities account for a variety of processes with genes involved in transcriptional regulation, transporters and membrane proteins (e.g. acetyltransferases, glycosidases and ATPases; more information in supplementary Table 2). Since all the Thermosynechococcus analyzed are from hot springs, these genes potentially provide some basis for that lifestyle.
Adaptation of Thermosynechococcus to their respective environments
We were especially interested in those CLOGs that are shared between the strains from environments with elevated iron concentrations (Jinata and OHK) but which are not present in any other cyanobacteria. Previous studies have shown that some cyanobacteria express higher levels of genes involved in iron ion homeostasis which are expressed in iron limiting conditions (Cheng and He, 2014), and here we looked at the presence or absence of iron related gene products in Thermosynechococcus compared to other cyanobacteria using FeGenie and BLAST comparisons. 24 CLOGs are specific to Jinata and OHK at the class level, some of which show high partial identity but low coverage matches with genes from other Thermosynechococcus (supplementary Table 4). It is notable that these 24 CLOGs do not appear in the third strain of the same species (NK55a). With our current understandings after considering results from BLAST and FeGenie, none of the 24 CLOGs comprises genes that could explain the organisms adaptation to elevated iron concentrations. 3 unique CLOGs are found in the 5 Thermosynechococcus that do not include Jinata and OHK, and we confirmed that none of those CLOGs are related to iron regulation as best as can be assessed from the sequence, similar to the Jinata and OHK CLOGs (supplementary Table 4). There is no sequence resolvable genomic signature specific to Jinata and OHK related to iron tolerance.
These genomes lack genes coding ferrous iron transport and uptake proteins EfeB, EfeO and EfeU, the metal transport gene ZupT, cellular iron storage proteins Bfr and the iron regulator under iron limiting conditions PfsR (Table 4). Genes encoding these proteins are found in other species of cyanobacteria, but not in the Thermosynechococcus genus members with currently available genomes. Within the genus there are no differences with regard to genes encoding proteins involved in iron regulation (PfsR), and in all cases the same genes encoding proteins related to ferrous iron uptake (FeoA, FeoB, YfeA and YfeB), ferric iron uptake or transport (ExbD, FutA, FutB and FutC), siderophore iron acquisition (FpvD), metal ion binding (Ho1 and Ho2) and iron starvation acclimation (IsiA) are present (Table 4).
The conserved genomic core of Thermosynechococcus in relationship to environmental distribution is unique
In addition to resolving a genetic basis for widespread environmental distribution, our work is also relevant to historical proliferation of cyanobacteria, since some modern-day hot springs and their biogeochemistries can be used as historical process analogues (Brown et al., 2005, 2007; Ward et al., 2019). Considering contemporary environments, the analysis of Thermosynechococcus also provide insight into island biogeography of microbes. Ionescu et al. (2010) observed that the speciation patterns of microorganisms are shaped by local community structures and environmental influences, and Bahl et al. (2011) additionally suggest a positive correlation between geographic and genetic distance. Papke et al. (2003) found that isolated environments such as geothermal springs may lead to evolutionary divergence of closely related Thermosynechococcus strains due to an island effect. Our analysis suggests that geographically widespread organisms belonging to the genus inhabit hot springs with varying geochemistries without genomically recognizable adaptations specific to their site of origin. Instead, the finding of highly conserved genomes within the genus, and furthermore, that the genetic content of the genus is not markedly different from other cyanobacteria, implies that the genus is inherently flexible and viable in the geochemical regimes studied. The large portion of shared genes within the genus provides a genetic basis for the lack of correlation between geographic and genetic distances within the genus found by Papke et al. (2003).
This genomic coherence at the genus level is in contrast to other studies, such as Reno et al. (2009; Wu et al. (2018) or Barajas et al. (2019)as mentioned above. Furthermore the organisms tested by Reno et al. (2009) are obligate thermoacidophilic archaea at the species level which are environmentally restricted similar to organisms here. They found that the core and pan genomes of these organisms are shaped by their geographical distribution and relatedness within and across different environments. Apparently, Thermosynechococcus is environmentally promiscuous, and have fewer restrictive requirements concerning their distribution.
Outlook
One aim of this study was to identify those proteins that make hot-spring inhabiting Thermosynechococcus genus members unique. Although the genus Thermosynechococcus is rather genomically conserved, 47 out of the 79 CLOGs unique to these genomes are identified as hypothetical proteins of unknown function and thus biochemical studies on these proteins is warranted.
Based on the genome comparisons presented here, a viability test of isolates in environments other than those of their origin is suggested as future work. For example, since Thermosynechococcus is lacking genes that are involved in iron regulation under limiting conditions, genus level iron tolerance experiments are proposed to test if strains from low-iron environments can also withstand elevated iron without any adaptations in their genome. In a similar way thermotolerance of these organisms could also be investigated. This could help us understand if Thermosynechococcus is indeed less restrictive or to identify mechanisms unresolvable by the CLOG approach that account for the geographical distribution.
Traits which are unresolvable with an orthology comparison approach like that employed here include amino acid substitutions or sequence changes that lead to variations in enzyme activity, structure, and regulation. Furthermore our analysis is not sensitive to differences in gene regulation or gene copy number. Access to solvents, solubility, active side peptides and differential folding can influence a protein’s functionality and therefor may have implications for adaptability or infer evolutionary history (Lesk and Chothia, 1986; Pandey and Braun, 2020). If the orthology approach employed here is correct, the geochemical parameters that are often thought to be important in microbial selection might simply be not as important as we assume them to be in the environment. Here, orthology between genes was assessed based on sequence comparison (proteinortho, BLAST) and sequence and structure comparison (FeGenie, HMMer). However, there exists a diverse set of methods and approaches to determine orthology between genes (Lechner et al., 2011; Forslund et al., 2018), and as orthology definition techniques change, our results could be interpreted.
From an Earth history perspective, the presence of cyanobacteria in high-iron environments today could indicate that ancient cyanobacteria may not have been limited by elevated iron concentrations on early Earth, whereas previous studies have reported on the toxicity of high iron concentrations to cyanobacteria (Swanner, Mloszewska, et al., 2015; Swanner, Wu, et al., 2015). We could not find striking adaptive mechanisms on a genome level that can explain the tolerance of strains found in environments with elevated iron concentrations, however the presence and perseverance of cyanobacteria in these environments have important implications when considering the earliest organisms capable of oxygenic photosynthesis on early Earth. Ionescu et al. proposed that simply by increasing photosynthetic rate and oxygen production, cyanobacteria might protect themselves from ferrous iron by promoting its precipitation at some distance from the cell (Ionescu et al., 2014). In line with this observation, is worthwhile to note that the biomass accumulation in high iron environments like Jinata hot spring is appreciable, with co-occurring visible biomass and super saturated dissolved oxygen concentrations from cyanobacterial activity (Ward et al., 2019).
Data availability for newly described strains
The Whole Genome Shotgun project for OHK43 has been deposited at DDBJ/ENA/GenBank under the accession JACOMP000000000. The version described in this paper is version JACOMP010000000.
Supplemental Material
SI Figure 1 – class and genus level comparison of core- and pan-genomes; distribution of number of CLOGs with number of genomes at class and genus level. Core TS corresponds to the CLOGs found in all Thermosynechococcus genus members, and shared TS corresponds to CLOGs that are shared by at least 2 and at most 6 genus level genomes.
SI Table 1 – Thermosynechococcus core CLOGs at class level and their hits in the BLAST protein database.
SI Table 2 – Jinata and OHK specific CLOGs at class level with coverage and identity scores, 5 TS specific CLOGs at class level
SI Table 3 – genome information for all family level genomes.
Acknowledgements
LMW was supported by a Simons Foundation Postdoctoral Fellowship in Marine Microbial Ecology. Metagenomic sequencing of OHK was supported by NSF grant #OISE 1639454. SEM acknowledges support from the Astrobiology Center Program of the National Institutes of Natural Sciences (grant no. AB311013). We would also like to thank the group around M. Daroch for providing us with an updated version for the T. elongatus PKUAC-SCTE542 genome which increased the strain’s coherence compared to previous versions of the genome.
Footnotes
↵* prondzinsky{at}elsi.jp / mcglynn{at}elsi.jp