Abstract
The evolution and spread of antibiotic resistance is often mediated by mobile geneticelements. Integrative and conjugative elements (ICEs) are the most abundant conjugativeelements among prokaryotes. However, the contribution of ICEs to horizontal gene transferof antibiotic resistance has been largely unexplored. Here we report that ICEs belonging tomating-pair formation (MPF) classes G and T are highly prevalent among the opportunisticpathogen Pseudomonas aeruginosa, contributing to the spread of carbapenemase-encodinggenes (CEGs). Most CEGs of the MPFG class were encoded within class I integrons, which co-harbour genes conferring resistance to other antibiotics. The majority of the integrons werelocated within Tn3-like and composite transposons. A conserved attachment site could bepredicted for the MPFGclass ICEs. MPFTclass ICEs carried the CEGs within compositetransposons which were not associated with integrons. The data presented here provides aglobal snapshot of the different CEG-harbouring ICEs and sheds light on the underappreciatedcontribution of these elements for the evolution and dissemination of antibiotic resistanceon P. aeruginosa.
Introduction
Among the non-fermenting Gram-negative bacteria, the Pseudomonas genus is the one withthe highest number of species [1, 2]. Pseudomonas aeruginosa, an opportunistic humanpathogen associated with an ever-widening array of life-threatening acute and chronicinfections, is the most clinically relevant species within this genus [3–5]. P. aeruginosa is oneof the CDC “ESKAPE” pathogens – Enterococcus faecium, Staphylococcus aureus, Klebsiellapneumoniae, Acinetobacter baumannii, P. aeruginosa and Enterobacter species –,emphasizing its impact on hospital infections and the ability of this microorganism to “escape”the activity of antibacterial drugs [6]. P. aeruginosa can develop resistance to a wide range ofantibiotics due to a combination of intrinsic, adaptive, and acquired resistance mechanisms,such as the reduction of its outer membrane permeability, over-expression of constitutive orinducible efflux pumps, overproduction of AmpC cephalosporinase, and the acquisition ofantibiotic resistance genes (ARGs) through horizontal gene transfer (HGT) [4, 7, 8]. P.aeruginosa has a non-clonal population structure, punctuated by specific sequence types(STs) that are globally disseminated and frequently linked to the dissemination of ARGs [4, 9]. These STs have been designated as high-risk clones, of which major examples are ST111,61 ST175, ST235 and ST244.
Due to its high importance for human medicine, carbapenems are considered by the WorldHealth Organization (WHO) as Critically-Important Antimicrobials that should be reserved forthe treatment of human infections caused by MDR Gram-negative bacteria [10], such as P.aeruginosa. Carbapenem-resistant P. aeruginosa is in the “critical” category of the WHO’spriority list of bacterial pathogens for which research and development into new antibioticsis urgently required [11]. Besides P. aeruginosa, carbapenem resistance has been reported inother Pseudomonas spp. and is often mediated by the acquisition of carbapenemase-encoding genes (CEGs) [12–14]. Carbapenemases are able to hydrolyse carbapenems andconfer resistance to virtually all ß-lactam antibiotics [15]. When it comes to the Pseudomonasgenus, CEGs are mostly present on class I integrons within the chromosome [4]. Class Iintegrons are genetic elements that carry ARGs and an integrase gene, which controlsintegration and excision of genes [16–18]. Mobile genetic elements (MGEs) such astransposons, plasmids and integrative and conjugative elements (ICEs), are responsible forthe spread of ARGs [19–23]. The phage-inducible chromosomal islands are a recently reportedfamily of MGEs, but unrelated to the carriage of ARGs [24].
Usually, the genes acquired by HGT are integrated in common hotspots in the host’schromosome, comprising a cluster of genes designated by genomic islands (GIs) [19, 25, 26]. This broad definition may also encompass other MGEs, such as ICEs and prophages. Althoughthe exact origin of these elements remains unknown, a growing body of evidence shows thatphages are one of the likely major ancestors of ICEs [27] [28]. ICEs are self-transmissiblemosaic and modular MGEs that combine features of transposons and phages (ICEs canintegrate into and excise from the chromosome), and plasmids (ICEs can also exist as circularextrachromosomal elements, replicate autonomously and be transferred by conjugation) [21,25, 29–31]. Integrative and mobilizable elements (IMEs) encode their own integration andexcision systems, but take advantage of the conjugation machinery of co-resident conjugativeelements to be successfully transferred [32]. ICEs usually replicate as part of the host genomeand are vertically inherited, remaining quiescent, and with most mobility genes repressed [33,34]. These elements also encode recombinases related to those in phages and othertransposable elements. Conjugation involves three mandatory components: a relaxase(MOB), a T4SS and a type-IV coupling protein (T4CP) [35, 36]. Four mating-pair formation(MPF) classes cover the T4SS among Proteobacteria: MPFT, MPFG, MPFF and MPFI [37]. Thefirst is widely disseminated among conjugative plasmids and ICEs, while MPFF is moreprevalent in plasmids of γ-Proteobacteria and MPFG is found essentially on ICEs. MPFI is rarelyidentified. Guglielmini et al. constructed a phylogenetic tree of VirB4, a highly conservedATPase from the T4SS apparatus of different conjugative plasmids and ICEs, and formulatedthe hypothesis of interchangeable conjugation modules along their evolutionary history [38].A close interplay between these elements in the ancient clades of the phylogenetic tree wasobserved, suggesting that plasmids may behave like ICEs and vice-versa, reinforcing thecommon assumption that the line separating ICEs and conjugative plasmids is blurring [30,39]. These authors also searched more than 1000 genomes and found that ICEs are presentin most bacterial clades and are more prevalent than conjugative plasmids [38]. It was alsoobserved that the larger the genome, the higher the likelihood to harbour a conjugativeelement at a given moment, which supports the common assumption that bacteria with largegenomes are more prone to acquire genes by HGT [40, 41].
Delimiting ICEs in genomic data remains particularly challenging [26]. Some signaturesfeatures are frequently observed, such as a sporadic distribution, sequence composition bias,insertion next to or within a tRNA gene, bordering attachment (att) sites and over-representation of mobility genes of the type-IV secretion system (T4SS). However, some ICEspresent atypical features and may not be detected by these approaches [26, 40]. In P.aeruginosa, most ICEs fall into three large families: the ICEclc, pKLC102 and Tn4371. ThePAGI2(C), PAGI3(SG), PAGI-13, PAGI-15 and PAGI-16 were previously described as membersof the ICEclc family, while the PAPI-1, PAPI-2, PAGI-4 and PAGI-5 were linked to the pKLC102family [19]. The ICETn4371 family also represents a large group of ICEs with a common backboneand which are widely distributed, such as in P. aeruginosa UCBPP-PA14, PA7 and PACS171bstrains [21]. These ICEs have been frequently implicated in virulence [42, 43].
Previous reports characterized the complete nucleotide sequence of extra-chromosomalgenetic elements housing different CEGs in pseudomonads [20, 44–47]; however, theassociation of CEGs with chromosome-located MGEs has rarely been investigated [48–50]. Taking into consideration that i) in pseudomonads, CEGs are frequently located within thechromosome, ii) ICEs are the most abundant conjugative elements in prokaryotes and iii) ICEsare more frequently identified in large bacterial genomes, such as in pseudomonads, wehypothesize that ICEs may play a key role in the horizontal spread of CEGs. To investigate thishypothesis, we developed an in silico approach to explore the association between ICEs andCEGs in pseudomonads.
Methods
Carbapenemases database
Antimicrobial resistance translated sequences were retrieved from the BacterialAntimicrobial Resistance Reference Gene Database available on NCBI(ftp://ftp.ncbi.nlm.nih.gov/pathogen/Antimicrobial_resistance/AMRFinder/data/2018-04-16.1/). The resulting 4250 proteins were narrowed down to 695 different carbapenemases tocreate a binary DIAMOND (v. 0.9.21, https://github.com/bbuchfink/diamond) database [51]. Only the sequences presenting ‘carbapenem-hydrolyzing’ or ‘metallo-beta-lactamase’ onfasta-headers were used to build this local database.
Genome collection and blast search
A total of 4565 Pseudomonas genomes was downloaded from NCBI (accessed on the 24th ofApril, 2018). These genomes were blasted against the local carbapenemase database usingthe following command: ‘diamond blastx –d DB.dmnd –o hits.txt --id 100 --subject-cover 100-f 6 --sensitive’.
Bioinformatic prediction of ICEs and genetic environment analyses
The CEG-harbouring Pseudomonas genomes were annotated through Prokka v. 1.12(https://github.com/tseemann/prokka) [52]. The translated coding sequences were analysedin TXSScan/CONJscan platform to inspect the presence of ICEs(https://galaxy.pasteur.fr/root?tool_id=toolshed.pasteur.fr%2Frepos%2Fodoppelt%2Fconjscan%2FConjScan%2F1.0.2) [37]. All ICEs harbouring CEGs predicted by TXSScan/CONJscanwere inspected for direct repeats that define the boundaries of the element. The completenucleotide sequence in Genbank format of corresponding records was imported intoGeneious v. 9.1.8 to help delimiting genomic regions flanking the ICEs [53]. Complete ICEsequences were aligned with EasyFig v. 2.2.2 (http://mjsull.github.io/Easyfig/files.html) [54]. Screening of complete ICEs for ARGs was achieved by ABRicate v. 0.8(https://github.com/tseemann/abricate). Phage and insertion sequences were inspectedthrough PHASTER (http://phaster.ca/) and ISfinder (https://www-is.biotoul.fr/), respectively[55, 56]. Multiple Antibiotic Resistance Annotator (MARA, http://mara.spokade.com) wasused to explore the genetic background of the CEGs [57]. Orthologous assignment andfunctional annotation of integrase sequences was achieved through EggNOG v. 4.5.1(http://eggnogdb.embl.de/#/app/home) and InterProScan 5(https://www.ebi.ac.uk/interpro/search/sequence-search) [58, 59].
Phylogenomics
All CEG-harbouring P. aeruginosa genomes were mapped against the P. aeruginosa PAO1reference strain (accession number NC_002516.2), to infer a phylogeny based on theconcatenated alignment of high quality single nucleotide polymorphisms (SNP) using CSIPhylogeny and standard settings [60]. The phylogenetic tree was plotted using the iTOLplatform (https://itol.embl.de/).
MLST and taxonomic assignment of unidentified species
To predict the sequence type (ST) of the strains harbouring ICEs, the P. aeruginosa MLSTwebsite (https://pubmlst.org/paeruginosa/) developed by Keith Jolley and hosted at theUniversity of Oxford was used [61]. Taxonomic assignment of unidentified species carryingICEs was achieved by JSpeciesWS v. 3.0.17 (http://jspecies.ribohost.com/jspeciesws/#home)174 [62].
Results
A plethora of carbapenemase-encoding genes was identified in a subset of Pseudomonasspecies
From the total Pseudomonas genomes analysed (n=4565), 313 CEGs were identified in 297genomes (Figure 1 and Table S1). As expected, blaVIM-2 represents the majority of the CEGsfound among Pseudomonas spp., being detected mainly in P. aeruginosa, followed by P.plecoglocissida, P. guariconensis, P. putida, P. stutzeri and 16 genomes corresponding tounidentified species (Table S1). Curiously, some strains presented two CEGs, either presentinga duplication of the same gene, such as blaIMP-34 from NCGM 1900 and NCGM 1984 Japaneseisolates, or harbouring different CEGs, such as blaIMP-1 and blaDIM-1 in isolates 97, 130 and 142recovered in Ghana (Table S1, highlighted in red). A wide variety of STs was also observed,including the high-risk clones ST111, ST175 and ST244.
Detection of ICEs encoding carbapenemases in Pseudomonas spp
65.5% (205/313, Table S1) of the CEG hits are located within small contigs, with a sequencesmaller than 20kb in length. The presence of repeated regions, such as those encoding fortransposases, tend to split the genome when second-generation sequencing approaches areused. Based on information retrieved from NCBI (accessed on the 24th of May, 2018), the totalnumber of bacterial genomes sequenced at the chromosome/complete genome level is12,077, while the number of genomes sequenced at the scaffold/contig is much larger(127,231). With this sequencing limitation, we were still able to identify 49 ICEs associatedwith CEGs (n=20 with complete sequence) among all pseudomonads genomes (Table 1, TableS1 and Figure 1). When an ICE location was attributed to a CEG located on a small contig, theassumption was based on previously published data, as pointed out on Table 1. Besides theaforementioned ICEs, we also identified a putative MGE within Pseudomonas sp. NBRC111143 strain (Table S1). The T4CP-encoding gene was absent from this blaIMP-10-carryingelement, which could be due to contig fragmentation or gene absence. In case the gene isactually missing, this element could still be mobilized by the conjugation machinery of an ICEor conjugative plasmid(s) present in the host, and should be classified as an IME.
The ICEs identified here were all integrated within P. aeruginosa genomes (with the exceptionof the one element identified in Pseudomonas sp. PONIH3 genome) and AT-rich whencompared to their host’s chromosome; the mean GC value for this species is 66.2% accordingto EZBioCloud (https://www.ezbiocloud.net/taxon?tn=Pseudomonas%20aeruginosa) (Table 1).
All ICEs identified here possessed only one tyrosine integrase (Figure 2). ICEs belonging to the ICEclc family (MPFG class) carried an integrase belonging to the bacteriophage P4-like family, while ICEs belonging to the ICETn4371 family (MPFT class) carried an integrase belonging to shufflon-specific DNA recombinase Rci and Bacteriophage Hp1-like family (Table 1).31 Rci and Hp1-like were only distantly related (13% amino acid identity) to P4-like integrases. Orthologous assignment of these integrases revealed that the former and the later integrases identified were present in more than 100 and 400 proteobacteria species, respectively. While P4-like integrases were more prevalent on γ-proteobacteria, half of the strains carrying Rci and Hp1-like integrases belong to the α-proteobacteria.
We observed that MPFG class ICEs tend to integrate next to a single copy of tRNAGly or a cluster of two tRNAGlu and one tRNAGly genes, which is in agreement with previous findings [26, 40]. A conserved 8-bp att site (5′-CCGCTCCA) flanked all complete ICEs of the MPFG class identified here (Table 1). Notably, most ICEs of this class were adjacent to phages (either at the 5’-or the 3’-end) targeting the same att site as the neighbour ICE. No att site could be identified for the integration of MPFT class ICEs. A gene encoding for a catechol 1,2-dioxygenase and a gene encoding for a protein with no described conserved domain were found flanking the blaSPM-1-harbouring ICEs. Regarding the elements carrying blaNDM-1, a gene encoding for a different protein also with no conserved domain identified and a gene encoding for the type III secretion system adenylate cyclase effector ExoY were separated upon insertion of these ICEs. Integration next to hypothetical proteins or tRNA genes was commonly observed.
Carbapenemases are frequently encoded within transposons
CEGs were associated with class I integrons frequently co-harbouring aminoglycoside resistance genes when associated with MPF class G ICEs (Table 1). Class I integrons were often associated with a wide array of transposons, such as the Tn3 superfamily transposons and the IS6100 composite elements (Table 1). MPFT class ICEs were targeted by more complex elements, such as the composite transposons carrying blaSPM-1 and blaNDM-1 (Table 1). The blaNDM-1 gene was identified in Singapore in ICETn43716385 and associated with ST308, as recently reported [63]. The blaNDM-1 was flanked by two ISCR24-like transposases. blaSPM-1 waslinked to ICETn43716061, a recently described ICE [64]. Again, the CEG was located within an ISCR4-like composite transposon. ISCR elements are atypical elements of the IS91 family which represent a well-recognized system of gene capture and mobilization by a rolling-circle transposition process [21, 70].
Besides previously described blaNDM-1 and blaSPM-1 harbouring ICEs, we characterize here new ICE elements of MPFG and MPFT classes (Table 1 and Figure 3). The blaDIM-1-harbouring ICE from IOMTU 133 strain was integrated between the 3’-end of a tRNAGly gene (IOMTU133_RS11660) and a gene encoding for the R body protein RebB (IOMTU133_RS12085). blaDIM-1 was first described as a single gene cassette located within a class I integron associated with a 70-kb Pseudomonas stutzeri plasmid recovered in the Netherlands [13]. However, the integron carrying blaDIM-1 in strain IOMTU 133 was unrelated to the one from the P. stutzeri plasmid, harbouring genes encoding for aminoglycoside (aacA4-C329 and rmtf), trimethoprim (dfrB5) and chloramphenicol (catB12) resistance (Figure 3A). Direct repeats (DRs) were found flanking the entire IS6100 composite transposon (5’-TTCGAGTC), indicating the transposition of this element into the ICE element. Besides being identified as a composite transposon, IS6100 was frequently observed as a single copy at the 3’end of the class I integron (Figures 3B and 3C), suggesting that these elements were derived from the In4 lineage [71]. The blaIMP-1 from the NCGM257 strain identified in Japan belonged to a different ST (ST357) than the frequently identified ST235 associated with the spread of this CEG in this country [72]. The CEG was also shown to be associated with a novel complex class I integron, co-harbouring aadB, cmlA9 and tet(G) genes encoding resistance to aminoglycosides, chloramphenicol and tetracyclines, respectively (Figure 3B). This integron was inserted (DRs 5’-GAGTC) within a mercury resistance transposon. This genetic organization was frequently recovered among other ICE-harbouring strains, such as the onesassociated with blaGES-5, blaIMP-13 and blaIMP-14 (Table 1). The entire ICE was integrated into the chromosome of NCGM257 strain between the 3’-end of a tRNAGly gene (PA257_RS24790) and the aforementioned Pseudomonas phage Pf1-like element. The new ICE identified on the P1_London_28_IMP_1_04_05 strain presented blaIMP-1 in a different In4-like integron than that observed for the NCGM257 strain, even though both elements were associated with a Tn3-like transposon (Figure 3C). Unlike most ICEs of the MPFG class, its integration occurred between a gene encoding for a LysR family transcriptional regulator (AFJ02_RS19410) and a gene encoding for a hypothetical protein (AFJ02_RS19770). Regarding the blaKPC-2-harbouring Pseudomonas sp. PONHI3 strain, a tetra correlation search revealed that this strain was highly similar (Z-score above the 0.999 cut-off) to Pseudomonas mosselii SJ10 (accession number NZ_CP009365.1). Average nucleotide identity based on BLAST (ANIb) analysis of these genomes revealed that both strains belong to the same species, since the ANIb value was above the 95% cut-off for species delineation [73]. However, the ANIb value for both strains was below the cut-off when compared with the P. mosselii DSM 17497 type train (accession number NZ_JHYW00000000.1), suggesting that both strains may comprise novel species within the Pseudomonas putida phylogenetic group [2]. The PONHI3 strain carried a double copy of blaKPC-2 within an ICE from MPFT class. A complex genetic environment was found surrounding these genes (Figure 3D). This ICE was integrated between a gene encoding for a biopolymer transport protein ExbD/TolR (C3F42_RS18665) and a gene encoding for an alpha/beta hydrolase (C3F42_RS18995).
An atypical GI encoding carbapenemases
Besides ICEs, we also identified an atypical 19.8-kb long GI harbouring blaVIM-2 in P. aeruginosa AZPAE13853 and AZPAE13858 strains from India (Figure S1). A similar element was also observed in P. aeruginosa BTP038 strain from the USA, with the exception that the Tn402-like transposon harbouring blaVIM-2 was orientated in an inverted position. Five base-pair DRs (5’-CTCTG in AZPAE13853 and AZPAE13858 and 5’-CTGAG in BTP038 strains) were found flanking this transposon structure. Importantly, in these strains the GIs were flanked by identical signal recognition particle RNAs (srpRNAs), indicating a strong site preference for these elements.
Discussion
Our results show that blaVIM and blaIMP are widely disseminated, both geographically and phylogenetically (across Pseudomonas spp.). Moreover, and as previously described, blaVIM-2 was the most frequently reported CEG (Figure 1 and Table S1) [4]. On the other hand, blaSPM-1 is still restricted to P. aeruginosa and Brazil (or patients who had been previously hospitalized in Brazil) [65]. Even though ST235 has been frequently linked to the dissemination of ARGs, no CEG-harbouring strains belonging to this ST were identified. Curiously, some strains (highlighted on Figure 1) belong to a double ST profile(ST235/ST2613), since the strains carry a double copy with different allele sequences of the house-keeping gene acsA, encoding for an acetyl-coenzyme A synthetase. These genes only display 80,3% nucleotide identity. We plan to conduct comparative genomic studies to explore the idiosyncrasies of these double ST profile strains.
Not all CEGs are likely to be geographically and phylogenetically disseminated, but those that are more promiscuous present a serious threat. The geographical distribution of the high-risk clones and the diversity of CEGs propose that the spread of these STs is global and the acquisition of the resistance genes is mainly local [4, 68]. Previous studies suggest that environmental species may pose an important reservoir for the dissemination of clinically relevant carbapenemases, which are vertically amplified upon transfer to P. aeruginosa high-risk clones [12, 14]. The high prevalence of these elements among high-risk clones may be partially explained by the genetic capitalism theory, given that a widely disseminated ST should have a greater probability of acquiring new CEGs and to be further selected and amplified due to the high antibiotic pressure in the hospital environment [74]. Other theories support that the high-risk clones have a naturally increased ability to acquire foreign DNA, since these STs appear to have lost the CRISPR (clustered regularly interspaced short palindromic repeats)-Cas (CRISPR associated proteins) system, which act as an adaptive immune system in prokaryotic cells and protects them from invasion by bacteriophages and plasmids [75–77].
This study underestimates the extent of host range because only ICEs in sequenced genomes were detected. Also, identification of new ICEs could only be achieved in complete genomes or contigs with a sequence length large enough to include the full (nor near complete) sequence of the ICE. As so, it is important to highlight the need to perform third generationsequencing on CEG-harbouring genomes to avoid fragmentation of the genetic environment surrounding the gene and to provide a wider view of complete supporting ICEs and other MGEs. All ICE elements here identified fulfilled the criteria to be considered conjugative as proposed by Cury et al.: a relaxase, a VirB4/TraU, a T4CP and minimum set of MPF type-specific genes [40]. ICEs tend to integrate within the host’s chromosome by the action of a tyrosine recombinase, even though some elements may use serine or DDE recombinases instead [30]. Though rare, some elements encode for more than one integrase, most likely resulting from independent integration of different MGEs [40]. Conserved sites are hotspots for ICE integration due to their high conservation among closely related bacteria, and so expanding the host range and be stably maintained after conjugative transfer [78, 79]. ICEs were often integrated next to phages highly similar to the Pseudomonas phage Pf1 (NC_001331.1), a class II filamentous bacteriophage belonging to the Inoviridae family [77]. Pf1-like phages are widely disseminated among P. aeruginosa strains and may have a role in bacterial evolution and virulence [80–82]. Interestingly, no representative of the pKLC102 family was linked to the dissemination of CEGs. This may be explained due to a higher affinity of the transposons carrying the CEGs for hotspots located within representatives of the other two families.
MGEs specifically targeting conserved regions of the genome such as tRNAs are common and this specificity represents an evolutionary strategy whereby the target site of an element is almost guaranteed to be present, due to its essentiality, and very unlikely to change due to biochemical constraints of the gene product. We think a similar situation exists for the elements found between the small srpRNAs described on the atypical GI element here identified and is in contrast to the more permissive nature of target site selection shown for example, by elements of the Tn916/Tn1545 family [83].
Here, we revealed that different Tn3-like and composite transposons harbouring a wide array of CEGs were transposed into MPF G and T ICE classes, which were most likely responsible for the dissemination of these genes through HGT and/or clonal expansion of successful Pseudomonas clones. This study sheds light on the underappreciated contribution of ICEs for the spread of CEGs among pseudomonads (and potentially further afield). With the ever-growing number of third-generation sequenced genomes and the development of more sophisticated bioinformatics, the real contribution of these ICEs will likely rapidly emerge.
Recently, it was shown that interfering with the transposase-DNA complex architecture of a Tn916-like conjugative transposon (also known as ICE) lead to transposition inhibition to a new host [84]. In the future, it would be interesting to determine if the same mechanism is observed for tyrosine recombinases present in ICEclc and Tn4371 derivatives, as well as in other MPF ICE classes, as a potential approach to interfere with the spread of antimicrobial resistance.
Author contributions
JB, APR, FG and LP designed the study; JB and RLS performed the in silico analysis; JB wrote the manuscript. All the authors approved the final manuscript.
Competing interests
None to declare.
Acknowledgments
This study received financial support from the European Union (FEDER funds POCI/01/0145/FEDER/007728) and National Funds (FCT/MEC, Fundação para a Ciência e Tecnologia and Ministério da Educação e Ciência) under the Partnership Agreement PT2020 UID/MULTI/04378/2013. J. B. and F. G. were supported by grants from Fundação para a Ciência e a Tecnologia (SFRH/BD/104095/2014 and SFRH/BPD/95556/2013, respectively). We thank Álvaro San Millan for helpful discussions. We also thank Benjamin Buchfink (DIAMOND), Jean Cury (TXSScan/CONJscan) and Sally Partridge (MARA) for their valuable assistance.
References
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.
- 46.
- 47.↵
- 48.↵
- 49.
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.
- 67.
- 68.↵
- 69.
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.
- 82.↵
- 83.↵
- 84.↵