Functional evolution of Nodulin26-like Intrinsic Proteins: From bacterial arsenic detoxification to plant nutrient transport.

Nodulin26-like-intrinsic-proteins (NIPs) play essential roles in transporting the nutrients, silicon and boron, in seed plants, but the evolutionary origin of this transport function and the co-permeability to toxic arsenic remains enigmatic. Horizontal gene transfer of a yet uncharacterized bacterial AqpN-aquaporin group was the starting-point for plant NIP evolution. We combined intense sequence-, phylogenetic and genetic context analyses and a mutational approach with various transport assays in oocytes and plants to resolve the trans-organismal and functional evolution of bacterial and algal and terrestrial plant NIPs and to reveal their molecular transport specificity features. We discovered that aqpN genes are prevalently located in arsenic-resistance-operons of various prokaryotic phyla. We provided genetic and functional evidence that these proteins contribute to the arsenic detoxification machinery. We identified NIPs with the ancestral bacterial AqpN selectivity filter composition in algae, liverworts, moss, hornworts and ferns and demonstrated that these archetype plant NIPs and their prokaryotic progenitors are almost impermeable to water and silicon but transport arsenic and boron. With a mutational approach, we demonstrated that during evolution, ancestral NIP selectivity shifted to allow subfunctionalizations. Together, our data provided evidence that evolution converted bacterial arsenic efflux channels into essential seed plant nutrient transporters.


Introduction
The metalloids boron (B) and silicon (Si) are fundamental for the development of vascular plants not only because they ensure proper differentiation, structural support and elasticity of plant cell walls, but also because they contribute to pathogen defence and general stress tolerance (Ma et al., 2006;Bienert & Chaumont, 2011). Consequently, corresponding transport mechanisms for these elements are essential for sufficient plant B and Si nutrition. Nodulin 26-like intrinsic proteins (NIPs) mediate B and Si transport and thereby sustain growth, fertility and yield of terrestrial plants. NIPs belong to the major intrinsic protein (MIP) superfamily (also termed as aquaporins, AQPs), which comprises channels for the diffusion of small neutral and mostly polar molecules across various biological membranes in all kingdoms of life. MIP channels possess six transmembrane helices connected by five loops and the two termini facing the cytoplasm. The so-called aromatic/arginine (ar/R) constriction region acts as an MIP selectivity filter (SF) and comprises four amino acids of transmembrane helices 2 and 5 (positions R1 and R2) and loop E (positions R3 and R4) representing the narrowest part of the channel pore and forming a size exclusion barrier conferring selectivity to particular substrates (Murata et al., 2000).
Sequences for NIP subfamily members have not been found in animal and fungal genomes and it has, therefore, been hypothesised that the first plant NIP was the result of a horizontal aqpN gene transfer (HGT) event during the origin of plant life forms (Finn & Cerd a, 2015;Zardoya et al., 2002;Danielson & Johanson, 2010;Abascal et al., 2014). However, the exact origin of the plant NIP subfamily and the evolution and diversification of functions within early plant lineages are completely unclear and unresolved.
NIPs form one of the isoform-richest subfamily in plants with a high diversity regarding their amino acid sequences and substrate specificities. Besides their vast substrate spectrum (Bienert & Chaumont, 2011;Bienert & Bienert, 2017), the physiological relevance of NIPs in planta may be restricted to the transport of the essential and beneficial metalloids B and Si, but also to the extrusion of the toxic minerals arsenic (As), antimony (Sb), and germanium (Ge) (Pommerrenig et al., 2015;Bienert & Bienert, 2017). NIPs subdivide into three groups termed NIP-I, NIP-II and NIP-III (Wallace & Roberts, 2004;Mitani et al., 2008). This grouping is based on common substrate selectivities, sequence similarities and an amino acid composition consistency of their ar/R SF. The NIP-I and NIP-II subgroups are present in all higher plants, while NIP-III isoforms are largely, but not exclusively, confined to Liliopsida species (Ma & Yamaji, 2015;Trembath-Reichert et al., 2015;Deshmukh & B elanger, 2016).
Multiple NIP isoforms belonging to the NIP-II group (such as AtNIP5;1, AtNIP6;1, ZmNIP3;1, and OsNIP3;1) have been unambiguously demonstrated to be essential for the root uptake and translocation of B in various monocot and dicot plant species and therefore for B nutrition in seed plants (reviewed in Pommerrenig et al., 2015;Bienert & Bienert, 2017). NIP-III group isoforms are mainly found in plant species that are biosilicifying or which benefit from high Si concentrations in their plant body such as Poaceae or Cucurbitaceae species. Accordingly, NIP-III group members (such as OsNIP2;1, OsNIP2;2, HvNIP2;1, HvNIP2;2 and CmNIP2;1) were shown to be crucial for an efficient root uptake and translocation of Si and therefore for Si homeostasis in such plant species (reviewed in Ma & Yamaji, 2015;Bienert & Bienert, 2017).
NIPs differing from these three subgroups have been identified in the moss Physcomitrella patens and the lycophyte Selaginella moellendorffii (Danielson & Johanson, 2008;Anderberg et al., 2012). Those NIPs form a fourth NIP group (NIP-IV) and have a SF composition consisting of F R1 -A R2 -A R3 -R R4 residues. As this SF layout is restricted to bacterial AqpNs and a few NIP isoforms found in extant members of early land plant lineages, it might represent the archetype SF layout of all plant NIPs and the original NIP channel selectivity (Finn & Cerd a, 2015;Danielson & Urbanson, 2010;Abascal et al., 2014;Trembath-Reichert et al., 2015). It is unlikely that boric acid and silicic acid transport, the essential functions of modern seed plant NIP-IIs and NIP-IIIs, respectively, represent the ancestral function of bacterial AqpN proteins as neither B nor Si essentiality is common in bacteria. Initial analyses of seed plant NIPs identified glycerol as the first common substrate and suggested a role for bacterial AqpNs and NIPs in early nonvascular terrestrial plants in glycerol uptake and distribution (Zardoya et al., 2016;Roberts & Routray, 2017). However, the exact origin of plant NIPs and the evolution and diversification of functions within early plant lineages are completely unclear and neither bacterial AqpNs nor plant NIPs with a F-A-A-R-type SF composition have been characterised so far.
The present work sheds light into the functional evolution of NIPs. We combined intense sequence-and genetic context analyses with transport assays to resolve the functional evolution of NIPs within plants and reveal molecular features of bacterial AqpNs and algal and terrestrial plant F-A-A-R-type NIP channels. AqpN channels are permeable to arsenous acid and are frequently part of arsenic (As) resistance operons (ars operons) of diverse bacterial phyla. This suggests that an intrinsic function of the ancestors of plant NIPs namely the bacterial AqpNs resides in As detoxification processes and that modern NIP channels underwent subfunctionalisation and neofunctionalisation and turned from As effluxers into essential and beneficial plant nutrient channels. Permeability of ancestral plant NIPs was a premise for the ability of vascular plants to efficiently take up and translocate B and Si into and within the plant body. This transport ability first allowed plants to use these elements efficiently, a crucial prerequisite for terrestrial upright growth and stress tolerance.

Arabidopsis
Arabidopsis thaliana Col-0 wild-type (NASC: N60000) and Atnip5;1 knock-down mutant (SALK_122287; NASC: N622287) seeds were obtained from The European Arabidopsis Stock Centre. For in vitro culture, seeds were sterilised with 70% ethanol plus 0.05% Triton X-100 followed by three washes of 99% ethanol. Seeds were plated on half-strength Murashige and Skoog (½MS) medium (2.2 g l À1 MS salts with minimal organics (Sigma), 1% sucrose, 0.7% agar, pH 5.8 (KOH)) with or without antibiotics and vernalised for 2 d at 4°C. In vitro cultures were grown in a photoperiod of 10 h : 14 h, 22°C : 19°C, light : dark (120 lmol m À2 s À1 ). For soil growth experiments, seedlings were transferred into pots after 2.5 wk of in vitro growth and grown in a phytochamber at 18°C and a 17 h light period for another 2.5 wk.

Physcomitrella
Physcomitrella patens (Gransden, IMSC no. 40001) was obtained from the International Moss Stock Center and grown at 22°C under a 14 h : 10 h, light : dark cycle and a photosynthetic photon flux density of 90 µmol m À2 s À1 . Moss cultures were cultivated on a modified Knop medium (Reski & Abel, 1985). Final nutrient concentrations were: KH 2 PO 4 (250 mg l À1 ), KCl (250 mg l À1 ), MgSO 4 (250 mg l À1 ), Ca(NO 3 ) 2 Á 4H 2 O (1000 mg l À1 ), FeSO 4 Á 7H 2 O (12.5 mg l À1 ). 1% Phytagel was used as solidifying agent. Before sterilisation of the medium, the pH was adjusted to 5.8 with KOH. To remove traces of B, the B chelating agent Amberlite IRA-743 (Sigma) was used. Amberlite IRA-743 was washed three times with MilliQ water and incubated together with the Knop medium at a concentration of

Sequence retrieval, gene context analysis and phylogenetic analysis
Multiple public-accessible gene, operon and genome databases have been used for sequence retrievals and gene context analyses. Bayesian phylogenetic analyses and tree computation have been performed with curated protein alignments. Detailed information on these procedures is provided in Supporting Information Methods S1. Sequences which have been used in this study are given in Dataset S1.

Cloning and vector construction
Detailed information about vector constructions and the procedures for molecular cloning techniques is provided in Methods S1.

Subcellular localisation and confocal microscopic imaging
Detailed information on transient transformation protocols and subsequent subcellular localisation analyses of YFP-tagged PpNIP5 proteins using a Zeiss LSM 780 confocal laser scanning microscope can be found in Methods S1.

Oocyte transport assays
In vitro cRNA synthesis, oocyte handling procedures, and various oocyte uptake assays with subsequent determination of permeability coefficients or element levels of oocytes are described in detail in Methods S1.

Complementation analysis of Atnip5;1 T-DNA insertion mutants
Detailed information on the T-DNA insertion line, vector constructions, the procedures for transgenic Arabidopsis generation and selection is provided in Methods S1.

RNA extraction, cDNA synthesis and real-time quantitative PCR
Various experimental information related to the different working steps necessary for a RT-qPCR as well as RT-qPCR-related specification details are described in detail in Methods S1.

Statistical analysis
The two-sample Student's t-test was applied to find out which independent sample sets were significantly different from each other. Two-sided testing was used. In all cases, statistical significance was defined as: *, P < 0.05; **, P < 0.01; or ***, P < 0.001. Tukey's test was used to compare means at a probability level of 5%. In this case, levels of significance are represented by letters as indicated in the captions.

Results
Phylogenetic analysis of NIP-type aquaporins from prokaryota and eukaryota Our phylogenetic analyses using 366 bacterial, archaeal and plant NIP-type MIPs showed that the major prokaryotic MIP groups, which are AQPZs, AQPMs, AQPNs and GLPFs, formed well supported clades (Fig. 1). Six major AQPN clades (I-VI) with a full node support were detected. Two of these AQPN clades do not encode F-A-A-R-type SFs but display a wide variety of SF residue combinations: AQPN-I is mainly composed of archaeal and AQPN-II of bacterial sequences (Fig. 1).
The remaining four AQPN clades are characterised by a predominant occurrence of F-A-A-R-type SFs. All plant NIP-type sequences as well as three bacterial AqpNs formed together the AQPN-III clade (Figs 1, S1). Within this clade, algal NIPs (KnNIP6;1 and KnMIP) can be clearly separated from the NIP-I, NIP-II, NIP-III groups and the F-A-A-Rtype NIPs of land plants (Figs 1, S1). Interestingly, the F-A-A-R-type NIPs of land plants cluster closest with the NIP-III group members, which are crucial for Si transport in seed plants.
AQPN-IV clade is formed by sequences deriving mainly from Firmicutes but also Proteobacteria that are F-A-A-R-type SFdominated (Fig. 1). The AQPN-V clade is exclusively formed by cyanobacterial sequences, which all encode F-A-A-R-type SFs (Fig. 1). The member-rich AQPN-VI clade is formed by sequences from various bacterial phyla either encoding F-A-A-R-type or an W-A-A-R-type SFs (Fig. 1).
Two additional MIP groups, AQPNx and AQPZx cluster apart from other AqpZ and AqpN-like sequences. These sequences derive from various different bacterial and archaeal species, and encode MIPs with very diverse SFs and form a phylogenetically loose cluster. The overall backbone of the phylogenetic tree is unresolved.
Bacterial AqpNs with an F-A-A-R-type selectivity filter are abundantly located in arsenic resistance operons Very few aqpNs have previously been identified in the genomes of few bacterial phyla (Finn & Cerd a, 2015;Danielson & Johanson, 2010;Abascal et al., 2014). To obtain more information on the uncharacterised bacterial AQPN clade, which is of high significance for the evolution of plant nutrient NIP channels, we studied AqpNs sequences in more detail. The phylogenetic analysis revealed that only 57 (56 bacterial and one archaeal) AQPNlike sequences are present in the 4576 prokaryotic genomes deposited in the KEGG database (Fig. 1). All of these AqpNs are classified as 'AQPZs' or 'MIPs' in the database. Therefore, only 1.3% of the prokaryote genomes of the database encode for AqpNs (Fig. 2).
By contrast, 1879 and 2893 sequences were classified as aqpZ and glpF genes, respectively. Identified AqpNs derived from 10 and one bacterial and archaeal phyla, respectively, with the bacteria themselves sharing no distinct lifestyle or habitat. The SFs of AqpNs are variable but many of these are composed of the amino acid residues F R1 -A R2 -A R3 -R R4 ( Fig. 1; Dataset S1).
In order to infer a potential role of AqpNs, we investigated their genetic context providing information on operons and functional units. Remarkably, 19 out of the 57 aqpNs (33.3%) were New Phytologist (2020)

Research
New Phytologist located in arsenic resistance (ars) operons, while the remaining aqpN isoforms do not seem to be associated to any other nonmetalloid regulatory genetic unit (Dataset S1). All except one of the 19 ars operon-encoded AqpNs contained the F-A-A-R SF layout. AqpNs were part of diversely constituted ars operons, consisting of genes encoding at least an ArsR transcription activator, an arsenate reductase (ArsC), and sometimes an arsenite Sadenosylmethyltransferase (ArsM) or an arsenical resistance protein (ArsH). The observation that AqpNs were frequently located in ars operons suggests that they play a physiologically important role in As detoxification as effluxers of As out of bacterial cells.
To further investigate the abundance of aqpNs in ars operons, a protein BLAST search retrieved 36 AqpN sequences that clustered with AqpN isoforms, for which the genetic context could be accessed, and which were not present in the KEGG database ( Fig. 1; Dataset S1). In this case, 50% of the aqpN sequences (18 out of 36) were found in ars operons, and all of them code for channels having the F-A-A-R SF-motif. The finding that aqpNs are present with a high frequency in ars operons was further supported by the reanalysis of 18 MIPs that have been identified in 685 ars operons (Yang et al., 2015). We assigned six AqpNs (all encoding for an F-A-A-R-type SF) and 12 GlpFs (Fig. S2a). Moreover, in an operon database covering 2072 prokaryotic genomes, we identified 31 MIP gene containing ars operons. Amongst them, our analyses identified three aqpNs encoding an F-A-A-R-type SF and 28 glpF genes (Fig. S2b).

Occurrence of F-A-A-R-type NIPs in land plants
The prevalence of F-A-A-R residues constituting the SF in bacterial AqpNs and in early diverging plant NIPs such as S. moellendorffii and P. patens suggests that the last common ancestor to the plant NIPs encoded a F-A-A-R-type SF. To study the frequency and evolution of this SF composition, we screened various plant genome databases for NIPs with this SF composition. We identified a NIP with an F-A-A-R-type SF in the genome of the charophyte Klebsormidium nitens and thereby discovered the presence of NIPs also in green algae (Fig. 1). This finding suggests that the HGT of an aqpN happened before plants invaded the land and associate with the estimation of Zardoya et al. (2002). Additionally, we found such F-A-A-R-type NIPs in liverwort (Marchantia polymorpha), moss (Sphagnum fallax) and fern (Salvinia cucullata and Adiantum capillus-veneris) taxa (Figs 1, 2). These findings closed the knowledge gap on the NIP lineage throughout the phylogeny from bacteria, green plants, over land plants to vascular plants. Seed plants, both gymnosperms and flowering plants, encode various NIPs with SFs typical for the NIP-I, NIP-II and NIP-III group. However, the F-A-A-R SF-group has not been identified in seed plants in any of the analysed genome-rich databases.

Identification of F-A-A-R-type MIPs allowing studying the functional evolution of NIP-type aquaporins
We subjected several MIPs to further analyses to resolve the functional evolution of F-A-A-R-type MIPs from bacteria along the plant lineage. KrAqpN is found in the aerobic, filamentous, nonmotile, Gram-positive bacterium Ktedonobacter racemifer belonging to the Chloroflexi phylum from which an aqpN gene may have been transferred via a HGT to plants (Fig. 1). Moreover, this F-A-A-R SF-encoding aqpN is located in a typical ars operon (Fig. 2). The F-A-A-R SF-carrying KnNIP6;1 is encoded by the filamentous terrestrial alga K. nitens, which belongs to the Charophyta green algae group from which all terrestrial plants, the Embryophyta, emerged (Hori et al., 2014). The name KnNIP6;1 was adopted according to a detailed phylogenetic analysis (U. Johanson, per. commun.: H. I. Anderberg & U. Johanson, unpublished). PpNIP5;1, PpNIP5;2, and PpNIP5;3 are encoded by the moss P. patens and represent members of the F-A-A-R-type NIPs of early land plant lineages.

Bacterial, algal and moss F-A-A-R-type NIPs do not conduct water in Xenopus oocytes
The water channel activity of KrAqpN, KnNIP6;1, PpNIP5;1, PpNIP5;2, and PpNIP5;3 was tested by Fig. 1 Phylogeny of bacterial major intrinsic proteins (MIPs) and plant nodulin 26-like intrinsic proteins (NIPs). Consensus tree derived from archaeal, bacterial, plant and putative metazoan amino acid sequences using Bayesian phylogenetic inference. Numbers beside the nodes indicate the posterior probability values if larger than 0.9. For the presentation of the tree midpoint-rooting was applied. To the left, fully supported clades were collapsed. The size of the triangles corresponds to the number of sequences (given in the triangle if larger than three) comprised in the clade. Prokaryotic GLPF, AQPM, AQPN and AQPZ clades are labelled and the amino acid residue compositions which comprise the aromatic/arginine selectivity filter (SF) in the corresponding clade are given below in the one-letter-code. Six well supported AQPN/NIP clades (I-VI) were formed. These clades are coloured differently AQPN-I: bluearchaeal sequences, AQPN-II: greybacterial sequences, AQPN-III: greenmainly plant and three bacterial sequences, AQPN-IV: orange -Firmicutes and Proteobacteria sequences, AQPN-V: purplecyanobacterial sequences, AQPN-VI: yellowsequences from various different prokaryotic phyla. AQPNx and AQPZx clades (brown triangles) neither clearly belong to the AQPZ nor AQPN clades. A few genes that belong to these clades have been found in arsenic resistance operons (ars operons). SF compositions found in these clades are displayed next to the clade name. Blue boxed SFs occur in ars operons. To the right, a detailed representation of the AQPN/NIP clade is depicted. Instead of the prokaryotic AqpN protein name/identifier, the organism by which the MIP is encoded is displayed. Different font colours indicate different taxonomic groups: brown, archaea; black, bacteria; red, metazoan; green, plants. If multiple individuals for a taxon were included then they are distinguished by number. Different vertical line colours indicate whether the proteins of this AQPN clade are classified as AqpN (black) or plant NIP (green) sequences. For each clade, the amino acid residues that determine the SFs are shown in the one-letter-code (green and black writing indicates NIP and AqpN SF compositions, respectively). Blue boxed SF combinations are found in ars operons. The first digit in parentheses indicates how many genes of the labelled clade with the indicated SF are located in ars operons while the second digit is the total number of genes with the indicated SF composition within the labelled AQPN clade. AQPN-III clade sequences that have been functionally analysed in this study are underlined. Sequences, identifiers and SF motifs can be found in Supporting Information Dataset S1.  (Fetter et al., 2004). By contrast, no significant P f increase was observed for cells expressing the different F-A-A-R-type MIPs from bacteria, algae or moss compared to control oocytes (Fig. 3a).
Bacterial, algal and moss F-A-A-R-type NIPs are permeable to arsenous acid and boric acid when expressed in Xenopus oocytes We determined the transport activity for arsenous acid in oocytes expressing KrAqpN, KnNIP6;1, PpNIP5;1, PpNIP5;2, PpNIP5;3, or the positive control OsLsi1 (Ma et al., 2006). Prokaryotes (light grey circle) possess ars operons for As detoxification and export. Instead of active As efflux transporters, some bacteria possess an ars operon-located aqpN gene (red), which facilitates the transmembrane diffusion of As(III) species. In prokaryotes, arsenate (As(V)) enters the cells via phosphate transporters (purple box). Transcription of arsR (orange) is induced by As(V) and the protein produced, ArsR, regulates the expression of aqpN (red) and arsC (olive-green). The chemical gradient needed to cause As efflux through AqpNs is maintained by the reduction of As(V) to As(III), catalysed by the arsenate reductase ArsC. The SF of many AqpNs is composed of F-A-A-R residues. An AqpN with a F-A-A-R-type SF has been suggested to be transferred to plants by a horizontal gene transfer (HGT) event. During land plant evolution (central panel) AqpNs have been employed by plants and underwent neo-and subfunctionalisation. While in primitive green plants only NIPs with a F-A-A-R-type SF are found (red shading) other functional NIP groups (green shading) such as the NIP-II group amplified. When silicon (Si) and boron (B) became beneficial and essential elements for vascular plant groups, respectively, F-A-A-R-type NIPs disappeared from the genomes of these plants. The change of F-A-A-R-type SFs to other SF compositions in evolving NIP-II and NIP-III group NIPs of higher plants contributed to a shift in metalloid transport properties. Si permeability formed and As transport ability probably diminished in favour of B permeability. The

New Phytologist
Expression of all MIPs resulted in a significant increase in As transport activity compared to water-injected oocytes quantified via ICP-MS analysis (Fig. 3b).
The permeability to boric acid was tested in direct uptake assays into oocytes in the presence or the absence of MIP expression. AtNIP5;1 which is physiologically important for the uptake of B into Arabidopsis roots was used as a positive control (Takano et al., 2006). Quantification of B levels in oocytes showed that the bacterial, algae and moss MIPs significantly increased the B uptake over that of water-injected oocytes (Fig. 3c,d).

Bacterial, algal and moss F-A-A-R-type NIPs do not facilitate the diffusion of silicic acid in Xenopus oocytes
To examine the silicic acid permeability of KrAqpN, KnNIP6;1, PpNIP5;1, PpNIP5;2, and PpNIP5;3 isoforms, we measured Ge accumulation in oocytes expressing these MIPs or OsLsi1, a NIP-III group NIP, as a positive control (Fig. 3e). Germanium dioxide (GeO 2 ), generating germanic acid in solution, proofed to be an excellent tracer for silicic acid in plant-and oocyte uptake assays (Ma et al., 2006). Oocytes expressing OsLsi1 accumulated about 300-fold more Ge than water-injected control oocytes when incubated with 1 mM GeO 2 . None of the F-A-A-R-type MIPs from bacteria, algae or moss increased the Ge uptake in a similar manner to the positive control OsLsi1 or to a level which would point to a physiologically significant Si channel (Fig. 3e). and plant F-A-A-R-type NIPs (KnNIP6;1, PpNIP5;1, PpNIP5;2, PpNIP5;3) in a hypo-osmotic Xenopus laevis oocyte swelling assay. ZmPIP2;5 cRNAand water-injected oocytes were used as positive and negative controls, respectively. Bar charts show means AE 95% CIs of n = 9-16 oocytes. (b) Arsenous acid uptake rates of Xenopus oocytes expressing bacterial (KrAqpN) and plant F-A-A-R-type NIPs (KnNIP6;1, PpNIP5;1, PpNIP5;2, PpNIP5;3). Oocytes expressing OsLsi1 and water-injected oocytes were used as positive and negative controls, respectively. Oocytes were exposed to a 0.1 mM NaAsO 2 containing buffer solution for 30 min. Data represent means AE SD of four pools of oocytes (n = 5-8 oocytes per pool). (c, d) Uptake of 10 boric acid by KrAqpN or KnNIP6;1 (c) and PpNIP5;1, PpNIP5;2, and PpNIP5;3 (d) expressing oocytes in a direct uptake assay. Oocytes expressing AtNIP5;1 and water-injected oocytes were used as positive and negative controls, respectively. Oocytes were exposed to a 5 mM 10 boric acid containing buffer solution for 20 min. Data represent means AE SD of 3-10 pools of oocytes (n = 8-10 oocytes per pool) (c) or 3-5 pools of oocytes (n = 9-11 oocytes per pool) (d). Differences in absolute uptake rates in (c,d) are due to different oocyte batches. (e) Permeability of a bacterial AqpN and plant F-A-A-R-type NIPs to germanic acid (a chemical analogue of silicic acid) in Xenopus oocyte direct uptake assays. Oocytes expressing KrAqpN, KnNIP6;1, PpNIP5;1, PpNIP5;2 and PpNIP5;3 and the control OsLsi1 as well as water-injected oocytes were exposed to 1 mM GeO 2 (which forms in solution Ge(OH) 4 , as the analogue of Si(OH) 4 ) for 30 min in a modified Barth's saline. Bar charts represent means AE SD of 4 pools of oocytes (n = 4-7 oocytes per pool). For metalloid quantifications (b-e), the assayed oocytes were washed, dried, digested and the As, 10 B and Ge concentrations were determined by ICP-MS. Significant differences in transport rates compared to waterinjected negative control oocytes were assessed in (a-e) using Student's ttest: *, P < 0.05; **, P < 0.01; ***, P < 0.001. All oocyte uptake assays have been repeated two to three times with independent oocyte batches and consistent results.  (Figs 1, 2). The number of non-F-A-A-R-type isoforms exceeds the number of F-A-A-Rtype NIPs in vascular plants, for example in horsetails and lycophytes. No F-A-A-R-type NIP was identified in seed plant genomes. The stepwise disappearance of F-A-A-R-type NIPs along the evolution of land plants may imply an evolutionary disadvantage of that SF composition in seed plants compared to NIP-I to NIP-III group SF compositions. F-A-A-R-type NIPs disappeared when B and Si acquired a dominant nutritional role in land plants (Fig. 2), possibly because they are efficient As channels but are suboptimal to serve as efficient B/Si channels. Therefore, we tested the hypothesis whether a switch from an F-A-A-R-to a NIP-II-type SF results in a decrease in arsenous acid transport rates relative to boric acid transport rates. To this aim, SFs were mutated in KrAqpN (a bacterial ars operon-residing AqpN channel with permeability to As), PpNIP5;3 (a F-A-A-Rtype moss NIP of unknown physiological function) and AtNIP5;1 (a physiologically important NIP-II type B channel). The SFs of KrAqpN FAAR and PpNIP5;3 FAAR were mutated into the SF of AtNIP5;1 AIGR (SF of the NIP-II group) resulting in KrAqpN AIGR and PpNIP5;3 AIGR . The SF of AtNIP5;1 AIGR was mutated into AtNIP5;1 FAAR . Oocytes expressing these channel proteins were simultaneously exposed to an arsenous acid and boric acid containing buffer to avoid a quantitative bias on metalloid uptake levels due to potentially different active channel numbers in individual oocyte batches. The absolute values of As and B uptake rates cannot be compared with each other, given that different As and B concentration gradients have been applied for experimental reasons (see the Materials and Methods section). The expression of native AtNIP5;1 AIGR resulted in higher B levels in oocytes than the expression of native PpNIP5;3 FAAR and KrAqpN FAAR channel proteins (Fig. 4a, left panel). Inversely, the exactly same PpNIP5;3 FAAR and KrAqpN FAAR expressing oocytes which have been analysed for B uptake,  Fig. 4 Influence of the ar/R selectivity filter (SF) composition of KrAqpN, PpNIP5;3 and AtNIP5;1 on the permeability to arsenous acid and boric acid. (a) 10 Boric acid (left panel) and arsenous acid (right panel) transport activity was simultaneously determined in direct uptake assays using Xenopus laevis oocytes expressing native AtNIP5;1, KrAqpN and PpNIP5;3 channel proteins. Water-injected oocytes were used as negative controls. Oocytes were exposed to a 5 mM 10 boric acid and 0.1 mM NaAsO 2 containing buffer solution for 30 min. Oocytes were washed, dried, digested and the 10 B and As content was determined by HR-ICP-MS analysis. The As content of water-injected negative control oocytes were below the detection limit of the high-resolution mass spectrometer. Grey chart bars represent the means of metalloid uptake rates AE SD of seven pools (n = 10 oocytes per pool), four pools (n = 8 oocytes per pool), nine pools (n = 9-10 oocytes per pool) and five pools (n = 9-10 oocytes per pool) of water-injected oocytes or oocytes having expressed AtNIP5;1, KrAqpN or PpNIP5;3, respectively. Significant differences in metalloid uptake rates between negative control oocytes or AtNIP5;1 and KrAqpN or PpNIP5;3 expressing oocytes were calculated using the Tukey's test to compare means at a probability level of 5%. Levels of significance are represented by lower-case letters. (b, c) 10 Boric acid (b) and arsenous acid (c) transport activity was determined as in (a) for oocytes expressing native AtNIP5;1 AIGR , KrAqpN FAAR or PpNIP5;3 FAAR , or SF-mutated proteins AtNIP5;1 FAAR , KrAqpN AIGR or PpNIP5;3 AIGR . Grey chart bars in (b) and (c) represent the same data as in the left and right panel in (a), respectively. White chart bars represent the metalloid uptake rate means AE SD of four pools (n = 8 oocytes per pool), 10 pools (n = 10-11 oocytes per pool) and 10 pools (n = 10 oocytes per pool) of oocytes having expressed the mutated proteins AtNIP5;1 FAAR , KrAqpN AIGR or PpNIP5;3 AIGR , respectively. Significant differences in metalloid uptake rates between native AtNIP5;1, KrAqpN or PpNIP5;3 channels and their corresponding SF mutant variants in (b,c) were calculated using Student's t-test: **, P < 0.01; ***, P < 0.001; ns, not significant. The uptake assay has been repeated twice with independent oocyte batches and consistent results. (2020)

Research
New Phytologist possessed significantly higher As levels than the oocytes having expressed AtNIP5;1 AIGR (Fig. 4a, right panel). These results indicate that the significantly higher B and lower As uptake rates, respectively, of AtNIP5;1 compared to KrAqpN and PpNIP5;3 cannot be explained by different expression levels of the three MIP constructs, but must have resulted from different selectivity properties such as a differing SF composition. When the native SF of KrAqpN FAAR was mutated into that of the physiological B channel AtNIP5;1 (KrAqpN AIGR ), significantly decreased As levels (Fig. 4c) and significantly increased B levels (Fig. 4b) were observed in oocytes suggesting a quantitative shift in the transport rates. The expression of the mutant PpNIP5;3 AIGR also resulted in a significant relative decrease of As levels in oocytes compared to that of the native PpNIP5;3 FAAR channel, while B transport rates remained similar (Fig. 4b,c). The expression of mutated AtNIP5;1 FAAR compared to that of AtNIP5;1 AIGR resulted in significantly decreased As and B levels indicating an overall negative impact of that SF composition on AtNIP5;1 metalloid transport capacity (Fig. 4b,c).

YFP:PpNIP5 fusion proteins localise to the plasma membrane
To study subcellular localisation, constructs carrying fusions of YFP and the coding sequences of PpNIP5;1, PpNIP5;2 or PpNIP5;3 were transiently expressed in epidermal tobacco leaf cells. In addition, transformed cells were stained with the fluorescent plasma membrane (PM) marker dye FM4-64. YFP-or FM4-64-dependent fluorescence appeared mainly at the borders of the cells. Overlay of YFP-and FM4-64-channels merged fluorescence signals and indicated that YFP:PpNIP5s fusion proteins localised to the PM (Fig. 5a).

F-A-A-R-type PpNIP5 isoforms partially rescue the boron deficiency phenotype of the Atnip5;1 knock-down mutant
To test whether plant F-A-A-R-type NIPs are functional metalloidoporins in plants and to have an experimentally independent confirmation of the B transport ability, we performed an in planta growth complementation assay. We transformed Atnip5;1 knock-down Arabidopsis lines with genetic constructs in which PpNIP5;1, PpNIP5;2 and PpNIP5;3 were cloned behind the control of the AtNIP5;1 promoter sequence. The Atnip5;1 plants were unable to take up sufficient amounts of B from the soil substrate under our standard growth conditions, resulting in a strong B-deficiency phenotype (Fig. 5b). This growth deficit was complemented when the Atnip5;1 mutant was transformed with the native AtNIP5;1 cDNA (Fig. 5b,c). Independent PpNIP5;1 and PpNIP5;2 expressing lines also significantly complemented the Atnip5;1 mutant growth phenotype (Fig. 5b,c) and had a significantly higher B uptake ability, indicated by the higher shoot B concentrations, than the Atnip5;1 knock-down mutants, unambiguously demonstrating their ability to facilitate the uptake of B into plant roots (Fig. 5d). However, the PpNIP5;1 and PpNIP5;2 expressing lines exhibited still slight B-deficiency symptoms when compared to the wild-type (Fig. 5b). It was previously observed that the expression of functional NIP5 isoforms or the AtNIP5;1pro:AtNIP5;1 construct do not fully restore the shoot B concentrations of Atnip5;1 mutants compared to the wildtype (Diehn et al., 2019). All generated PpNIP5;3 harbouring lines spliced the PpNIP5;3 mRNA resulting in a cDNA sequence which is not encoding a full-length MIP (Fig. S3). Accordingly, PpNIP5;3 expression failed to complement the growth of the Atnip5;1 mutant. In P. patens, PpNIP5s transcript abundances were not regulated in response to varying metalloid feeding and exposure levels (Fig. S4).

Discussion
Bacterial aqpNs are rare, abundantly located in ars operons and facilitate the transport of arsenous acid In accordance with previous studies considering few aqpNs of few bacterial phyla, our phylogenetic analyses suggested that plant NIPs descend from an ancestral aqpN gene of extant Chloroflexi bacteria (Zardoya, 2005;Danielson & Johanson, 2010;Abascal et al., 2014;Finn & Cerd a, 2015;Zardoya et al., 2016;Roberts & Routray, 2017). Alternatively, plant NIPs, as many other 'plant' genes, may derive from cyanobacteria, which represent the ancestors of plant plastids and which frequently possess aqpNs in their genomes. Due to low node support, the exact evolution of NIPs from their bacterial AqpN ancestors across the prokaryoteeukaryote border could not be resolved. The reason for this probably lies in the paucity of NIP sequences from nonseed plants and the long period since these taxa shared a most recent common ancestor. Our database survey illustrated that aqpN genes are much less abundant than genes clustering with the intensively characterised GLPF and AQPZ clades. The total number of identified aqpN sequences is extremely small (Fig. 1). Throughout different prokaryotic phyla, aqpN genes are strikingly located in ars operons. GlpF genes are also found in ars operons but the majority is found in carbon utilisation-related operons (Bienert et al., 2013). None of the (by us) identified aqpNs was part of a carbon regulon and this contrasts with the genetic context of the glycerol channel-encoding glpFs arguing against the hypothesis that glycerol transport constitutes the physiological function of AqpNs and thereby accounts for the original function of the archetype of plant NIPs (Roberts & Routray, 2017).
In agreement with the ars operon localisation of aqpNs, we demonstrated their permeability to As as exemplified by KrAqpN. Together, functional and genetic data suggest that AqpNs play a physiological role in As efflux. AqpNs of diverse bacteria are placed together with arsM or arsH genes in ars operons. ArsH oxidises trivalent to relatively nontoxic pentavalent methylated As species, the latter representing potential substrates for AqpNs/NIPs given that a rice NIP can indeed transport such organic As species (Li et al., 2009;Chen et al., 2015). This may indicate that certain AqpNs and NIPs mediate the efflux of organic in addition to inorganic As species to detoxify cells. Based on the observation that the tested KrAqpN was permeable to boric acid, we also hypothesise that AqpNs of cyanobacterial origins are permeable to boric acid since many cyanobacteria also require B for metabolism (Bonilla et al., 1990). As AqpNs are closest related to members of the AQPZ clade, which play key roles in bacterial water regulation, we also tested the bacterial AqpN for water transport. Despite the phylogenetic vicinity, KrAqpN was not facilitating the diffusion of water (Fig. 3a). This is in agreement with the poor water permeability often observed for the closely related plant NIP-I to NIP-III group isoforms (Pommerrenig et al., 2015).
Surprisingly, BLAST searches identified aqpNs in the genomes of the monocot seagrass Z. marina and the metazoan sea anemone N. vectensis (Fig. 1). The genome database information on these intron-lacking sequences does not allow a conclusion whether they are prokaryotic DNA contaminations, more likely or whether these MIPs have arisen from an unlikely, late HGT. Both scenarios can be explained by the fact that Z. marina and N. vectensis share aquatic habitats with members of the Bacteroidetes, which encode for aqpN genes (Fig. 1).

Occurrence and potential role of F-A-A-R-type NIPs in algae and early land plants
To gain a better understanding on NIP evolution, we compared archetype NIP sequences throughout different taxa of land plants. We identified F-A-A-R-type NIPs in almost all groups of plants up to ferns with the exception of horsetails and hornworts, which might be due to insufficient genome coverage and sequence accessibility for species belonging to these taxa (Fig. 2). The occurrence of NIPs comprising F-A-A-R residue SFs does not correlate with the environmental habitats of the corresponding plant taxa. There is strong evidence that the ancestors of current terrestrial plants were closely related to the ancestors of present-day charophyte taxa (Hori et al., 2014). We identified F-A-A-R-type NIPs in the genome of the charophyte K. nitens. In our transport assays, the algal KnNIP6;1 does not transport water but is permeable to arsenous acid and boric acid, essentially displaying a similar transport profile compared with the tested bacterial KrAqpN. Comparing the existence of NIP genes in the genomes of diverse plant species and the amount of Si found in their tissues suggested the existence of an ancestral Si transport ability of archetype plant NIPs (Trembath-Reichert et al., 2015).

New Phytologist
Our results from the Si uptake assays, however, did not support this hypothesis. The neofunctionalisation of NIPs, namely to conduct Si, must have occurred later in evolution and was probably the driving force for the expansion of the NIP-III group in seed plants.
B is essential only for a few, but not for all, red, brown and green algae, and for diatom species (Carrano et al., 2009). Constantly high concentrations of B in seawater (0.4 mM) suggests that B transport proteins may not be needed for B uptake into algae. Moreover, Klebsormidium species, including K. nitens, lack the cell wall polysaccharide RG-II, which represent the only characterised physiologically relevant B-binding molecules in green plants (Sørensen et al., 2011;Domozych et al., 2012). This suggests that charophytic NIPs are not needed for B uptake into charophytic algae.
With respect to As detoxification, K. nitens encodes an Sadenosylmethionine methyltransferase (AS3MT; equivalent to ArsM in bacteria), the main metabolic enzyme that methylates As in sequential steps for detoxification in chlorophyta, bacteria and humans (Palmgren et al., 2017). Seed plants have lost AS3MT genes (Palmgren et al., 2017). Methylated pentavalent As species such as methylarsonic acid (MMA(V)) and dimethylarsinic acid (DMA(V)) might therefore represent substrates for KnNIP6;1, as it is the case for the plant NIP OsLsi1 (Li et al., 2009). Klebsormidium nitens also encodes an arsenate reductase and two ACR3 efflux transporters, which together represent an efficient arsenite efflux pathway and detoxification machinery. We suggest that the physiological role of KnNIP6;1 may predominantly be As efflux. NIPs or AqpNs have not been identified in sister species to land plants such as glaucophytes, red algae or chlorophytes, which are closely related to charophytes (this study; Anderberg et al., 2011).

F-A-A-R-type plant NIP features suit the molecular demands to sub-and neofunctionalise into physiologically important metalloid channels
To fulfil an ancient role in metalloid uptake or efflux a localisation in the PM is necessary. Our analysis demonstrated that PpNIP5s, which are able to complement the Atnip5;1 B-deficiency phenotype in Arabidopsis, are indeed localised to the PM (Fig. 5). Unlike seed plant NIPs, which respond to metalloid levels (Takano et al., 2006;Tanaka et al., 2011), PpNIP5 transcript abundances were unresponsive to low B levels in the growth medium. Toxic externally applied arsenate concentrations did not result in an increased expression of any of the PpNIP5s, which would be expected for playing a role in detoxification processes (Fig. S4). Under highly toxic external arsenite concentrations, P. patens only formed protonema tissue in which all three PpNIP5 transcript levels were very low also under control growth conditions (Chen et al., 2012;Xiao et al., 2011). These observations suggest that the F-A-A-R-type PpNIP5s have an impact on the metalloid status of P. patens but are not transcriptionally regulated by the tested metalloid treatments. The physiological function of individual PpNIP5 isoforms remains to be identified in future analyses.

F-A-A-R-type NIPs disappear in seed plants for which B and Si is nutritionally important
Bacterial type AqpN-like NIPs in liverworts, mosses, clubmosses and ferns with a F-A-A-R-type SF, remained existent up to the origin of vascular plants. In addition to F-A-A-R-type NIPs, the genomes of the liverwort M. polymorpha, the moss P. patens, the clubmoss S. moellendorffii and the fern S. cucullata possess typical NIPs (MA_158586g0010, MA_60111g0010, PHYPADRAFT_ 147365, SmNIP3;1 to SmNIP3;5, Sacu_v1.1_s0057.g014809, respectively), which cluster with B-allocating NIP-II group isoforms of seed plants (Fig. 2). By contrast, F-A-A-R-type NIP genes are absent in the published genomes of both angiosperms and gymnosperms. This suggests that F-A-A-R-type NIP transport selectivity and/or ability had no evolutionary advantage for these plant species. The demand for B significantly correlates with the amount of borate ester cross-linked RG-II found in the plants' pectin cell wall fraction (Matsunaga et al., 2004;O'Neill et al., 2004). Our data showed that the taxa-specific occurrence and proliferation of NIP-II and NIP-III isoforms, at the expenses of F-A-A-R-type-NIPs, either positively match RG-II levels of the corresponding plant taxa (Fig. 2) or the Si demand of plant taxa such as Poaceae or Cucurbitaceae species, respectively.
We tested whether the evolutionary switch from F-A-A-Rtype to NIP-II-type SFs might have had an impact on pore selectivity characteristics influencing, for instance, As and B transport rates. Indeed, the mutagenesis approach corroborate the hypothesis, as it demonstrated that MIPs of bacteria and plants encoding the F-A-A-R-type SF have a higher permeability to As compared to their mutants encoding an A-I-G-R-type SF being intrinsic to typical seed plant B channels, such as AtNIP5;1. Moreover, the mutated variant, KrAqpN AIGR , of the ars operonlocated native KrAqpN FAAR had a relatively increased B permeability when it contained the SF residues of the physiological B channel AtNIP5;1 AIGR . The permeability of AtNIP5;1 AIGR to both As and B was reduced when mutated into AtNIP5;1 FAAR . This, in AtNIP5;1 FAAR , decreased metalloid permeability was expected as another mutational study demonstrated that the native R1 position of the SF of AtNIP5;1 (A R1 -I R2 -G R3 -R R4 ) is essential for its metalloid transport ability (Mitani-Ueno et al., 2011).
Our mutational approach demonstrated (1) that ars operon-localised AqpN As channels encoding a F-A-A-R-type SF have a higher As transport capacity compared to their mutants encoding a NIP-II-type SF and (2) that physiologically important NIP-II B channels encoding a NIP-II-type SF have a higher B transport capacity compared to their mutants encoding a F-A-A-R-type SF. These results indicate that plant AtNIP5;1 and bacterial KrAqpN possess a SF pore layout which serves an optimal transport regulation of their physiological relevant substrate, namely boric acid and arsenous acid, respectively (Fig. 4a).
A high permeability to As in F-A-A-R-type NIPs might represent a physiological disadvantage for instance under metalloid toxic conditions. In the genomes of species with F-A-A-R-type MIPs such as bacteria, charophytes (K. nitens), mosses (P. patens), ferns (Ceratopteris richardii, Pteris vittata), and lycophytes www.newphytologist.com (S. moellendorffii) ACR3 genes are present. ACR3 proteins are PM or tonoplast-localised trivalent As antiporters which belong to the BART (bile/arsenite/riboflavin transporter) superfamily and are crucial for As detoxification. No ACR3 homologs have been identified in angiosperms yet (Mansour et al., 2007;Indriolo et al., 2010). Therefore, the occurrence of F-A-A-R-type NIP channels correlating with that of ACR3 transporters in plant taxa, suggests that F-A-A-R-type NIPs did not replace ACR3 efflux transporters in As detoxification processes in plants, while F-A-A-R-type AqpNs have replaced active As efflux transporters in bacterial ars operons. It is tempting to hypothesise that the presence of F-A-A-R type NIPs in plants necessitate additional efficient As detoxification mechanisms such as ACR3 transporters. The physiological function of individual F-A-A-R-type NIPs in old terrestrial plant lineages remains to be elucidated. In summary, the transport functions and the phylogeneticand genetic context analyses are consistent and suggest that today's seed plant NIP function has evolved from bacterial As efflux channels and that amino acid changes in the SF were necessary but not sufficient for that change. In charophytic algae, the original role of ancestral NIPs may remain conserved. While the role of NIPs in mosses remains enigmatic, in seed plants SF alterations together with currently uncharacterised molecular channel changes converted bacterial As efflux proteins into essential plant nutrient transporters. The functional analyses strongly suggest that nutritional demands of terrestrial plants were a strong driver for functional divergence of NIPs. Plant NIP paralogues of AqpNs underwent subfunctionalisation by specialising on B transport regulation and specificity. The ancestral, however, physiologically insignificant B transport ability of NIPs of early land plants gained high importance when stable and flexible Bdependent cell wall properties became crucial for upright plant growth. In addition, neofunctionalisation of NIPs occurred when a few NIP-II (in horsetails) but in particular NIP-III group isoforms developed their ability to transport Si, which was not generally intrinsic to ancestral F-A-A-R-type NIPs.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.    Dataset S1 Information on major intrinsic protein (MIP) sequences which have been used for the phylogenetic analyses of this study.
Methods S1 Detailed information on methods.
Table S1 Nucleotide sequences of codon-optimised MIP genes.  New Phytologist is an electronic (online-only) journal owned by the New Phytologist Trust, a not-for-profit organization dedicated to the promotion of plant science, facilitating projects from symposia to free access for our Tansley reviews and Tansley insights.
Regular papers, Letters, Research reviews, Rapid reports and both Modelling/Theory and Methods papers are encouraged.
We are committed to rapid processing, from online submission through to publication 'as ready' via Early View -our average time to decision is <26 days. There are no page or colour charges and a PDF version will be provided for each article.
The journal is available online at Wiley Online Library. Visit www.newphytologist.com to search the articles and register for