The protective role of aquaporins in the freeze-tolerant insect Eurosta solidaginis: functional characterization and tissue abundance of EsAQP1

Organisms that experience excessive shifts in solute concentration and water distribution during osmotic challenge risk sustaining structural damage to their cells and/or disruption of proper cell function (Hadley, 1994). One adaptation that is well known to protect cells against the osmotic challenges of freezing and desiccation is the accumulation of low molecular weight compounds, such as glycerol, urea and trehalose (Ring and Danks, 1994). In preparation for winter, some insects amass multi-molar levels of these molecules, which colligatively reduces the rate of ice formation and the amount of ice that develops (Lee, 1991; Lee, 2010). In addition, these compounds decrease the rate of water loss and serve as molecular substitutes for water during dehydration (Ring and Danks, 1994).

Recently, aquaporins (AQPs) have also been implicated in protecting cells against osmotic damage [see Borgnia et al. and Campbell et al. (Borgnia et al., 1999; Campbell et al., 2008) and references therein]. These proteins, which are water-selective, transmembrane channels found in both prokaryotic and eukaryotic cells, act as conduits for rapid water flux (∼3×10⁹ water molecules subunit^–1 cell^–1) greatly exceeding simple diffusion across the membrane (Borgnia et al., 1999; Preston et al., 1992). During freezing, concentration gradients build between the intracellular and extracellular fluid compartments, when water becomes osmotically inactive as ice. As the extracellular concentration of solutes increases, intracellular water must rapidly leave the cell to alleviate the increasing osmotic pressure and prevent damage to the membrane (Mazur, 1984; Muldrew et al., 2004). Linking AQPs to this rapid redistribution of water may ultimately provide another dimension to our understanding of adaptations for freeze tolerance and survival of osmotic stress.

Cells of freeze-tolerant organisms must simultaneously coordinate the efflux of water and influx of cryoprotectants, such as glycerol, in response to the concentration of extracellular solutes during ice formation. Aquaglyceroporins (GLPs), a subgroup of AQPs permeable to water and small uncharged solutes (Borgnia et al., 1999), are implicated in the cellular movement of glycerol during freezing (Izumi et al., 2006; Philip and Lee, 2010; Philip et al., 2008; Zimmerman et al., 2007). The importance of glycerol uptake through GLPs in freezing survival is suggested by the improved cryopreservation survival of mouse oocytes that artificially express GLPs (Edashige et al., 2003). The ability of AQPs and GLPs to redistribute molecules may also explain why the rate of freezing can influence the survival of insects (Bale et al., 1989) and frogs (Costanzo et al., 1991). During rapid freezing, the rate of ice formation may simply overwhelm the cell's ability to redistribute water and cryoprotectants effectively through AQPs and GLPs. The role of these proteins in water and low molecular weight solute movement, as well as their ability to allow cells to rapidly respond to osmotic stress, suggests that AQPs and GLPs play a significant role in protecting cells during freezing.

One organism that naturally survives freezing is the goldenrod gall fly, Eurosta solidaginis (Baust and Nishino, 1991). Larvae of these flies are exposed to environmental extremes as they overwinter inside the stem of goldenrod (Solidago spp.) (Layne, 1993). Northern populations experience temperatures as low as –40°C within their galls (Baust and Lee, 1981). Consequently, E. solidaginis have developed the remarkable capacity to withstand extreme winter conditions; fully freeze-tolerant larvae can survive exposure to –80°C (Bennett and Lee, 1997; Philip and Lee, 2010). As such, E. solidaginis serves as an excellent model organism not only to investigate natural adaptations to freezing but also to potentially provide insight for the clinical cryopreservation of mammalian tissues.

Our previous study demonstrated that tissues from E. solidaginis are unable to survive freezing in the presence of an AQP/GLP inhibitor (Philip et al., 2008). This result suggests a critical role for AQPs/GLPs in facilitating water exosmosis and/or cryoprotectant (e.g. glycerol) uptake during freezing. To further our understanding of their role in meeting osmotic challenges, we isolated, cloned and sequenced an AQP from E. solidaginis and functionally characterized its water and solute permeability. Additionally, we investigated the abundance of this AQP among different tissues during seasonal cold-hardening in E. solidaginis larvae. The functional characterization and observed changes in seasonal protein abundance for this AQP provide further evidence that these proteins are important in promoting freeze tolerance.

Spherical galls containing the larvae of E. solidaginis (Diptera: Tephritidae), Fitch 1855, were collected from goldenrod plants (Solidago sp.) near Oxford, OH, USA, including at the Miami University Ecology Research Center (39°31′57″N, 84°43′23″W). Larvae were collected in late October (2007) for cDNA cloning, and early October and late December (2007) for immunoblots. Larvae for immunoblots were snap-frozen and stored at –80°C.

Larvae were homogenized in Ultraspec RNA isolation reagent (Biotecx, Houston, TX, USA). The poly(A)⁺ mRNA fraction was isolated from the total RNA using oligo(dT)-cellulose (Biotecx) following standard protocols (Sambrook and Russell, 2001). cDNA was synthesized from mRNA using the Reverse Transcriptase System (Promega, Madison, WI, USA). The cDNA was reverse transcribed with a cDNA cloning primer (cDNA cloning primer: 5′-GGCCACGCGTCGACTAGTACTTTTTTTTTTTTTTTTTV-3′, Integrated DNA Technologies, Coralville, IA, USA) and used in subsequent PCR amplifications.

PCRs were carried out using degenerate primers [MIP/NOD (F): 5′-GG(GATC)G(GC)(GATC)CA(TC)(TCA)T(GATC)AA(TC)CC(GATC)GC(GATC)GT(GATC)AC-3′; MIP/NOD (R): 5′-C(TG)(GATC)GC(GATC)GG(AG)TTCAT(GATC)(GC)(ATC)(GATC)GC(GATC)CC(GATC)GT-3′] designed to conserved regions of AQP sequences (Elvin et al., 1999). Thermocycler conditions were: 94°C for 2 min with 40 cycles of 94°C for 1 min, 50°C for 1 min, 72°C for 1 min. The ends of the amplicon were finished during a final incubation at 72°C for 10 min. All reactions were analyzed by agarose gel electrophoresis; bands between 400 and 500 base pairs were excised, processed using Wizard SV Gel and PCR Clean-Up System (Promega), and subcloned into a pGEM-T Easy vector (Promega) as per the manufacturer's instructions. Vectors were transformed into JM109 competent cells (Promega) and white colonies were selected for by α-complementation of LacZ on LB plates containing 100 μg ml^–1 ampicillin supplemented with IPTG and X-gal. White, insert-containing colonies were grown in 4 ml of LB media (containing 100 μg ml^–1 ampicillin) overnight and plasmid DNA was isolated using Wizard SV Minipreps DNA Purification System (Promega) and sequenced using a BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems, Foster City, CA, USA) and standard T7 or SP6 sequencing primers (Promega). Products were analyzed on an Applied Biosystems 310 or 3130× DNA Sequencer in the Center for Bioinformatics and Functional Genomics at Miami University.

Complete cDNA sequences were obtained by 5′ and 3′ rapid amplification of cDNA ends (RACE). The 3′ RACE forward primer was complementary to the putative AQP (EsAQP1 3′ RACE primer: 5′-GGCGGGGCGTGTAAGCGTTTTGCGTGCC-3′) and the reverse primer (3′ RACE PCR: 5′-GGCCACGCGTCGACTAGTAC-3′; Integrated DNA Technologies) corresponded to the cDNA cloning primer. This method yielded the remainder of the 3′ end, as well as the untranslated region (UTR). 5′ RACE was performed on terminal deoxynucleotidyl transferase (TdT)-tailed cDNA [EsAQP1 5′ RACE (F): 5′-GGCCACGCGTCGACTAGTACGGGIIGGGGIGGGIIG-3′, EsAQP1 5′ RACE (R): 5′-GCCCAAATGACCGAGGGTGACAGAGAGACCGATGGCGAG-3′] to obtain the previously unidentified 5′ end of the AQP. The entire sequence was amplified with a gene-specific forward primer [EsAQP1 Whole Sequence (F): 5′-GCAAGCAATTTCACGAGAAGCTTATTAAATTTAAAACACTGCCGG-3′] and the 3′ RACE PCR reverse primer, and was cloned into a pGEM-T Easy vector. The rat AQP3 plasmid was constructed from RNA isolated from rat kidneys, using rat AQP3 forward (5′-GCCTCGCCATGGGTCGACAGAAGGAG-3′) and reverse (5′-GCATCACAAGGGGGCTTTTATGGGGTGTCCC-3′) primers based on Edashige et al. (Edashige et al., 2003). Human AQP1 plasmid (Preston et al., 1992) was a generous gift of Prof. Peter Agre (The Johns Hopkins University, Baltimore, MD, USA). Human AQP1 was used as a positive control for the water permeability assay and rat AQP3 was used as a positive control for both the water and glycerol permeability assays.

DNA sequences were edited for quality and aligned by ClustalW in BioEdit 7.0.9.0 (Hall, 1999). In silico translated aquaporin cDNAs from E. solidaginis were compared by BLASTp with GenBank at NCBI. Amino acid analysis was carried out using BioEdit 7.0.9.0. Hydropathy plots (http://ca.expasy.org/tools/protscale.html) (Kyte and Doolittle, 1982) were constructed to predict the number of membrane-spanning regions, as well as to make comparisons to known aquaporins from GenBank/NCBI. For the predicted transmembrane regions, we evaluated the primary structure regions using TMHMM Server, V. 2.0 (http://www.cbs.dtu.dk/services/TMHMM/) as well as structural comparisons at ExPaSy SwissModel (http://swissmodel.expasy.org/). Based on these results, we constructed a two-dimensional membrane topology plot using the T_EXtopo package (Beitz, 2000) and employed a group of selected, aligned aquaporins for the purpose of shading via the T_EXshade package. Based on the alignment, individual amino acids were shaded if the majority of the residues had similar characteristics (‘similar’), the majority were the same residue (‘conserved’), or all residues were identical (‘invariable’) at a particular location. Putative phosphorylation and glycosylation sites were identified by ScanProsite (http://expasy.org/tools/scanprosite/) (De Castro et al., 2006) on the ExPaSy server.

A three-dimensional analysis of structural conservation among AQPs was performed with the structural alignment of multiple proteins (STAMP) tool within the Visual Molecular Dynamics (VMD) program (Humphrey et al., 1996). Residues were colored based on their Q-value, with blue assigned to structurally conserved regions and red to sites with no correspondence in structure.

Phylogenetic analysis of the putative E. solidaginis AQP was performed by Bayesian analysis using aligned amino acid sequences of insects and selected taxa for which functional data have been published (supplementary material Table S1). Nucleotide sequences were aligned based on a ClustalX amino acid alignment using a codon-constraint approach implemented by CodonAlign2.0 (Hall, 2004). Analysis was performed using PAUP4.0*b10 and MrBayes 3.1.2. Tree construction was done by MrBayes 3.1.2 using a ‘Generalised Time Reversible plus Invariant plus Gamma’ (GTR+I+Γ) evolutionary model (Waddell and Steel, 1997) as selected by MrModeltest v2.3 (Nylander, 2004). Four runs using four chains and 10,000,000 generations were performed. All other parameters in MrBayes 3.1.2 were left at default. Stationarity was assessed using the ‘sump burnin=2500’ command and by examining the plot of the generation versus the log likelihood values. A 50% majority rule consensus tree was generated using the command ‘sumpt burnin=2500’. A newick tree file, including branch lengths and Bayesian consensus values, was imported into the tree explorer of MEGA4 (Kumar et al., 2008) to draw the phylogram.

Before cRNA for oocyte injections was transcribed, a fragment of the Xenopus β-globin 5′-UTR was subcloned into the pGEM-T Easy plasmid, downstream of the T7 promoter site and upstream of the ATG start site, as described in detail elsewhere (Zimmerman et al., 2007). This step was necessary because our preliminary swelling assays suggested that proteins were not translated efficiently by Xenopus oocytes when the cRNA did not contain the Xenopus β-globin 5′-UTR fragment. EsAQP1 and rat AQP3 plasmids were linearized with SpeI and cRNA was synthesized with the mMessage mMachine T7 kit (Applied Biosystems) as described by the manufacturer. Human AQP1 was linearized with SmaI and cRNA was synthesized with the mMessage mMachine T3 kit (Applied Biosystems). cRNA was treated with TURBO DNase (Applied Biosystems), recovered with a phenol:chloroform extraction and ethanol precipitation, and resuspended in RNase-free water. The cRNA from each reaction was verified on a denaturing agarose gel to contain only a single product, then aliquoted and frozen at –80°C.

Ovaries from double-pithed, mature female Xenopus laevis frogs were removed according to a protocol approved by the Institutional Animal Care and Use Committee at Miami University (Protocol 750). The tissue was minced into small pieces using fine forceps and treated with 2 mg ml^–1 collagenase A (Roche, Mannheim, Germany) in Ca²⁺-free modified Barth's solution (Ca²⁺-free MBS; 88 mmol l^–1 NaCl, 1 mmol l^–1 KCl, 2.4 mmol l^–1 NaHCO₃, 10 mmol l^–1 Hepes, 0.82 mmol l^–1 MgSO₄, pH 7.5) for 3 h at 15°C with gentle agitation. Oocytes were washed repeatedly with fresh MBS (Ca²⁺-free MBS with 0.33 mmol l^–1 Ca(NO₃)₂, 0.41 CaCl₂; 177 mOsm) to remove residual collagenase, and Stage V and VI oocytes were incubated overnight at 15°C in MBS containing 10 mg l^–1 gentamicin. Any follicular layers that remained attached were manually removed from oocytes using fine forceps and oocytes were injected (Nanoject II, Drummond, Broomall, PA, USA) with either 50.6 nl of 500 ng μl^–1 cRNA recovered from the mMessage mMachine reactions or the same volume of RNase-free water. Oocytes were incubated in MBS containing 10 mg l^–1 gentamicin for 3 days (with daily changes of media) at 15°C before being used in swelling assays.

To examine whether mercury is inhibitory to these AQPs, a subset of oocytes (N=6) were incubated in 0.3 mmol l^–1 HgCl₂ for 10 min before hypotonic challenge in (Hg²⁺-free) diluted MBS. To determine whether mercury inhibition was reversible, another group of oocytes (N=5) were incubated in 0.3 mmol l^–1 HgCl₂ for 10 min followed by 15 min in 5 mmol l^–1 β-mercaptoethanol (BME), a reducing agent.

To determine the abundance of EsAQP1 protein in tissues, larvae that had been collected in October and December and immediately frozen at –80°C were thawed and quickly dissected. Fat body, gut, Malpighian tubule, salivary gland and brain were removed and placed on ice in protein isolation medium (250 mmol l^–1 sucrose and 10 mmol l^–1 triethanolamine adjusted to pH 7.6 with 1 mol l^–1 NaOH) containing Protease Inhibitor Cocktail (Sigma P8340; 1:100 dilution). Each tissue type was pooled (N=4 individuals) and proteins were extracted by grinding tissues with a Teflon pestle and sonicating the solution. Homogenates were centrifuged (4°C) for 10 min at 1000 g to remove large cellular debris, the protein concentration of the supernatants (whole protein fractions) was determined with a Bradford assay (BioRad, Hercules, CA, USA) and 3 μg of protein were analyzed through immunoblotting and densitometry techniques described previously (Philip and Lee, 2010). A polyclonal rabbit EsAQP1 antibody was developed using a peptide corresponding to the C-terminus of the predicted EsAQP1 protein (Cys-KYRTHADDREMRKLDSTRDYA) (Proteintech, Chicago, IL, USA). Rabbit serum was affinity purified and used in immunoreactions at a final concentration of 0.5 μg ml^–1 in TBS-T containing 5% non-fat milk. The specificity of this and other primary antibodies was characterized by probing Xenopus oocytes expressing either EsAQP1 or rat AQP3. Oocyte proteins were isolated in media containing 20 mmol l^–1 Tris-HCl, pH 7.6, 0.1 mol l^–1 NaCl and 1% Triton X-100 (Bossi et al., 2007) with protease inhibitors (Sigma; 1:100 dilution). Isolates from EsAQP1- and rat AQP3-expressing oocytes were probed with EsAQP1, rat AQP3 (Sigma A0303, 1:300 dilution) or rat AQP4 (Sigma A5971, 1:1000 dilution) primary antibodies.

Changes in relative volume within each group of AQP-injected oocytes during the swelling assay are plotted as means ± 1 s.e.m. Water (P_f) and solute (P_sol) permeabilities among different AQPs were compared using an ANOVA and Bonferroni–Dunn post hoc test.

We isolated, cloned and sequenced a partial cDNA of an AQP from E. solidaginis using degenerate primers designed against conserved regions of AQPs. The full sequence (1305 bp), including the UTRs, was obtained through 5′ and 3′ RACE (GenBank accession no. FJ489680). In silico translated cDNA revealed that this new AQP possesses two aspargine–proline–alanine (NPA) amino acid motifs (Fig. 1A). A Kyte–Doolittle hydropathy plot (supplementary material Fig. S1) suggests there are six transmembrane regions connected by five loops, such that the N- and C-termini are located in the cytosol. Two-dimensional topology of this novel AQP showed the locations of the transmembrane regions, the NPA motifs and putative post-translational modification sites (Fig. 1B). Predicted cytosolic casein kinase II (Thr 251, Ser 263) and protein kinase C (Ser 246, Ser 263) phosporylation sites and an extracellular N-glycosylation (Asn–Ile–Thr–Glu 132) site are indicated on the diagram (Fig. 1B).

The phylogenetic relationship of this E. solidaginis aquaporin to functionally characterized AQPs from human (Homo sapiens sapiens), quail (Coturnix coturnix), Cope's gray tree frog (Hyla chrysoscelis), Mozambique tilapia (Oreochromis mossambicus), vinegar fly (Drosophila melanogaster), yellow fever mosquito (Aedes aegypti), mulberry silkworm moth (Bombyx mori), green leafhopper (Cicadella viridis), triatomid bug (Rhodnius prolixus), sleeping chironomid (Polypedilum vanderplanki), pea aphid (Acyrthosiphon pisum), brown dog tick (Rhipicephalus sanguineus), nematode roundworm (Caenorhabditis elegans) and Escherichia coli was determined by Bayesian analysis (see supplementary material Table S1 for GenBank accession numbers). The outgroup in our phylogram was a bacterial aquaporin (E. coli AQPZ). A phylogram based on the Bayesian analysis is presented as a 50% majority rule consensus tree in Fig. 2. The evolutionary model indicated as most suitable by MrModelTest (Nylander, 2004) was the GTR+I+Γ model of substitution. In this analysis, we used codon-constrained nucleotide alignments of the coding regions of the aquaporins (supplementary material Fig. S2). The newly sequenced E. solidaginis AQP is most closely related to the dehydration-inducible P. vanderplanki AQP1 (PvAQP1), to which it is 52.3% identical at the amino acid level. Based on its identity and phylogenetic relationship to PvAQP1, we designated this AQP cloned from E. solidaginis as EsAQP1. Both EsAQP1 and PvAQP1 are ancestral to a clade of AQPs in organisms ranging from insects to mammals. According to this phylogenetic reconstruction, EsAQP1 is distantly related to GLPs from both human (HsAQP3) and Cope's gray tree frog, H. chrysoscelis (HcAQP3). The alignment of the EsAQP1 amino acid sequence to the most closely related insect (PvAQP1) and human (HsAQP4) AQPs using ClustalW (BLOSUM62) demonstrated high homology among the three proteins, especially near the NPA motifs (Fig. 1). The predicted three-dimensional structures of EsAQP1, PvAQP1 and humanAQP4 were compared using the VMD program to determine whether the amino acid differences affected the overall structure of the proteins. Areas of high structural conservation, as predicted by the Q-values, predominated in the transmembrane helices (Fig. 1C). In contrast, the extracellular connecting loops were more structurally divergent, as demonstrated by the prevalence of red residues (Fig. 1C).

Although the amino acid sequence, alignments, thread-through modeling and phylogenetic analysis suggested that EsAPQ1 was an aquaporin, it was important to confirm this prediction by functionally characterizing the water and solute permeability of this protein. We measured water permeability by translating exogenous AQPs in Xenopus oocytes and exposing the cells to a hypotonic challenge. Oocytes translating EsAQP1 cRNA responded to hypotonic extracellular fluid by rapidly swelling (Fig. 3; see supplementary material Movies 1–4), thereby demonstrating the protein's permeability to water. The mean osmotic water permeability coefficient (P_f) for EsAQP1 using Eqn 1 was more than 9-fold that of sham-injected oocytes (P<0.0001; Fig. 4A). Despite this, EsAQP1 demonstrated only 57% of the water permeability of human AQP1 (P<0.0001; Fig. 4A).

Exposure to mercury-containing compounds, including HgCl₂, inhibits water transport in most AQP isoforms, as these molecules are thought to covalently bond to the AQP channel and occlude molecular movements through the pore (Preston et al., 1993). Mercury sensitivity of EsAQP1 was demonstrated by a significant decrease in water permeability following exposure to mercuric chloride (P<0.0001; Fig. 4B). Generally, mercury inhibition of an AQP can be partially or totally reversed if the oocytes are exposed to a reducing agent, such as BME (Preston et al., 1993). Despite an increase in water permeability of mercury-inhibited EsAQP1 oocytes after exposure to BME, this change was not statistically significant (as measured by the conservative Bonferroni–Dunn post hoc test; P=0.035; Fig. 4B).

In addition to permitting water flux, GLPs are a subset of AQPs that permit small, uncharged solutes, such as glycerol or urea, to pass through the plasma membrane. Apparent uptake, as measured by changes in oocyte volume in response to glycerol- or urea-containing isosmotic media, was not significantly different from that of sham-injected oocytes in either experiment (P>0.05; Fig. 5). In contrast, rat AQP3 oocytes serving as a positive control, swelled because of their permeability to glycerol and urea (Fig. 5) (Borgnia et al., 1999).

After characterizing the molecular structure and physiological function of EsAQP1, we examined the tissue-level translation of this protein. The predicted amino acid sequence for EsAQP1 was used to design a polyclonal antibody against an antigenic portion of the AQP corresponding to amino acids 248–268 on the C-terminus. To verify the antibody reacted with EsAQP1, we probed protein isolated from EsAQP1 and rat AQP3 cRNA-injected Xenopus oocytes and found that the EsAQP1 antibody only reacted with EsAQP1 proteins (Fig. 6A). Immunoblots probed with the EsAQP1-specific antibody resulted in a single band at 25 kDa, which is slightly lower than the in silico predicted molecular mass of 29 kDa. Additionally, proteins isolated from EsAQP1 cRNA-injected Xenopus oocytes did not cross-react with commercially available rat AQP3 and rat AQP4 antibodies (Fig. 6A).

The newly characterized antibody to EsAQP1 was used to assay differences in the abundance of EsAQP1 in fat body, gut, Malpighian tubule, salivary gland and brain during the cold and desiccation hardening that occurs between October and December (Fig. 6B). When standardized to total protein content, the abundance of EsAQP1 was highest in the brain. The Malpighian tubules had the next most abundant amount of EsAQP1, albeit 6-fold less than levels found in brain. When the same tissues were examined in late December, there was an overall increase in the abundance of EsAQP1 in the tissues assayed (Fig. 6B), including a nearly 4-fold increase in abundance of EsAQP1 in the salivary gland.

Surviving osmotic challenge requires the regulation and maintenance of water and solute balance between body compartments. AQPs, which rapidly redistribute water and certain small solutes across the membrane, are associated with the physiological response to stresses that require the management of cellular water, such as anhydrobiosis (Kikawada et al., 2008), feeding on osmotically challenging foods (Le Caherec et al., 1996; Shakesby et al., 2009) and survival of freezing (Izumi et al., 2006; Philip et al., 2008; Philip and Lee, 2010). To better understand the role of these proteins in osmotic stress, we identified, functionally characterized the water permeability and determined the tissue distribution of an AQP during seasonal cold acclimation in E. solidaginis.

We determined that a novel E. solidaginis AQP, EsAQP1, was similar in sequence, predicted topology and function to previously identified and reported AQPs (Campbell et al., 2008). This novel sequence contained two NPA motifs, six membrane-spanning regions and five connecting loops, all hallmarks of AQP proteins (Fig. 1). Additionally, the amino acids predicted to make the ar/R (aromatic/arginine) selectivity filter in PvAQP1 (Phe 56, His 181, Arg 196) (Kikawada et al., 2008) were in nearly identical positions in EsAQP1 (Phe 56, His 180, Arg 195).

Sequence comparisons among different AQPs can indicate their relationship to each other and suggest how the proteins have evolved. Alignments of EsAQP1 at both the nucleotide and amino acid levels with other functionally characterized AQPs indicate it shares a high percentage of identity with PvAQP1 from the sleeping chironomid P. vanderplanki (Fig. 1A). The closest human ortholog to EsAQP1 was AQP4, with an identity of 27.8% at the amino acid level (Fig. 2). EsAQP1, PvAQP1 and human AQP4 shared transmembrane regions that are well conserved, especially near the pore-forming residues of the NPA motifs (Fig. 1A,C). In contrast, both the N- and C-termini were more divergent among these three AQPs, suggesting these regions are under less selective pressure. Although amino acid changes in the N- and C-termini do not directly change the functional transmembrane pore, these alterations can have significant regulatory effects on the aquaporin. Drosophila melanogaster BIB, a member of the AQP family, has a C-terminus that is ∼10-fold longer than that of ‘traditional’ AQPs and contains many targets for tyrosine kinase regulation (Campbell et al., 2008). Although EsAQP1 does not have an extended C-terminus, we predicted several putative phosphorylation sites that could enable functional regulation of EsAQP1 during physiological processes (Fig. 1B). These putative sites and their potential role remain to be elucidated in further regulatory studies, which are currently underway.

Despite the existence of a large number of AQP isoforms among diverse forms of life, these proteins are highly conserved. Phylogenetic analysis of EsAQP1 and a representative group of other functionally characterized AQPs was performed using Bayesian methodology and a GTR+I+Γ model of molecular evolution (Waddell and Steel, 1997). This model of evolution is particularly well suited for ancient, conserved protein-coding sequences, as it takes into account the possibility of sequence reversion, as well as the likelihood of invariant sites and the recognition that not all sites will be under the same selective pressure. Additionally, using nucleotides that code for amino acids constrained in the alignment as triplet codons likely facilitates much greater depth of resolution in phylogenetic reconstructions (Simmons et al., 2002). The clade of EsAQP1 and PvAQP1 is intriguing because, unlike many of the other organisms in the phylogenetic tree, both insects are tolerant of extreme environmental conditions. Larvae of P. vanderplanki can survive exposure to extreme cold (–270°C) and heat (102°C) (Hinton, 1960), and have a dehydration-inducible AQP (PvAQP1) that is presumed to facilitate rapid water loss during entry into anhydrobiosis (Kikawada et al., 2008). In addition to EsAQP1, we analyzed the partial sequence from another E. solidaginis AQP cDNA and found it was most closely related to another AQP of P. vanderplanki (PvAQP2; data not shown). Accordingly, the close phylogenetic relationship of these AQP proteins from two insects that experience similar osmotic challenges suggests that evolutionary pressures drive adaptations at the molecular level to promote organismal tolerance of extreme stress.

Water permeability measurements through artificial expression of AQPs in Xenopus oocytes provided evidence that EsAQP1 is a functional water channel. It should be noted that the rate of water flux was lower in EsAQP1 than in both human AQP1 and rat AQP3 (Fig. 4A). While EsAQP1 may naturally possess below-average water permeability, other factors could explain its reduced water flux rates. One such explanation is a change in the amino acid sequence of EsAQP1 from an ancestral protein. A substitution that changes an alanine (A) to a valine (V) in one of the normally conserved NPA motifs is thought to contribute to low water permeability in a R.prolixus AQP (Rp-MIP) (Echevarria et al., 2001). Although this particular substitution does not explain the reduction in permeability of EsAQP1, which contains two NPA motifs, an amino acid substitution(s) elsewhere in the protein could alter its water transport abilities. Changes to residues outside the active site of proteins can still impact their activity, as demonstrated by a single amino acid substitution altering the kinetic properties of lactate dehydrogenase A in barracudas (Holland et al., 1997). Another possible cause of the relatively lower permeability of EsAQP1 is the non-native environment in which it was tested. The permeability of some AQPs is based on post-translational modifications in response to environmental signals, such as temperature (Azad et al., 2004). If similar cues are required by EsAQP1, we may be unintentionally influencing its permeability by expressing the proteins in Xenopus oocytes. Understanding the mechanism by which EsAQP1 is regulated in vivo may provide valuable insight into the mechanism by which this protein functions in E. solidaginis.

Survival of freezing requires cells to exchange both water and cryoprotectants between intracellular and extracellular compartments. The permeability of AQP/GLPs to these molecules implicates these proteins in protecting against freeze-induced damage in a range of organisms, including yeast (Tanghe et al., 2002), insects (Izumi et al., 2006; Philip et al., 2008) and amphibians (Zimmerman et al., 2007). The ability of EsAQP1 channels to redistribute water across the membrane faster than by simple diffusion through membrane, as shown by our in vitro assay (Fig. 4), would facilitate cellular dehydration that is caused by the formation of concentrated extracellular fluid during freezing. Therefore, we predict that EsAQP1 is similar to other AQP/GLPs, and serves as a pathway for water flux to dissipate osmotic pressures across the membrane during freezing.

In addition to their important role in protecting cells against the rapid development of osmotic gradients during freezing, AQPs appear to be regulated during exposure to desiccating conditions in insects (Kikawada et al., 2008; Philip and Lee, 2010). During the autumn, the microhabitat of E. solidaginis becomes highly xeric as the water content of the surrounding gall tissue decreases to ∼20% (Rojas et al., 1986). While the protective mechanism of AQPs during exposure to desiccating conditions is not known, these proteins, including EsAQP1, may serve as a pathway by which water movement between body compartments is controlled.

Cells not only need to rid themselves of excess water during freezing but also must accumulate cryoprotectants to protect the conformation of proteins (Carpenter and Crowe, 1988) and membrane structure (Karow, 1991) against osmotic extremes. Because glycerol is an important cryoprotectant in E. solidaginis, we tested whether EsAQP1 provided a pathway for its flux across the membrane. Our experimental evidence demonstrated that EsAQP1 is impermeable to glycerol (Fig. 5), and therefore we suggest another AQP isoform is responsible for the movement of this cryoprotectant during larval freezing. Based on previous studies, which identified multiple AQP- and GLP-like proteins in E. solidaginis (Philip et al., 2008; Philip and Lee, 2010), and the putative number of AQPs existing in insect genomes [seven in Drosophila spp. and five in mosquito (Campbell et al., 2008)], EsAQP1 is likely only one of several AQP/GLP isoforms involved in glycerol and/or water redistribution during osmotic challenge.

The diverse physiological functions and osmotic requirements among different tissues result in differential expression of AQP isoforms (Borgnia et al., 1999). Using an antibody against EsAQP1 to determine its distribution at the tissue level in E. solidaginis, we found it was most abundant in the brain. In contrast, the few studies that have functionally characterized arthropod AQPs found a relatively low abundance in the brain or head (Ball et al., 2009; Shakesby et al., 2009). AQPs in mammalian brains are critically important in regulating water movement (Tait et al., 2008); AQP4 is the most abundant AQP in the brain, where it is thought to prevent edema by eliminating excess water from brain parenchyma (Papadopolous and Verkman, 2007). In a study comparing AQP4-null and wild-type mice, application of a frozen probe to the surface of the brain resulted in more severe edema in the AQP4-null mice (Papadopolous et al., 2004). The authors concluded that the freeze injury caused the blood–brain barrier to become leaky, thereby allowing water to accumulate around the brain. Because the AQP4-null mice were unable to void excess water through AQP4, they developed damagingly high intracranial pressure. Similar to mammals, insects possess a blood–brain barrier that prevents hemolymph from directly contacting their nervous tissue and disrupting the highly regulated ionic composition of the microenvironment surrounding the neurons (Carlson et al., 2000; Chapman, 1971). If the integrity of the insect blood–brain barrier is compromised through similar mechanisms during freezing, they too would experience an accumulation of water in their nervous tissue. Damage to the nervous system is a leading candidate to explain organismal death following freezing injury in E. solidaginis (Collins et al., 1997); therefore, we propose the abundance of EsAQP1 in brain protects by mediating rapid dissipation of water that accumulates during freezing.

As the fat body is one of the most cold-hardy tissues in E. solidaginis (Bennett and Lee, 1997), we did not expect EsAQP1 to be virtually absent from this tissue (Fig. 6B). Because the fat body of E. solidaginis is tolerant of intracellular freezing (Lee et al., 1993), and has an intrinsically low water content (i.e. <35%) (Davis and Lee, 2001), one could argue this tissue does not require EsAQP1 in high abundance for rapid redistribution of water during freezing. Although this is a plausible explanation, the results from previous studies suggest that AQPs and/or GLPs are present in fat body, as this tissue loses its ability to tolerate freezing when frozen in the presence of an AQP inhibitor (Izumi et al., 2006; Philip et al., 2008). Furthermore, fat body is the site of glycerol biosynthesis (Kukal et al., 1988), which likely necessitates a high abundance of GLPs. Therefore, we suggest that AQPs are important to the fat body in E. solidaginis; however, the active AQP isoform(s) is not EsAQP1.

Seasonal cold hardening of E. solidaginis occurs gradually through the autumn to prepare larvae for the cold and dry winter. During this period, larvae accumulate cryoprotectants (Morrissey and Baust, 1976), enter diapause (Irwin et al., 2001) and remodel the fatty acid composition of their membranes (Bennett et al., 1997). Because E. solidaginis AQPs promote freeze tolerance (Philip et al., 2008) and are implicated in seasonal cold hardening (Philip and Lee, 2010), we expected to observe an increase in the abundance of EsAQP1 in tissue from October to December. The abundance of EsAQP1 increased in all surveyed tissues during this cold-hardening period (Fig. 6B). By increasing the number of AQPs, cells likely improve their ability to rapidly redistribute water, and can better protect themselves against the build-up of osmotic pressures across the membrane during freezing in winter.

The results from the current study, in which we determined the sequence, phylogenetic relationship, osmotic permeability, tissue distribution and seasonal abundance of EsAQP1, complement our previous studies that link AQPs to the other mechanisms promoting freezing tolerance in E. solidaginis (Philip et al., 2008; Philip and Lee, 2010). Similar to natural freeze tolerance, successful clinical cryopreservation demands rapid water efflux and cryoprotectant influx during the freezing process. Although there have been a few attempts to enhance cryopreservation success by modifying AQP abundance (Edashige et al., 2003; Tanghe et al., 2002), this approach has not resulted in greater overall success in freezing tissues and organs. Therefore, studying natural systems in which the challenges of surviving freezing have been solved at the organismal level may uncover important details concerning AQPs/GLPs that could improve clinical cryopreservation in the future.