Abstract
For the past 50 years, it was believed that all bats, like humans and guinea pigs, did not synthesize vitamin C (Vc) because they lacked activity of L-gulonolactone oxidase (GULO) in their livers. Humans and guinea pigs lack the activity due to pseudogenization of GULO in their genomes, but there is no genetic evidence to show whether such loss in bats is caused by pseudogenization. Unexpectedly, our successful molecular cloning in one frugivorous bat (Rousettus leschenaultii) and one insectivorous bat (Hipposideros armiger) ascertains that no pseudogenization occurs in these species. Furthermore, we find normal GULO protein expression using bat-specific anti-GULO polyclonal antibodies in bats, evaluated by Western blotting. Most surprisingly, GULO activity assays reveal that these two bat species have retained the ability to synthesize Vc, but at low levels compared with the mouse. It is known that bats in the genus Pteropus have lost GULO activity. We then found that functional constraints acting on the GULO of Pteropus vampyrus (which lost its function) are relaxed. These results imply that the ability to synthesize Vc in bats has not been lost completely in species as previously thought. We also suggest that the evolution of bat GULO genes can be a good model to study genetic processes associated with loss-of-function.
Introduction
Vitamin C (Vc) (or L-ascorbic acid) is required by all organisms. Vc functions as an antioxidant to prevent tissues from oxidative damage and is necessary for a wide range of metabolic processes: a lack of Vc can lead to scurvy (Padh 1990). The ancestor of vertebrates was capable of synthesizing Vc de novo from glucose, and most mammals synthesize Vc in their livers, whereas fishes, amphibians, and reptiles synthesize it in their kidneys (Chatterjee 1973; Moreau and Dabrowski 1998). It is widely claimed that only a limited number of mammalian species, that is, the guinea pigs (Family Caviidae, Order Rodentia), primates in the suborder Haplorrhini (humans, apes, prosimian tarsiers, New World, and Old World monkeys), and bats (all families, Order Chiroptera) are defective in their ability to synthesize Vc, due to a lack of the activity of L-gulonolactone oxidase (GULO), the enzyme catalyzing the last step in the biosynthesis of Vc, in their livers (Burns et al. 1956; Burns 1957; Roy and Guha 1958a, 1958b; Chatterjee et al. 1961; Chatterjee 1973; Birney et al. 1976; Milton and Jenness 1987). Indeed, the enzyme could not be expressed in mammals incapable of Vc synthesis (Odumosu and Wilson 1973; Nishikimi and Udenfriend 1976), and hence, Vc must be obtained from exogenous sources such as fruits in these species (Pauling 1970).
Pseudogenization serves as an engine of evolutionary change, and it has been proposed that gene losses provide opportunities for phenotypic adaptations (Wang et al. 2006). During evolution, the gene encoding GULO has become a pseudogene in humans and guinea pigs that are defective in Vc synthesis (Nishikimi et al. 1992, 1994). Hence the GULO protein cannot be expressed in their livers, and de novo synthesis of Vc stops before the last step occurs. The consequence of pseudogenization is not lethal because these species are able to obtain enough Vc from their diets (Milton and Jenness 1987).
It has long been believed that bats, regardless of diets, are unable to synthesize Vc because they lack the activity of the GULO enzyme (Roy and Guha 1958b; Chatterjee et al. 1961; Chatterjee 1973; Birney et al. 1976; Milton and Jenness 1987). The loss of the capacity to synthesize Vc endogenously in humans, guinea pigs, and bats is frequently cited as an example where pseudogenization leads to loss-of-function because a functional pathway for Vc synthesis is unnecessary for these species and may indeed be costly to maintain (Chatterjee 1973; Deacon 2010). However, the bat species studied previously were nearly all from the New World, and many bat genera from the Old World (e.g., Rousettus, Cynopterus, Rhinolophus, and Hipposideros) were not included in analyses. In addition, bats show remarkable longevity for their body size (Brunet-Rossinni and Austad 2004), and a recent study found that Vc-binding genes had undergone increased selective pressure in long-lived mammals (Jobson et al. 2010). Furthermore, no direct evidence was provided that the loss of the Vc-synthesizing ability in bats was due to the pseudogenization of the GULO gene or the production of deficient GULO proteins. Therefore, we employed molecular cloning, together with protein expression and enzyme activity analyses, to reevaluate whether all bats have lost the ability to synthesize Vc and if so, to determine how they lost it.
Materials and Methods
Taxonomic Coverage
Bat liver tissue samples were kept at School of Life Sciences, East China Normal University, Shanghai, China, comprising six species (Rousettus leschenaultii, Cynopterus sphinx, Hipposideros armiger, Rhinolophus ferrumequinum, Myotis ricketti, and Scotophilus kuhlii). We also retrieved an additional seven mammalian sequences from GenBank (http://www.ncbi.nlm.nih.gov/Genbank/) for evolutionary analyses (Equus caballus [horse], XM_001492727; Bos taurus [cow], NM_001034043; Sus scrofa [pig], NM_001129948; Canis familiaris [dog], XM_543226; Mus musculus [mouse], NM_178747; Rattus norvegicus [rat], NM_022220; Monodelphis domestica [opossum], XM_001380006; Ornithorhynchus anatinus [platypus], XM_001521551). To expand our coverage of bat taxa, we explored the low-coverage megabat genome (2× Pteropus vampyrus) using search tool–BLAT (http://www.ensembl.org/Multi/blastview). We also screened other non-bat genomes (Macaca mulatta [macaque], 1.90× Spermophilus tridecemlineatus [squirrel], Pan troglodytes [chimpanzee], 2× Macropus eugenii [wallaby], Gorilla gorilla [gorilla], and Pongo pygmaeus [orangutan]) in the database. Basically, we used the individual 12 exons of mouse and bat GULO genes or the two pseudogenes [Cavia porcellus [guinea pig], D12762; Homo sapiens [human], OTTHUMT00000337557) for inquiries. We then searched the above genomes using individual exon sequences with search sensitivity set as “allow some local mismatch” and selected those matches with cutoff E-value below 1×10−3.
Reverse Transcription-Polymerase Chain Reaction and Sequencing
Total RNAs were isolated from bat liver tissue using RNAiso kit (TaKaRa, Japan) and reverse transcribed to first-strand cDNA with Invitrogen SuperScript III RT kit and oligo-dT primers (Invitrogen). We designed one pair of degenerated primers based on the published cDNA sequences for amplification of the complete coding region of GULO. Other two paired primers within the coding region of GULO were also designed for double checking. All primers are available upon request. Polymerase chain reaction products from cDNA were ligated into a pMD-19T vector (TaKaRa), and the positive clones were sequenced using a Big Dye Terminator kit on an ABI 3730 DNA sequencer (Applied Biosystems). All the new sequences are submitted to GenBank with accession numbers HQ415789 and HQ415790.
Production of Anti-Bat GULO Polyclonal Antibodies
Partial gene sequences (nucleotide positions 9–744) of R. leschenaultii GULO were amplified and subcloned into the bacterial expression vectors (pET28a (+); Novagen, Merck, Germany) that contained His-tag for purification and detection. The recombinant proteins achieved in BL21 bacteria after isopropyl beta-D-1-thiogalactopyranoside induction for 4 h were affinity purified by using AKTA purifier (GE Healthcare, UK). The purified proteins were used to immunize a rabbit three times, and antisera were collected as the anti-bat GULO polyclonal antibodies.
Sample Preparation, Protein Quantification, and Western Blotting
Liver tissue (0.1 g) of six species of bats, guinea pig (Cavia porcellus), human (Homo sapiens), and mouse (M. musculus) were collected and homogenized separately in 1 ml of lysis buffer (2.2% sodium dodecyl sulfate [SDS], 62.5 mM Tris-HCl, pH 6.8) with Precellys 24 (Bertin Technologies, France) four times using a program of 5500-1X15-005. The homogenates were boiled for 10 min at 100 °C after cooling on ice and following a quick spin; each homogenate was diluted 30-fold in ddH2O, and the total amounts of proteins in each sample were then determined with Quick Start Bradford protein assay kit (Bio-Rad). Equal amounts of total liver proteins (30 μg) from each species were loaded and separated on 12.5% SDS–polyacrylamide gel electrophoresis (PAGE) at 125 V using an SE 260 electrophoresis unit (GE Healthcare). After electrophoresis, the separated proteins on SDS-PAGE were transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore) in a WEALTEC semidry transfer system (Wealtec Corp.) with semidry transfer buffer (0.192 M Glycine, 25 mM Tris-base, 1.3 mM SDS, 200 ml methanol, and 800 ml ddH2O). Anti-bat-GULO antibodies were produced in this study as described above. The procedure for Western blotting was as previously described (Pan et al. 2006). Anti-GULO and anti-beta actin (ab8227; Abcam) antibodies were used as 1:2000 and 1:2500 dilutions, respectively. The transferred membrane was cut into half, the upper part of the membrane where molecular weight higher than 46 kDa was used for anti-GULO detection and the lower part of the membrane was used for beta-actin detection. Secondary antibodies, anti-rabbit immunoglobulin G-horseradish peroxidase (Santa Cruz Biotechnology), were diluted 5000-fold. Western blotting detection reagents (Amersham ECL Plus, GE healthcare) were used, and images on PVDF membrane were acquired by using a ChemiDoc XRS imager (Bio-Rad).
GULO Activity Assay
Methods for the assay of GULO activity were modified from previous studies (Moreau and Dabrowski 1998; Hasan et al. 2004). GULO activity was measured in each liver sample, including those from six bat species and the guinea pig, human, and mouse. Liver tissue (0.2 g) samples in 1.6 ml buffer containing 0.25 M sucrose and protease inhibitor cocktail tablets (Roche, Switzerland) were homogenized by using Precellys 24 (Bertin Technologies). Homogenates were centrifuged at 14,000 × g for 25 min at 4 °C. Pellets were discarded and supernatants were centrifuged at 100,000 × g for 1 h at 4 °C. The pellets (microsomes) were dissolved in 0.8 ml of 20 mM Tris-acetate (pH 8.0) containing 10 mM KCl, 1 mM ethylenediaminetetraacetic acid, and 0.3% (w/v) sodium deoxycholate by pipetting to release the enzyme into solution. The suspensions were then centrifuged at 14,000 × g for 30 min at 4 °C. The supernatant was used for the GULO activity assay. The blank control was reactant without microsome addition. The 5-ml reaction mixture contained 50 mM potassium phosphate buffer (pH 7.0), 50 mM sodium citrate, and 600 μl of solubilized microsomes (about 3 mg proteins). The enzymatic reaction was initiated with the addition of 20 mM L-gulonolactone (Sigma-Aldrich) as a substrate and incubated in aerobic conditions at 37 °C. Reactants were shaken at 180 revolutions per minute in a TS-2102 incubator (Chance International Croup Ltd, China) for 1 h and then centrifuged at 15,000 × g for 15 min at 4 °C. Total amounts of reduced and oxidized forms of ascorbate in the supernatants were determined with a Ferric Reducing Ascorbate Assay kit (BioVision). The absorbance of the final reaction products was measured at OD593 using a Synergy microplate reader (Biotec, UK).
Evolutionary Analyses
Sequences of bat and non-bat species were aligned using ClustalX (Thompson et al. 1997), with pseudogenes excluded and adjusted based on amino acid sequences. With this alignment, the phylogenetic tree was reconstructed by using MrBayes3.1 (Huelsenbeck and Ronquist 2001). For Bayesian inference, we evaluated the best-fit model as SYM+G using MrModeltest (Nylander 2004). The program was run with 400,000 generations with a burn-in of 25%.
Genus Pteropus has been reported to have lost the activity of GULO (Roy and Guha 1958b; Chatterjee et al. 1961; Chatterjee 1973; Birney et al. 1976; Milton and Jenness 1987). To investigate whether the GULO gene of P. vampyrus has undergone relaxed selection, compared with that of R. leschenaultii and H. armiger, we analyzed the selection pressures on GULO genes. We employed the codon-based method to estimate the ratio of nonsynonymous and synonymous substitutions using PAML4 (Yang 2007). Basically, we employed a free-ratio model to allow the ω (dN/dS) ratios (nonsynonymous to synonymous substitutions) to vary for each branch. We then used the likelihood ratio test (LRT) to evaluate whether this model fits the data significantly by comparing the likelihood values between free-ratio and one-ratio (one ratio for all branches) models. To further ascertain whether relaxed selection had occurred on the gene of P. vampyrus, we employed a two-ratio model to allow a difference in ω ratio between P. vampyrus and all other branches including R. leschenaultii, H. armiger, and non-bat species. We then employed another two-ratio model with ω fixed one for branch of P. vampyrus and different ω for the other branches. LRT was performed using likelihood parameters between the two models. If the GULO gene of P. vampyrus was under relaxed selection, the LRT should not have significance, suggesting that the ω ratio was not significantly lower than one. Ancestral reconstruction was conducted using the Bayesian method in PAML4 (Yang 2007).
Results
Molecular Cloning
To address whether bats have normal mRNA expression of GULO, we designed one pair of degenerate primers targeting the complete coding region of GULO. Unexpectedly, we successfully cloned this gene from two bat species—R. leschenaultii (Pteropodidae) and H. armiger (Rhinolophidae). Cloning for C. sphinx (Pteropodidae) and R. ferrumequinum (Rhinolophidae) failed even with other two paired primers.
Expression of GULO Protein in Bats
To evaluate the expression of GULO protein in bats, we prepared bat-specific anti-GULO antibodies. The results of anti-GULO blotting demonstrated clearly that the GULO protein was expressed in the mouse and in bats examined, but the signals of the protein band were not seen in the guinea pig and human (fig. 1), which are well known to have lost their ability to synthesize Vc. The signals of R. leschenaultii (Rl) and H. armiger (Ha) were of similar intensity as that of the mouse (Mm). However, the signals of C. sphinx (Cs) and R. ferrumequinum (Rf) were weaker than of the other species, and the former also had two distinct bands.
Expression patterns of GULO protein. Western blotting against the bat GULO and beta-actin antibodies are shown on the PVDF membrane. Protein bands of GULO are seen at a molecular weight around 50 kDa and those of beta-actin are at a size of 42 kDa. Species analyzed are guinea pig (Cavia porcellus, Cp), human (Homo sapiens, Hs), mouse (Mus musculus, Mm), and the bats Rousettus leschenaultii (Rl), Cynopterus sphinx (Cs), Hipposideros armiger (Ha), and Rhinolophus ferrumequinum (Rf).
GULO Activity Assay
Coincidentally, R. leschenaultii (Rl) and H. armiger (Ha)—the two bat species with normal protein expression—produced Vc, but C. sphinx (Cs) and R. ferrumequinum (Rf)—the two with weak or multiple-band protein expression—did not synthesize Vc, under the assay conditions (fig. 2). It was noteworthy that Rl and Ha, respectively, had 4-fold and 6-fold less Vc synthesis activity compared with the mouse. These results indicated that some species of bats have retained some ability for de novo synthesis of Vc, but the ability is low compared with the normal Vc-synthesizing mouse. We also constrained a species tree from published molecular phylogenies (Murphy et al. 2004; Teeling et al. 2005) to show clearly the evolution of Vc-synthesizing abilities in bats (fig. 3; supplementary table S1, Supplementary Material online).
GULO activity assay. All tests are carried out with more than three samples and three repeats for each sample. Y axis represents Vc synthesis (nmol) per assay volume (ml). Statistical significance (*P < 0.001) is determined by an analysis of variance test (Holm–Sidak method). Experimental data are expressed as means ± standard deviation. Species tested are guinea pig (Cavia porcellus, Cp), human (Homo sapiens, Hs), mouse (Mus musculus, Mm), and the bats Rousettus leschenaultii (Rl), Cynopterus sphinx (Cs), Hipposideros armiger (Ha), and Rhinolophus ferrumequinum (Rf).
Bat species tree that shows the GULO activities. Bat species are given as scientific names and non-bat species are given as common names. The green colors show the species known to have GULO activity, and the orange colors show the species that have lost GULO activity. The bat species assayed in this study are underlined. Human and guinea pig lineages are shown as dashed lines because of pseudogenization of their GULO genes. The maximum likelihood bootstrap support values on major nodes are retrieved from Teeling et al. (2005).
Evolutionary Analysis
We acquired most of the GULO exons (exons 3–8 and exons 11 and 12) of P. vampyrus (Supplementary fig. S1, Supplementary Material online) and did not find any indels (insertions/deletions) or premature stop codons in these exons. To infer the evolutionary history of bat GULO genes, we constructed a phylogeny using Bayesian inference. The tree obtained corresponded closely with phylogenetic relationships expected according to large-scale gene sequencing studies (Murphy et al. 2004; Teeling et al. 2005) (fig. 4). Regarding evolutionary rates, P. vampyrus had a higher ω (dN/dS) ratio (0.62), compared with the two bat species that retained Vc-synthesizing ability (0.05 and 0.16) (fig. 4) (P = 5 × 10−6, LRT [Yang 2007]), suggesting that relaxation of functional constraints could have shaped the evolution of GULO in this bat species. The two-ratio model also suggested that the foreground (P. vampyrus) had a significantly higher ω ratio (0.69) than background (all other branches, 0.06) (P = 2.9 × 10−5). The result of the LRT showed no significant difference in estimations using the two-ratio model in which ω for the branch of P. vampyrus is not fixed and fixed for 1 (P = 0.60). It suggests that the evolution of the GULO gene of P. vampyrus is close to be neutral. Interestingly, ω value for the branch of rat is quite high (0.53) (fig. 4). However, rat has normal Vc synthesis ability and there are not many amino acid changes that occur only in the rat lineage as in P. vampyrus (see fig. 5 and description below).
Evolutionary relationships of bat GULO genes. The tree is constructed using Bayesian inference. Bayesian posterior probabilities are given for each node. Values in the parentheses are dN, dS, and ω (dN/dS) ratios. For bat species, scientific names are given; for nonbat species, common names are used.
Alignment showing potential loss-of-function mutations. The asterisk means that a mutation has been retained in Pteropus (lost function) compared with all other species; the box means mutations have been specifically retained by all bats compared with non-bat species.
Mutations of P. vampyrus
After aligning most of the exons in GULO from P. vampyrus (which has lost GULO activity) with ten GULO genes from other mammals including two bat species—R. leschenaultii and H. armiger (which have retained lower activities) and the putative ancestral bat sequence inferred by the Bayesian method (see the Materials and Methods), we found that eight amino acid changes (positions 55, 66, 87, 92, 134, 170, 221, and 247, according to mouse GULO amino acid sequence) occurred only in P. vampyrus. Six of these sites (positions 55, 66, 87, 92, 134, and 170) belong to the flavin adenine dinucleotide (FAD)–binding region (the active center of the enzyme) (fig. 5). We noticed that mutations L55S (position 55), S134G, and S221F are related to polarity changes; H92L is related to a charge change. If a protein was expressed from this gene, the protein should be inactive due to mutations at these sites in the active region of the enzyme. Changes at two sites (positions 167 and 200) are unique to bats (P. vampyrus, R. leschenaultia, H. armiger, and ancestral bat), among which position 167 belongs to the FAD-binding region (fig. 5). Thus, the mutation of position 167 may be related to the low activity of bat GULO.
Discussion
Previous studies concluded that all bats had lost the ability to synthesize Vc by evaluation of GULO activity in species from 24 genera in 7 families (Pteropodidae, Noctilionidae, Mormoopidae, Phyllostomidae, Natalidae, Vespertilionidae, and Molossidae) (Roy and Guha 1958b; Chatterjee et al. 1961; Chatterjee 1973; Birney et al. 1976; Milton and Jenness 1987) (fig. 3). However, we observed that two bat species (R. leschenaultii and H. armiger) from the Old World have indeed retained GULO activity (fig. 2), and our observations are supported by molecular cloning and protein expression analyses. The failed cloning in C. sphinx and R. ferrumequinum suggested that genes in these bats may have trace levels of mRNA expression or that the genes in these two species are highly divergent and could not be amplified using the primers designed. In addition, there could be two alternative splicing forms in C. sphinx as there were two protein bands detected in this species (fig. 1). Nonetheless, normal gene expression was apparent in R. leschenaultii and H. armiger. Although expression was not obtained from P. vampyrus in this study due to unavailability of the sample tissue, we observed that this species possessed GULO in an intact, though modified form (fig. 5) in the genome. Evolutionary analysis suggested that the functional constraint of the gene in P. vampyrus is relaxed and it is close to be under neutral evolution (fig. 4), which seems to be related with its loss of GULO activity in this species. Mutations in the FAD-binding region could contribute to such loss of activity in P. vampyrus (fig. 5). We also observed some mutations in the FAD-binding region in R. leschenaultii and H. armiger and these could contribute to the lower levels of GULO activity in these species compared with the mouse.
Primates in the suborder Haplorrhini lost the ability to synthesize Vc at about the time (50–65 Ma) that the lineage split with the Strepsirrhini (Nakajima et al. 1969; Nishikimi et al. 1994). Most rodents (i.e., Myoprocta acouchy and Dasyprocta aguti, family Dasyproctidae) in the suborder Hystricomorpha appear to be capable of Vc synthesis (Yess and Hegste 1967), but the guinea pig (Cavia porcellus, family Caviidae) in the same suborder lost this ability during the past 20 million years (Nishikimi et al. 1992). Our screening of genome data verified that the GULO genes of the gorilla, macaque, chimpanzee, and orangutan have become pseudogenes (Supplementary fig. S2, Supplementary Material online), correlating with their loss of Vc synthesizing ability; whereas other species, such as the cow, dog, pig, horse, megabat (P. vampyrus), mouse, rat, opossum, and platypus have retained the intact genes, suggesting their maintenance of Vc-synthesizing abilities. For the squirrel and wallaby, several exons were located, and no indels or premature stop codons were found (Supplementary fig. S3, Supplementary Material online), suggesting that these taxa may have retained intact GULO genes too. By comparing the GULO genes of the guinea pig and human, it is apparent that these two species have gone through completely different pseudogenization processes: guinea pigs lack exon 5 and humans lack exons 2, 3, 6, 8, and 11 (Nishikimi et al. 1992, 1994). Even for their shared exons (exons 4, 7, 9, 10, and 12), guinea pigs and humans rarely share the same indels (deletions/insertions) or premature stop codons (Supplementary fig. S4, Supplementary Material online). These data suggest that the process of loss-of-function of the GULO gene of bats might have taken place differently, compared with humans and guinea pigs. Although the GULO genes in bats of which sequences were examined in this study are still intact, their reduced function may lead to them becoming pseudogenes in the future.
Ancestral bats should have possessed the ability to synthesize Vc, even though most of the species lost this ability over evolutionary time. Vc-deficient mammals can acquire Vc in their diets (e.g., fruit) (Pauling 1970; Milton and Jenness 1987). When the ability to synthesize Vc is not necessary, the GULO gene may degenerate via relaxed selection (Lahti et al. 2009). It was unexpected to find two species of frugivorous bats having different strategies to acquire Vc: whereas R. leschenaultii still synthesizes Vc, the closely related C. sphinx no longer does so. Both species eat similar fruits in southern China and show extensive dietary overlap in the rainy season, although R. leschenaultii migrates in the dry season when food is scarce (Tang et al. 2005). Two closely related insectivorous species—R. ferrumequinum and H. armiger—also exhibited different strategies to acquire Vc: the former has lost the ability to synthesize, whereas the latter still retains it at a relatively low level. We predict that R. leschenaultii and H. armiger should be in the process of pseudogenization of GULO and lose the ability to synthesize Vc in the future. Previous studies suggested that Vc-deficient species including frugivorous bats (genus Pteropus) have developed an alternative approach to synthesize Vc using dehydroascorbate (DHA), the oxidized form of Vc, which is transported by the Glut1 protein located in erythrocyte membranes to compensate for Vc deficiency (Rose 1988; Montel-Hagen et al. 2008). Investigations into whether the Vc-synthesizing bats still use Glut1 and DHA as a compensatory mechanism will be revealing. Frugivorous bats can easily absorb Vc from their diet. However, the mechanisms by which Vc can be obtained by insectivorous bats still remain unexplained at present.
In conclusion, although many bats have lost the ability to synthesize Vc as in humans and guinea pigs, the bat GULO genes are still intact. These bat genes can undergo transcription and translation, even though gene function is degenerating to different degrees, perhaps ultimately leading to pseudogenization. Thus, bat GULO genes are at the interface between functional genes and nonfunctional pseudogenes. Specifically, the GULO genes in R. leschenaultii and H. armiger are functional genes that express low levels of active proteins, but the equivalent genes in P. vampyrus, C. sphinx, and R. ferrumequinum are nonfunctional and may express inactive proteins. The results presented here enlighten the adaptive role played by gene loss for fitness and highlight how loss-of-function may evolve, potentially leading ultimately to pseudogenization.
We thank Dr Naoko Takezaki and two anonymous reviewers for valuable comments, thank Stephen J. Rossiter for technical advice, and Dong Xu, Jin Xing, and Lingjiang He for help in the laboratory. This study was supported by grants under the Key Construction Program of the National “985” Project and “211” Project to S.Z., the PhD Program Scholarship Fund of ECNU (2010044) to J.C., and a BBSRC China Partnering Award to G.J.
References
Author notes
These authors contributed equally to this study.
Associate editor: Naoko Takezaki
