The gene TaWOX5 overcomes genotype dependency in wheat genetic transformation

Although great progress has been achieved regarding wheat genetic transformation technology in the past decade1–3, genotype dependency, the most impactful factor in wheat genetic transformation, currently limits the capacity for wheat improvement by transgenic integration and genome-editing approaches. The application of regeneration-related genes during in vitro culture could potentially contribute to enhancement of plant transformation efficiency4–11. In the present study, we found that overexpression of the wheat gene TaWOX5 from the WUSCHEL family dramatically increases transformation efficiency with less genotype dependency than other methods. The expression of TaWOX5 in wheat calli prohibited neither shoot differentiation nor root development. Moreover, successfully transformed transgenic wheat plants can clearly be recognized based on a visible botanic phenotype, relatively wider flag leaves. Application of TaWOX5 improved wheat immature embryo transformation and regeneration. The use of TaWOX5 in improvement of transformation efficiency also showed promising results in Triticum monococcum, triticale, rye, barley and maize. Over-expressing TaWOX5 substantially increases the transformation efficiencies of wheat and other cereals, including barley and maize, with reduced genotype dependency, and transformed transgenic plants can readily be screened using a visible phenotype.

LEAFY COTYLEDON1 (LEC1 (ref. 5 )), LEAFY COTYLEDON2 (LEC2 (ref. 6 )), NiR 7 , BABY BOOM (BBM 8 ) and WUSCHEL (WUS 9,10 ). In particular, co-overexpression of genes BBM and WUS2 produced high transformation frequency in several previously transformation-recalcitrant inbred maize lines 11 . Moreover, mature maize embryos and leaf tissues were used to generate transgenic plants with these two morphogenic regulators, and co-overexpression of ZmBBM and ZmWUS2 clearly enhanced the transformation efficiency of certain recalcitrant genotypes of sorghum (Sorghum bicolor (L.) Moench), Indica rice and sugarcane (Saccharum officinarum L.). However, overexpression of WUS2 and BBM in cereal crops resulted in many negative effects including callus necrosis, compromised differentiation of shoots and roots, decreased fertility of transgenic plants and a variety of aberrant, stunted and twisted phenotypes 11 .
There are also reports of using the promoter of maize phospholipid transferase protein (PLTP) to drive BBM expression and an auxin-inducible promoter to drive WUS2 to increase transformation efficiency without resulting in aberrant phenotypes 12,13 . The PLTP promoter, incorporated with three viral enhancer elements, enhanced WUS2 expression and precluded the regeneration of cells with WUS2 integration, while strong transient expression of WUS2 stimulated somatic embryo formation in cells without WUS2 integration in the tissue, which was designated 'altruistic transformation' and can be used to obtain transgenic plants without WUS2 (ref. 14 ). Recently, a GRF4-GIF1 chimera construct was used to generate transgenic plants, which achieved an average transformation efficiency of 65% (with a range of 27-96%) in two tetraploid wheat varieties (Desert King and Kronos), even reaching 9-19% in two previously non-transformable common wheat varieties, Hahn and Cadenza 15 . In the present study, we report that overexpression of the wheat gene TaWOX5 dramatically improves the transformation frequency of wheat and five other cereal species, with less genotype dependence and no obvious negative effects on callus differentiation or plant phenotype.

Results and Discussion
The WUS gene is an important regulator of somatic embryogenesis in Arabidopsis 10,16 . Based on the Arabidopsis WUS sequence, two wheat homologous genes (TaWOX5 and TaWUS) were obtained. TaWOX5 is more closely related to Arabidopsis WOX5 (Extended Data Fig. 1) containing a WUS-related homeobox domain according to the description of the AtWUS protein structure 17 , which belongs to the WOX5 type in the WUS gene family and is specifically expressed in the root tip 18 Fig. 1). Furthermore, six different sequences corresponding to TaWOX5 were amplified by PCR from the common wheat line CB037 (Triticum aestivum, AABBDD, 2n = 42), and three TaWOX5 sequences were obtained fromTriticum monococcum accession CItrl3961 (AA, 2n = 14) and Aegilops speltoides accession PI554241 (SS, 2n = 14)), the two diploid species most closely related to common wheat (Supplementary Table 1 Fig. 1 | effects of TaWOX5 and TaWUS on transformation efficiency of selected wheat genotypes. a, Structure of the tDNA region on plasmids pWMB111, pWMB111-TaWOX5 and pWMB111-TaWUS. b, Transformation efficiency of Fielder and CB037 using vectors containing TaWOX5 and TaWUS, as well as control vector pWMB111. Data were tested using Student's t-test (two-sided), mean ± s.d. was plotted with all individual data points and exact P values are shown at the top of boxes. n (experimental (exp.) replicates) = 2 independent experiments. c, Transformation efficiency of Zhongmai895, Jimai22, Zhengmai9023, Lunxuan987 and Jing411 using vectors containing TaWOX5 and TaWUS-D, as well as control vector pWMB111. Data were tested using Student's t-test (two-sided), mean ± s.d. was plotted with all individual data points, exact P values are shown at the top of boxes and n (exp. replicates) is listed at the bottom of the figure. d, Transformation efficiency of 29 wheat varieties using vector containing TaWOX5 and control vector, in which different colours for selected varieties on the x axis represent different levels of P value significance (red, P < 0.001; blue, P < 0.05; black, no P value). Data were tested using Student's t-test (two-sided) and mean ± s.d. was plotted with all individual data points. Detailed information, including the number of embryos, exact P values and n, are shown in Supplementary Table 2. 0, transformation efficiency 0%; #, no data available.
The genomic sequences of TaWOX5 (MN412513), TaWUS-A, TaWUS-B and TaWUS-D were cloned into the pWMB111 expression vector under the control of the maize ZmUbi promoter and Nos terminator (Fig. 1a). Following the stable delivery of pWMB111-TaWOX5 into wheat genotypes Fielder and CB037 by Agrobacterium-mediated transformation (PureWheat), we detected transformation efficiencies significantly higher than those from the transformation experiments using the control (empty) pWMB111 vector (100% versus 50% efficiency, P < 0.05; Fig. 1b). Notably, transformation efficiencies obtained using the constructs for separate expression of TaWUS-A, TaWUS-B and TaWUS-D were lower than those using the pWMB111 control vector (Fig. 1b), because many of the regeneration shoots from those TaWUS transformation experiments did not produce roots and were therefore excluded from calculations regarding transformation efficiency. The shoots recovered from experiments using TaWOX5 were normal and developed healthy roots.
Over the past 5 years we have utilizeded TaWOX5 in the successful transformation of 29 common wheat varieties, including several known to be transformation recalcitrant ( Fig. 1d and Supplementary Table 2). Wheat variety Jimai22 is the most widely cultivated in China, with an annual planting area exceeding 2 million hectares, and the calli derived from Jimai22 immature embryos are very poor in quality. However, overexpression of TaWOX5 dramatically improved the quality of Jimai22 calli (Fig. 2). The transformation efficiency of this cultivar was significantly increased to 55.4 ± 17.9% from 5.8 ± 3.3% following the application of TaWOX5 (P < 0.001; Fig. 1d and Supplementary Table 2). Moreover, when TaWOX5 was not used in the transformation, wheat variety Ningchun4 did not regenerate green shoots, Jimai22 and Kenong199 generated only one to three green shoots and Fielder showed fewer than ten green shoots per immature embryo (Fig. 2b). When using TaWOX5, all tested varieties produced more than ten green shoots per immature embryo (Fig. 2d). Similarly, notable improvement in green shoot number was also observed in other wheat cultivars (Extended Data Fig. 2).

NATurE PlANTs
varieties Zhongmai895, Sumai3 and Jing411 (transformation efficiency <8% using a control vector) was significantly increased, to 82.7 ± 14.4%, 57.4 ± 13.0% and 17.5 ± 5.3% (P < 0.05 or 0.001), respectively. The use of TaWOX5 also significantly improved the transformation efficiency of previously non-transformable varieties including Bs366, Ningchun4, Aikang58, Xinong979 and Sunstate, to 83.5 ± 15.6%, 29.3 ± 13.6%, 21.8 ± 3.8%, 16.7 ± 3.1% and 9.1 ± 5.4% (P < 0.05 or 0.001), respectively. In addition, transgenic plants were also successfully generated from other important wheat varieties or germplasms including Zhengmai6694, Zhengmai9170, Zhongmai175, Cang6005 and Luohanmai (a landrace genotype) using TaWOX5, with efficiency ranging from 6.1% to 97.8%. In the course of our transformation work over the past few years we have used TaWOX5 in the transformation of 3,290 Fielder immature embryos and 5,459 Jimai22 immature embryos, the numbers of immature embryos used in these experiments being 1,000 greater than for each genotype among the majority of commercial wheat varieties examined in our study (Supplementary Table 2). Note that genotyping of transgenic plants overexpressing TaWOX5 was variously performed using PCR, droplet digital PCR (ddPCR) and QuickStix analyses (Extended Data Fig. 3 and Supplementary Table 3). The data from ddPCR showed that transgenic wheat plants recovered in TaWOX5 experiments contained one or more copies of the transgene (Supplementary Table 3). TaWOX5 was inherited by the T 1 generation based on Mendelian pattern (Supplementary Table 4). Collectively, our data with these large sample sizes and diverse common wheat genotypes demonstrate that the application of TaWOX5 greatly improves transformation efficiency.
It is well known that Agrobacterium-mediated plant transformation depends on both the infection efficiency of Agrobacterium and the regeneration ability of host cells. The application of TaWOX5 was able to enhance the ability of host tissues in callus induction and regeneration. In this case, the infection efficiency of Agrobacterium will determine the final transformation efficiency of wheat tissues. According to our evaluation in a previous study 19 , the ranked order of regeneration capacity for five widely cultivated Chinese commercial wheat varieties was Zhongmai895, Ningchun4, Aikang58, Xinong979 and Jimai22. However, the transformation efficiency of Jimai22 using TaWOX5 (55.4 ± 17.9%) was higher than that for Aikang58, Xinong979 and Ningchun4 (Supplementary Table 2). To explore this phenomenon, we introduced maize genes ZmR and ZmC1, involved in anthocyanin biosynthesis 20 , into three groups of wheat genotypes of varying transformation efficiency: a 'high' group with an efficiency of 80-100% (Fielder and Zhongmai895), a 'medium' group (40-80%, Jimai22) and a 'low' group (0-50%, Ningchun4, Aikang58 and Sunstate). When calculating transient transformation efficiency based on visualization of areas where anthocyanin expression exceeded 50% of the entire immature embryo, efficiency order from high to low was Fielder (87%), Zhongmai895 (73%), Jimai22 (54%), Ningchun4 (14%), Aikang58 (4%) and Sunstate (0%) (Extended Data Fig. 4), which is consistent with the stable transformation efficiency of these genotypes using TaWOX5 (Fig. 1d and Supplementary Table 2). While regeneration capacity remains an underlying requirement for successful transformation of all tested varieties, it appears that transient transformation efficiency is roughly correlated with final transformation efficiency when TaWOX5 is used.

NATurE PlANTs
The physiological status of wheat immature embryos greatly affects successful transformation. High temperature during the growth period of wheat mother plants, especially at the grain filling stage, negatively affects production and differentiation of embryonic calli derived from immature embryos 21 , which can often result in transformation failure 21,22 . In standard PureWheat methods, the physiological status of wheat immature embryos is also addressed to affect regeneration ability and transformation efficiency. When wheat mother plants are subjected to biotic or abiotic stress, the physiological state of immature embryos deteriorates. Typically, immature embryos from stressed wheat mother plants cannot generate embryonic calli 21 . There were no good-quality calli or transgenic shoots produced from the transformed immature embryos of CB037 and Kenong199 plants showing early senescence of the flag leaf in the transformation experiments with a control vector (Fig. 3a). However, when TaWOX5 was used in transformation, immature embryos from stressed CB037 plants produced brown calli and generated between one and three green shoots per callus (Fig. 3b,c) while those from stressed Kenong199 plants generated between five and ten shoots per callus. These results demonstrate that the application of TaWOX5 can enable the production by wheat immature embryos of poor status of transgenic plants with efficiency as high as 33.8% for CB037.
TaWOX5 has merits in the promotion of transformation efficiency. First, it dramatically increased the transformation efficiency of wheat varieties/lines to a high level in the present study. The transformation efficiency of selected wheat varieties, including Fielder, CB037, Zhoumai18 and Zhengmai6694, was as high as 100% and 90% in small-and large-scale studies, respectively, when using TaWOX5 in many repeated experiments conducted over the past 5 years (Supplementary Table 2). Second, wheat varieties/ lines previously recalcitrant to transformation, including Dwarfing Polish, Bs366, Aikang58, Sunstate, Ningchun4 and Xinong979, were readily transformed with TaWOX5. Third, TaWOX5 proved efficient in the transformation of T. monococcum, barley, rye, triticale and maize. Fourth, transgenic regeneration shoots TaWOX5 has considerable potential applications in genetic transformation and genome editing for cereal crops. We incorporated TaWOX5 into our vector containing a double tDNA region (pWMB248) and the CRISPR-associated protein 9 (Cas9) expression cassette (pWMB110-Cas9), in which TaWOX5 was linked with the Bar selection marker or Cas9, and TaWOX5 can be removed together with Bar or Cas9 in the progenies of transgenic or edited plants. A total of 51 and 33 wheat mutant plants for the TaQ gene were confirmed in 112 and 75 T 0 plants from Fielder and Jimai22, respectively, by PCR-restriction enzyme (PCR-RE). The editing efficiency of TaQ in Fielder was 45.5% with a control vector, and in Jimai22 was 44.0% with a TaWOX5 vector. Our results demonstrated that TaWOX5 is useful in the recovery of Cas9-edited events. Even though T 0 plants did not grow to maturity, we can assume that the tDNA locus would readily segregate from the edited locus in the next generation. Although the application of TaWOX5 cannot directly contribute to frequency improvement in the generation of marker-free or Cas9 cleavage plants, it is useful in regard to exclusion of most non-targeted candidate plants in segregating generations for obtaining marker-free or mutant plants, which can help to reduce workload by observing the TaWOX5 phenotype. For this workflow, we first selected plants without the phenotype of the wide flag leaf in the T 1 or T 2 generation then identified them for the absence of Bar and the presence of a target gene or edited sequence by PCR or PCR-RE and sequencing.

Conclusions
In summary, 31 common wheat cultivars were successfully transformed and developed into transgenic plants using TaWOX5. Overexpression of TaWOX5 also notablely increased the transformation efficiency of T. monococcum, triticale, rye, barley and maize. The application of TaWOX5 can enhance the efficiency of the genetic transformation and genome editing of wheat and other crops and improve cost effectiveness by improving plant regeneration, reducing the requirement for embryo quality and identifying marker-free transgenic or transgene-free edited plants based on the visible botanic phenotypes associated with TaWOX5 overexpression. Cloning of TaWOX5. The coding sequence of Arabidopsis WUS (AJ012310) in NCBI was used as a query to search the homologous gene in common wheat (T. aestivum) via tblastn, resulting in only one sequence, FN564431.1. Based on this sequence, primer pair TaWOX5F (5′-GTGTCAATGGAGGCGCTGAGCG-3′) and TaWOX5R (5′-ATGCGTGCGTGCGACGTTGATT-3′) was designed to amplify TaWOX5 from the genomic DNA of wheat line CB037, T. monococcum accession CItrl3961 and Ae. speltoide accession PI554241.

Plant materials and cultivation conditions. Common wheat cultivars/lines
Because TaWOX5 is not an AtWUS orthologue, the protein sequence of AtWUS was used as a query for tblastn in IWGSC, with three contigs obtained. According to contig sequences, a pair of specific primers (5′-ATGGACAAG CAGAGCGTC-3′ and 5′-TAGGACAATGACGGGAGCACT-3′) was designed to amplify the sequence, designated as TaWUS.
Vector construction. The primers CB1SmaF: 5′-AAACCCGGGATGGAG GCGCTGAGCGG-3′ and CB1KpnR: 5′-AAAGGTACCTTAGACCAGATAC CGAT-3′ were used to perform PCR amplification using pMD-18T-TaWOX5 as a template with a high-fidelity enzyme KOD (Toyobo, KOD-401). The PCR product and pWMB003 vector (containing ZmUbi promoter and Nos terminator) were then digested with KpnI and SmaI to obtain a 773-base-pair (bp) product and a 4,535-bp vector backbone. Next, the target PCR product and vector backbone were ligated to generate intermediate expression vector pWMB003-TaWOX5. Vectors pWMB003-TaWOX5 and pWMB111 (containing a Bar expression cassette controlled by ZmUbi promoter MZ458107) were digested with HindIII to produce a 3,033-bp TaWOX5 expression cassette and a 10,170-bp vector backbone, respectively. Finally, the two enzyme-digested products were ligated to generate the target expression vector pWMB111-TaWOX5 (Extended Data Fig. 6a) for the transformation of wheat, T. monococcum, rye, triticale and barley. Vector pWMB202, containing anthocyanin biosynthesis-related genes ZmR and ZmC1 (ref. 20 ), was used to detect transient transformation efficiency in different wheat genotypes. The TaWOX5 expression cassette was inserted into the vector pLC41 (GenBank accession no. LC215698.1) containing a Bar expression frame controlled by a 35S promoter and Nos terminator for transformation of wheat (varieties Chinese Spring and Norin61) and maize (inbred lines B73 and A188).
Vectors pWMB111-TaWOX5 and pWMB202 were introduced into Agrobacterium strain C58C1, and vector pLC41-TaWOX5 was introduced into Agrobacterium strains EHA105 and LBA4404 carrying pVGW9, for wheat and maize transformation, respectively. pVGW9 (SEQ ID 1 US 10266835 B2) was a helper plasmid for plant transformation and contained virB, virC, virD, virG and virJ from plasmid pTiBo542 (GenBank accession no. NC_010929.1). The plasmids pLC41-TaWOX5 and pVGW9 can be accessed for academic purposes with an MTA. According to our previously published methods 23 , the single-guide RNA of TaQ controlled by promoter TaU3 was constructed into vector pWMB110-SpCas9 containing SpCas9 and Bar genes driven by promoters ZmUbi and 35S, respectively, to generate vector SpCas9-TaQ for editing TaQ in the wheat cultivar Fielder. The whole expression cassette of TaWOX5 (ZmUbi-TaWOX5-NOS) was amplified from plasmid pWMB003-TaWOX5 and inserted onto vector pWMB110-SpCas9 to generate vector TaWOX5-SpCas9. The sgRNA of TaQ was constructed into TaWOX5-SpCas9 to generate vector TaWOX5-SpCas9-TaQ (Extended Data Fig. 6b) for editing TaQ in Jimai22.

Plant transformation.
Wheat. Wheat spikes were sampled at 14 days post anthesis (DPA) and immature grains were carefully collected. Under aseptic conditions, grains were surface sterilized with 70% ethanol for 1 min and 5% sodium hypochlorite for 15 min and rinsed five times with sterile water. Fresh immature embryos were isolated and underwent Agrobacterium-mediated transformation to obtain transgenic plants following the protocol described by Ishida et al. 1 , with slight modifications. In brief, immature embryos were incubated with Agrobacterium for 5 min in cocultivation WLS 1 liquid medium (1/10 Linsmaier and Skoog (LS) salts, 1/10 Murashige and Skoog (MS) vitamins, glucose 10 g l −1 , 2-(N-morpholino) ethanesulfonic acid (MES) 0.5 g l −1 and acetosyringone (AS) 100 μM, pH 5.8) at room temperature, and cocultivated for 2 days on cocultivation medium (WLS liquid medium plus AgNO 3 0.85 mg l −1 , CuSO 4 ⋅5H 2 O 1.25 mg l −1 and agarose 8 g l −1 ), with the scutellum facing upwards, at 25 °C under darkness. After cocultivation, embryonic axes were removed with a scalpel and remaining scutella were transferred onto plates containing callus induction medium (LS salts, MS vitamins, 2,4-d 0.5 mg l −1 , picloram 2.2 mg l −1 , AgNO 3 0.85 mg l −1 , ascorbic acid 100 mg l −1 , carbenicillin 250 mg l −1 , cefotaxime 100 mg l −1 , MES 1.95 g l −1 and agarose 5 g l −1 ) for delay culture for 5 days under the same conditions. Afterwards, tissues were cultured on selection medium (callus induction medium plus phosphinothricin (PPT, Sigma, no. 45520) 5 mg l −1 without cefotaxime) for further callus induction. Two weeks later, calli were placed on selection medium containing PPT 10 mg l −1 for 3 weeks for embryonic callus induction under darkness. Embryonic calli were then differentiated on 1/2 MS medium containing Letters NATurE PlANTs PPT 5 mg l −1 without zeatin (other than LSZ-P5 (ref. 1 ) medium in PureWheat methods containing zeatin) at 25 °C under 100 μmol m -2 s -1 light. Regenerated shoots were transferred into cups filled with rooting medium plus PPT 5 mg l −1 for elongation and root formation. Plantlets with well-developed root systems were transplanted into pots and cultivated in a growth chamber. Transformation of T. monococcum, rye and triticale was performed by the same methods used for wheat.
Maize. Maize transformation was performed following previously published protocols 25 . Between 8 and 15 DPA, maize spikes containing immature embryos were sampled. Immature embryos were isolated and infected by Agrobacterium for 5 min. Infected embryos were then transferred to LS-AS solid medium (LS medium plus 2,4-d 1.5 mg l -1 , sucrose 20 g l −1 , glucose 10 g l −1 , proline 0.7 g l −1 , MES 0.5 g l −1 , AgNO3 0.85 mg l -1 , CuSO 4 ⋅5H 2 O 1.25 mg l -1 , AS 100 μM, and agarose 8 g l −1 ) with scutella facing upwards, and incubated under darkness at 25 °C for 7 days. After cocultivation steps, embryos were transferred to the first selection medium (LS medium with modified vitamins 25 plus 2,4-d 1.5 mg l −1 , sucrose 20 g l −1 , proline 0.7 g l −1 , carbenicillin 250 mg l −1 , cefotaxime 100 mg l −1 , PPT 5 mg l −1 , MES 0.5 g l −1 , AgNO 3 1.7 mg l -1 and agar 8 g l −1 ) for 10 days, and then moved to the second selection medium (first selection medium plus PPT 10 mg l −1 ) for culture for 21 days. Calli were cut into pieces of diameter 3-5 mm and transferred onto a second, fresh selection medium. After culture for a further 21 days, embryonic calli were cut into pieces of diameter 2-3 mm and transferred to LSZ medium (LS medium with modified vitamins 25  ddPCR. Total genomic DNA was extracted following a standard CTAB method 26 . A single-copy wheat TaWaxy gene on genome D was used as reference gene for ddPCR, and its specific primers (TaWaxy-DF: 5′-GCCACGTCGAAGA AGGAGATC-3′ and TaWaxy-DR: 5′-GACAGGTTCAGCCGGTATGTG-3′) and probe (TaWaxy-DProbe: VIC-CCTGCACTGTTGCTCGC CGCT-BHQ) were employed 27 . The primer pairs (BarF: 5′-TCGTCAACCACTACATCGAGACA-3′ and BarR: 5′-GTCCACTCCTGCGGTTCCT-3′) and probe (BarProbe: FAM-ACTTCCGTACCGAGCCG-MGB) for the Bar gene were designed to detect the transgene copy number in transgenic wheat plants. Duplexing a TaWaxy-D primer pair with a Bar primer pair increased the accuracy of ddPCR assay by enabling normalization and direct amplicon comparison within and across samples. Each reaction mixture (20 µl) consisted of 10 µl of ddPCR TM supermix (Bio-Rad, no. 1863010), 0.9 µM primers, 250 nM probe and 200 ng of DNA. The mixtures and droplet generation oil (Bio-Rad, no. 1863005) were added in cartridges and loaded into a QX200 droplet generator (Bio-Rad) for droplet generation. Droplet emulsions were transferred to a 96-well PCR plate and sealed with a foil heat seal at 180 °C for 10 s. PCR amplification was run at 95 °C for 10 min, 40 cycles of 94 °C for 30 s and 54 °C for 40 s, with a final cycle of 98 °C for 10 min. A QX200 Droplet Reader (Bio-Rad) was used for automatic measurement of the fluorescence signal of each droplet. Quantification of target DNA was calculated using QuantaSoft v.1.7.4.0917 (Bio-Rad). A no-template negative control was adopted in all ddPCR assays. The test for each sample was performed in duplicate.

Statistical analysis.
Transformation efficiency is expressed as mean ± s.d., and data were analysed using SPSS 17.0. All data were tested using Student's t-test (two-sided), in which P < 0.05 was considered statistically significant.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Accession numbers and gene names are available from the phylogenetic tree in Extended Data Fig. 1 Tables 2 and 3. Transgenic lines and plasmids generated are available from the corresponding authors on request. Source data are provided with this paper.