Plant phosphomannose isomerase as a selectable marker for rice transformation

The E. coli phosphomannose isomerase (EcPMI) gene is widely used as a selectable marker gene (SMG) in mannose (Man) selection-based plant transformation. Although some plant species exhibit significant PMI activity and active PMIs were even identified in Man-sensitive plants, whether plant PMIs can be used as SMGs remains unclear. In this study, we isolated four novel PMI genes from Chlorella variabilis and Oryza sativa. Their isoenzymatic activities were examined in vitro and compared with that of EcPMI. The active plant PMIs were separately constructed into binary vectors as SMGs and then transformed into rice via Agrobacterium. In both Indica and Japonica subspecies, our results indicated that the plant PMIs could select and produce transgenic plants in a pattern similar to that of EcPMI. The transgenic plants exhibited an accumulation of plant PMI transcripts and enhancement of the in vivo PMI activity. Furthermore, a gene of interest was successfully transformed into rice using the plant PMIs as SMGs. Thus, novel SMGs for Man selection were isolated from plants, and our analysis suggested that PMIs encoding active enzymes might be common in plants and could potentially be used as appropriate genetic elements in cisgenesis engineering.

During plant transformation, transformed cells are identified from a matrix consisting primarily of untransformed cells using a selection system and then regenerated to transgenic plants. Numerous selection systems have been developed in the past 20 years; nevertheless, only a few of them have been practically used 1,2 . The most-used selection systems consist of an antibiotic or herbicide selective agent, which is toxic for the non-transformed cells, and the corresponding selectable marker gene (SMG) confers the tolerance to transformed cells 3 . These systems are stable and efficient in both scientific research and commercial breeding. However, relevant safety concerns have also been continually raised by the public and by regulatory committees in certain countries. One concern is that the use of toxic selection agents in factories and fields has potential risks to the environment. Another concern is the safety risks of SMGs. Because bacteria can import and integrate exogenous DNA 4,5 , the natural bacterial population may be exposed to recombinant DNA, including SMG fragments, during the cultivation and consumption of genetically modified (GM) crops and therefore lead to the possibility of unintended horizontal gene transfer (HGT) 6,7 . HGT may spread antibiotic resistance from the SMGs of GM plants to soil or gastrointestinal tract microbes, which may compromise the clinical effectiveness of antibiotic drugs. Another risk is caused by the gene flow between GM plants and their wild relatives 8 . Because SMGs do not naturally exist in plants, the spread of novel antibiotic and herbicide resistance traits may lead to the extinction of wild populations 9 or to increased fitness and persistence of wild relatives, potentially making them "super-weeds" 10 . Moreover, antibiotic or herbicide SMGs are generally isolated from bacteria 3 and thus are cited as "unnatural". Although the biological safety of SMGs for human and animal health has been validated by several independent risk-assessment tests [11][12][13] , their bacterial origins are easily associated with potential danger, causing rejection by consumers.
Rice is the most preferred staple food in Asia and a major target for plant genetic engineering. It has been more than a quarter century since the successful genetic transformation of rice. Countless studies have revealed that transgenes can effectively improve the yield, development, nutrition, stress resistance and other agronomic traits of cultivated rice varieties. However, even under strict monitoring, few types of GM rice have been commercially released to date 14 . Safety concerns regarding SMGs are major deterrents to taking advantage of the advancement of biotechnology in rice breeding. Several approaches, including co-transformation systems, recombination-mediated site-specific excision or integration systems, and even CRISPR-based targeted degradation systems, have been designed and/or applied to generate marker-free transgenic plants [15][16][17][18] . However, these approaches normally require the screening of a large number of transformation events or extra transformation steps, which would greatly increase labor and time costs. One alternative strategy is the application of a non-antibiotic and non-herbicide selection system during transformation. A phosphomannose isomerase (PMI)/ mannose (Man) combination is one of the most popular and highly used non-antibiotic systems in rice. Man itself is not toxic to plants; however, Man is phosphorylated to mannose-6-phosphate (M-6-P) by endogenous hexokinases. The accumulation of M-6-P in plants blocks glycolysis 19 , inhibits ATP production 19,20 , and represses photosynthetic gene transcription 21 . Consequently, plant cells cultivated in Man media grow much slower, and organogenesis and germination are strongly restrained 22,23 . PMI can convert M-6-P into fructose-6-phosphate (F-6-P), which is then metabolized via glycolysis in plants. Therefore, only cells expressing PMI can grow and be selected on medium that uses Man as a carbon source, which enables the use of PMI as a SMG in plant transgenes.
PMIs are widely found in bacteria, fungal and animal cells; however, PMIs were commonly considered nonexistent in higher plants, except in some soybean species 24,25 . Recently, a few active PMIs were identified in Arabidopsis and Chinese cabbage 26,27 , suggesting that functional PMI may exist in higher plants. The existence of PMI isoforms in plant genomes consequently raises a question: Are the higher plant-originated PMI genes sufficiently active to be SMGs for Man selection of plant transgenes? Here, we report that four novel PMIs were isolated from algae and rice. The corresponding proteins were expressed and purified from E. coli, and their enzymatic activities were identified. Furthermore, the active plant PMI genes were expressed in cells to select and regenerate transgenic rice under Man pressure. Our results suggested that plant PMIs could be applied as effective alternative SMGs in the genetic engineering of rice and other plants.

Results
Cloning of the plant PMIs and characterization of the sequences. The genomic sequence databases of Chlorella variabilis and Oryza sativa were searched to identify plant PMIs. Four putative PMIs were found, including a gene from Chlorella (CHLNCDRAFT_139231, designated as CvPMI) and three genes from rice (Os01g0127900, Os09g0389000 and Os11g0600900, designated as OsPMI1, OsPMI2 and OsPMI3, respectively). The coding sequences of the PMIs were cloned, and the predicted molecular masses for CvPMI, OsPMI1, OsPMI2 and OsPMI3 were 45.85, 46.93, 44.18 and 24.15 kDa, respectively. The alignment of amino acid sequences revealed that the sequences of these plant PMIs were highly similar to the classic type I PMIs of C. albicans, E. coli and humans (Fig. S1). As indicated in Fig. S1, all plant PMIs except OsPMI3 contained the eukaryotic PMI consensus sequence YXDXNHKPE in their primary structures 28 . Type I PMI is a zinc-dependent metalloenzyme. Four amino acids of C. albicans PMI (CaPMI) (Gln-111, Glu-138, His-113 and His-285) are essential for the ligand of zinc 29 . These residues were also conserved in the sequences of the plant PMIs except OsPMI3 (Fig. S1). Furthermore, the three-dimensional (3D) structures of the PMIs were predicted. Except for OsPMI3, these plant PMIs were likely to exhibit a structure highly similar to that of CaPMI (Fig. S2), suggesting that most plant PMIs may have the same biochemical activity as the typical type I PMI.

Identification of plant PMI isoenzymatic activity.
To determine whether plant PMIs have the ability to isomerize M-6-P and to estimate whether their activities are sufficient to be used as potential SMGs, the four PMIs were fused with a glutathione S-transferase (GST) tag at the N-terminal and expressed in E. coli cells. Two previously identified PMIs, EcPMI from E. coli and AtPMI2 from Arabidopsis, were expressed as controls. Five GST-PMI fusion proteins were purified from the soluble fraction of bacterium to homogeneity using affinity purification. Although many expression conditions had been tested, GST-OsPMI2 was not soluble. Therefore, the protein was denatured in inclusion bodies and subsequently recovered and purified. The purity of the recombined proteins was assessed via sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis (Fig. 1A).
To quantitatively evaluate the isoenzymatic activity in vitro, the values of Km and Vmax for the M-6-P to F-6-P isomerization reaction catalyzed by the expressed fusion proteins were measured by the phosphoglucose isomerase (PGI)/D-glucose-6-phosphate dehydrogenase (G6PDH)-coupled enzyme method 26 . As expected, the GST tag itself did not convert M-6-P (Table 1). Accordingly, the activity of the fusion proteins should be contributed by the corresponding PMIs. As shown in Fig. 1B and Table 1, GST-CvPMI had the highest Vmax (68.03 ± 2.22 μ mol min −1 mg −1 protein) among the plant PMIs, which was comparable with that of EcPMI (74.63 ± 5.34 μ mol min −1 mg −1 protein). However, the Km value of CvPMI was greater than that of EcPMI, suggesting lower substrate affinity of the recombinant CvPMI proteins. The Km and Vmax of GST-AtPMI2 were 4104.67 ± 53.47 μ M and 33.33 ± 2.42 μ mol min −1 mg −1 protein, respectively. In a previous study, the Km and Vmax parameters of the histidine-tagged AtPMI2 were determined to be 372 ± 13 μ M and 22.5 μ mol min −1 mg −1 protein, respectively 26 . The differences between the kinetic parameters may be due to the different environments of the enzymatic assays and/or to interference from the protein tags. The parameters of OsPMI1 were similar to those of AtPMI2, implying that they may have similar biochemical activity. Moreover, the activities of OsPMI2 were much lower than were those of OsPMI1 possibly due to denaturation during the protein purification process. Furthermore, the activities of OsPMI3 were not detected, which is consistent with the lack of the conserved type I PMI motif in its sequence.
Transformation of Japonica and Indica rice using three plant PMIs as SMGs. To assess whether plant PMIs could potentially function as SMGs for rice transformation, the three highly active PMIs (CvPMI, OsPMI1 and AtPMI2) were inserted into a pCAMBIA1381 binary vector under the control of the CaMV 35S promoter (35S) to replace the hygromycin phosphotransferase (HPT) SMG. A similar vector using EcPMI substituted  for HPT was constructed in parallel and used as the positive control. Previously, we optimized a high-throughput Agrobacterium-mediated rice transformation system for EcPMI selection 30 . In this study, the selection abilities of the three plant PMIs were initially examined in the Japonica cultivar Nipponbare following the same protocol. Similar to the effect of EcPMI, resistant calli grew from the calli transformed with the CvPMI, OsPMI1 and AtPMI2 vectors under selection with 12.5 g/L Man plus 5 g/L sucrose ( Fig. 2A-D). The selection rates of CvPMI were comparable with those of EcPMI and were slightly higher than the selection rates of OsPMI1 and AtPMI2 ( Table 2). The selection of the PMIs was also tested in an Indica variety, Kasalath. Following the Indica transformation protocol of EcPMI constructs, 7.5 g/L Man plus 15 g/L maltose were added to the medium as selection pressure 31 . Similar to the results in Nipponbare, the resistant calli were developed after being transformed with the PMIs, and the selection rates of CvPMI were even higher compared with those of EcPMI (Table 2). These results preliminarily suggested that the plant PMIs have the ability to be used as SMGs in rice transformation. Next, the resistant calli selected by the PMI SMGs were tested for regeneration. In both Nipponbare and Kasalath, plantlets were obtained (Fig. 2E-H), and the plant PMIs exhibit an efficiency comparable to that of EcPMI (Tables 2 and S1-S2). Then, a regenerated plantlet was transformed in rooting medium (Fig. 2I) and confirmed by sequence-specific PCR analyses. As indicated in Tables 2 and S1-S2, CvPMI provided transformation frequencies (TFs) similar to those of EcPMI in Nipponbare and much higher than those in Kasalath. Although AtPMI2 and OsPMI1 were less efficient than EcPMI in Nipponbare, their TFs were close to that of EcPMI in Kasalath. These results suggested that transgenic rice could be efficiently generated using the plant PMIs in a manner similar to EcPMI. Furthermore, the low-copy rates in transgenic events were investigated by a TaqMan assay of the 35S promoter. The construct with a reduced low-copy rate was believed to produce insufficient PMI activity and was not recommended for rice transformation 31 . However, low-copy rates varied significantly between independent experiments (Tables S1-S2), suggesting that the rates may also be related to reasons other than the constructs or SMG activity. The low-copy transgenic plants were transferred into soil, and no significant abnormality in growth or development was observed in the transgenic population of Nipponbare or Kasalath (Figs 2J and S3). The seeds of the T 0 plants were germinated on Man selection medium to further examine the selection effect of the plant PMIs. The germination of untransformed rice seeds was completely inhibited on the MS medium plus 12.5 g/L Man. By contrast, most seeds of the plant PMI transgenic lines could germinate and grow into plants (Fig. S4). These data further confirmed that the plant PMIs could be used to select the positive transgenic lines in rice transformation.

Optimization of regeneration medium for plant PMIs. The escape of untransformed cells under Man
selection frequently occurs when using the EcPMI SMG, as noted in previous reports, as well as when using the plant PMI SMGs in this study. One possible reason might be reduced Man pressure in the regeneration medium. Therefore, we tested the escape rate and regeneration rate of the three plant PMIs and EcPMI under different Man concentrations in the corresponding Nipponbare transformation. In general, the regeneration rates decreased as Man pressure increased, and escape was remarkably avoided by Man supplementation of the regeneration medium, as expected (Tables 3 and S3). In particular, the regeneration rates of the AtPMI2 and OsPMI1 transformants were similar to or slightly lower than that of EcPMI. Their transformants produced few plants or did not produce any plant under additional selection pressure in the regeneration medium (Table 3). However, the regeneration rate of the CvPMI transformants was much higher under Man pressure compared with EcPMI, and approximately 12.5% of plants could still be produced even in medium using Man as sole carbon source ( Table 3).

Expression of PMIs in the corresponding transgenic plants.
To further verify the transgenic plants selected by the PMI/Man system, reverse-transcription PCR (RT-PCR) was performed to examine the expression of the PMIs in the corresponding transgenic lines of Nipponbare. According to the semi-quantitative RT-PCR results, the exogenous CvPMI and AtPMI2 genes were expressed in the corresponding regenerated plants, and the abundance of the OsPMI1 transcript greatly increased (Fig. 3A). In addition, the PMI activities of the plants were examined in six independent low-copy transgenic lines of each construct and untransformed plants. As expected, the PMI activities of the proteins isolated from transgenic plants were enhanced (Fig. 3B) Table 2. Comparison of rice transformation using EcPMI and plant PMIs as SMGs. a HPT in the pCAMBIA 1381 vector was replaced by EcPMI or by plant PMIs. Then, the transformation of derivative constructs was examined individually. b The ratio of independent resistant calli (events) generated from agrobacteria-incubated calli. c The ratio of independent regenerated events to resistance events. d The ratio of PCR-positive events to total regenerated events. e The ratio of PCR-positive events to agrobacteria-incubated calli. f The ratio of lowcopy PCR-positive events to total PCR-positive events; low-copy events (harboring one or two copies) were determined by real-time PCR using a TaqMan probe for the 35S promoter, as described previously. The resistant calli or plants that developed from the same Agrobacterium incubation were defined as the same event. All data are presented as the mean ± SD (n = 2).   rice. These data suggest that the overexpression of CvPMI, AtPMI2 and OsPMI1 should enhance the PMI activity of rice cells; therefore, these genes are effective as SMGs for Man selection.

Transformation of a gene of interest (GOI) in rice using the plant PMI/Man system. To assess
whether plant PMIs could be used as SMGs for GOI transformation, three binary vectors that contained both the maize ubiquitin 1 promoter (Ubq)-driven individual plant PMI cassette (the SMG) and the 35S-driven HPT cassette (GOI) were constructed. The transformation of the corresponding PMI-HPT vectors into Nipponbare was similar to that of the above-described vectors containing each SMG alone. As listed in Table 4, no obvious difference in the selection and regeneration efficiency of the sole SMG vector and the corresponding SMG-GOI vector was observed. GOI transformations in the regenerated plants were molecularly identified by sequence-specific PCR of the PMI cassette and the HPT cassette (the Ubq promoter and HPT gene, respectively), and almost all PMI-positive plants had an HPT fragment (Table S4). We noticed that plants lacking HPT also lacked a PMI cassette, suggesting that the regenerated plants lacking the GOI may be caused by the escape of cells under Man pressure. Furthermore, six T 0 lines of each construct were randomly selected, and their seeds were collected and germinated under Man and hygromycin selection separately. As expected, the T 1 population could resist both selection agents, and the progeny segregation of Man resistance was similar to that of hygromycin resistance (Table S5). Taken together, these data suggested that the GOI could be reliably and efficiently transformed into rice by the plant PMI/Man system.

Discussion
All current PMI/Man systems use EcPMI as the SMG, except for a few examples that take advantage of the yeast PMI. Apparently, there are significant differences in the gene expression between prokaryotes or yeast and higher plants. In a previous study, we showed that the utilization of the plant codon-optimized EcPMI exhibited higher selection efficiency in rice transformation, which suggested that the expression of the original EcPMI may be partially inhibited in plants 32 . Successful applications of the PMI system are dependent on the plant's inability to utilize Man. The disappearance of PMI activity in plants was long presumed to stem from the absence of the PMI gene in the genome 33 . However, genome research has revealed that the homologous genes of type I PMI are universally observed in higher plants 27 . One conventional explanation was that the plant homologies of PMI might be deficient in isoenzymatic activity. In contrast to this hypothesis, biochemical assay of the histidine-tagged AtPMI1 and AtPMI2 proteins produced in E. coli indicated that these enzymes are activated in vitro 26 . The function of AtPMI2 was also examined in our study. The purified GST-AtPMI2 protein exhibited obvious M-6-P catalytic activity in vitro. In addition, AtPMI2 overexpression significantly enhanced the in vivo PMI activity in rice. Both of these results further confirmed that the AtPMI2 protein should be an active isoenzyme. Rice is a typical plant that is sensitive to Man accumulation 34,35 . OsPMI1 has a conserved sequence motif of the typical PMIs. Similar to AtPMI2, GST-OsPMI1 clearly displayed PMI activity. Although the activities of AtPMI2 and OsPMI1 were lower than that of EcPMI in vitro, their selection efficiencies of rice transformation were comparable. Additionally, the proteins of the transgenic plants expressing AtPMI2, OsPMI1 and EcPMI exhibited the same levels of biochemical activity regarding M-6-P conversion. These data imply that the in vivo activities of AtPMI2 and OsPMI1 may be similar to that of EcPMI. Because AtPMI2 and OsPMI1 also provide sufficient ability to metabolize Man in plants, we hypothesize that these two PMIs can be used as potential substitutes of EcPMI in the existing PMI/Man selection system. In addition, our results indicate that rice overexpressing OsPMI1 can grow under Man pressure, which suggests that the absence of PMI activity in rice should be caused by extremely low OsPMIs expression. Notably, background PMI activity can be detected in many Man-sensitive plants, such as Arabidopsis, papaya, sugarcane and Chinese cabbage [36][37][38] , and putative type I PMIs do exist in the genomes of plants. Therefore, we hypothesize that the overexpression of PMIs of Man-sensitive plants may also enhance the PMI activity of cells and thereby may be utilized as alternative SMGs in plant genetic engineering. Earlier reports suggested that most plants lack PMI or have very low PMI activity 39 , whereas a few species with high activity were noted in subsequent studies. Man is deeply involved in the metabolism of mannitol 40 . In some legumes and umbelliferous plants (such as Cassia coluteoides and celery), mannitol is the major photosynthetic product and/or translocated carbohydrate. Unsurprisingly, these plants have high levels of PMI 41,42 . Other plants, including certain algae as well as some varieties of grape and sweet potato, can easily overcome the growth toxic effect of high concentrations of exogenous Man and can even use Man as an exclusive carbon source during growth [43][44][45] . Although direct evidence is lacking, it is reasonable to presume that their PMIs should be highly active. In this study, CvPMI, which was encoded by the only copy of PMI in the genome of the Man-tolerant Chlorella, exhibited a strong activity in vitro. The PMI activities in most CvPMI-overexpressing plants were higher than were those in transgenic plants carrying EcPMI, AtPMI2 and OsPMI1 overexpression constructs. Although the protein levels of PMIs might differ in these plants, we could still infer that CvPMI was more active in vivo than EcPMI or the PMIs isolated from Man-sensitive plants because they were driven by the same promoter. In addition, rice cells carrying CvPMI regenerated more effectively than the other PMIs under a high concentration of Man, implying that CvPMI could confer stronger Man metabolism than EcPMI and the other plant PMIs. Taken together, our results suggest that the PMI homologs from Man-tolerant plants might be used as better SMGs than EcPMI in transformation systems of rice and other plants.
Because Man itself does not kill non-transformed cells, the positive selection of Man normally has a higher efficiency than the negative selection of antibiotic or herbicide systems. We showed that three plant PMIs all exhibited high selection efficiencies in Indica and Japonica in this study. Using the same binary vectors containing the plant PMI expression cassette and the HPT cassette, Man selection is much more efficient than hygromycin selection (HPT selection data are indicated in Table S6). The escape issue is much more serious in PMI/Man systems. Given that Man seriously affects plant organogenesis, the regeneration medium of the conventional rice transformation protocol using EcPMI normally contains no or a low concentration of Man. Generally, approximately 5% to 10% of regenerated plants (depending on constructs, transformation protocols and rice varieties) produced by EcPMI/Man systems are non-transgenic plants 30,31,34,46 . Our data indicate that CvPMI provided more efficient regeneration compared with EcPMI under a high concentration of Man and that the escape rates were greatly reduced with the increase in Man. These data suggest that the combination of optimized Man medium and CvPMI or other highly active plant PMIs may overcome the escape issue of PMI/Man systems.
The social concern of consuming GM crops is partly attributed to the notion of eating "unnatural" DNA and its products. The safety of EcPMI has been well evaluated by a series of risk assessments, and no adverse effects on the health of mammals or on the environment have been demonstrated 47,48 . However, this protein remains a "foreign" element in transgenic plants. We showed that at least three plant PMIs were sufficiently powerful SMGs for rice transformation, suggesting that plant PMIs can be used as substitutes of EcPMI in plant genetic  engineering. According to the principle of cisgenesis 49-51 , we believe that plant transformation systems using their own PMI or PMIs from closely related varieties could potentially alleviate the skepticism of the public regarding transgenic technology in plants.

Methods
Gene cloning and vector construction. To   PMI in vitro assay. The in vitro activity of PMI was measured at 30 °C using a coupled enzymatic assay based on a modified protocol 52 . The 300 μ L total volume reaction mixture contained 257 μ L of Tris-HCl buffer (50 mM, pH 7.5), 25 μ L of D-M6P (over 10 concentrations ranging from 4 μ M to 48000 μ M, Sigma), 10 μ L of NADP (13.5 mM, Sigma), 2 μ L of phosphoglucoisomerase (PGI, 1 kU/mL, Sigma) and 1 μ L of glucose 6-phosphate dehydrogenase (G6PDH, 1 kU/mL, Sigma). The mixture was equilibrated to 30 °C, and the absorbance was monitored at 340 nm (A 340 ) until constant. Then, 5 μ L of diluted extract was added to initiate the reaction. The increase in A 340 was recorded for approximately 10 min. Enzyme activities were calculated from the initial linear rates of cofactor reduction after the subtraction of endogenous activities (which were measured in assays without substrate) in a Lineweaver-Burk assay. The protein concentration of extracts was estimated by the Bradford method, with bovine serum albumin (BSA) as the protein standard. Enzyme assays were performed on freshly prepared extracts, and each value was an average of three experiments. One unit of enzymatic activity converted 1 μ mol of D-M6P to D-F6P per min.
PMI in vivo assay. The specific activity of PMI in vivo was determined according to the coupled enzymatic assay procedure, with a slight modification 53 . Young leaf tissues (500 mg) were extracted in 400 μ L of ice-cold Tris-HCl (50 mM, pH 7.5) containing 1% protease inhibitor cocktail (Sigma), 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM dithiothreitol (DTT) and 1 mM polyvinylpyrrolidone (PVP). One hundred microliters of crude protein extract was added to the substrate solution consisting of 75 μ L of NADP (13.5 mM), 1 μ L of PGI (1 kU/mL), and 0.5 μ L of GDPDH (1 kU/mL) and incubated for 30 min at 37 °C to remove any endogenous D-M6P. The reaction was initiated by adding 20 μ L of D-M6P (50 mM), and the ▵ A 340 was recorded for 60 min to calculate the activities.
Rice transformations. The transformations followed previously described protocols 30,31 with modifications.
The embryos were isolated from mature rice (Nipponbare and Kasalath) seeds, and explants were induced at 30 °C for two weeks. The Agrobacterium-incubated calli were transferred to the recovery medium containing 250 mg/L carbenicillin for five days and then selected on Man-containing media for three weeks. Calli on the selection media were counted, and the resistant calli that emerged from each callus were regarded as one event. Then, two to three callus nodules from one resistance event were transformed in the regeneration medium for three weeks. Only one shoot from each event was transferred to the rooting medium. Then, the plants grew and produced seeds in a greenhouse. The medium compositions were identical to previously reported recipes 30,31 . All chemical agents used in the rice transformation were obtained from Sigma-Aldrich Co., USA.