High frequency of phenotypic deviations in Physcomitrella patens plants transformed with a gene-disruption library

Background The moss Physcomitrella patens is an attractive model system for plant biology and functional genome analysis. It shares many biological features with higher plants but has the unique advantage of an efficient homologous recombination system for its nuclear DNA. This allows precise genetic manipulations and targeted knockouts to study gene function, an approach that due to the very low frequency of targeted recombination events is not routinely possible in any higher plant. Results As an important prerequisite for a large-scale gene/function correlation study in this plant, we are establishing a collection of Physcomitrella patens transformants with insertion mutations in most expressed genes. A low-redundancy moss cDNA library was mutagenised in E. coli using a derivative of the transposon Tn1000. The resulting gene-disruption library was then used to transform Physcomitrella. Homologous recombination of the mutagenised cDNA with genomic coding sequences is expected to target insertion events preferentially to expressed genes. An immediate phenotypic analysis of transformants is made possible by the predominance of the haploid gametophytic state in the life cycle of the moss. Among the first 16,203 transformants analysed so far, we observed 2636 plants ( = 16.2%) that differed from the wild-type in a variety of developmental, morphological and physiological characteristics. Conclusions The high proportion of phenotypic deviations and the wide range of abnormalities observed among the transformants suggests that mutagenesis by gene-disruption library transformation is a useful strategy to establish a highly diverse population of Physcomitrella patens mutants for functional genome analysis.


Background
The most informative approach to identify a function for a given gene is the precise inactivation or functional alteration of the gene, followed by the analysis of the phenotypic change resulting from this manipulation. Gene targeting based on homologous recombination between a targeting construct with altered or abolished gene function and its cognate endogenous gene has been a highly successful approach for gene function analysis in prokaryotes, lower eukaryotes, and mice. Unfortunately, in higher plants this approach is restricted by the very low ratio of 10 -3 to 10 -5 targeted relative to illegitimate recombination events. Although a few homologous recombination events between incoming targeting constructs and their cognate genomic sequences have been described, homologous recombination remains very inefficient and gene targeting thus is not routinely possible in higher plants [1,2]. In contrast, gene targeting via homologous recombination occurs with high frequency in the moss Physcomitrella patens [3,4]. After the first demonstration of high-frequency recombination between chromosomal sequences and homologous DNA introduced by transformation [5], gene targeting in Physcomitrella was used successfully to study the function of several genes by creating functional knockouts [6][7][8][9]. The high specificity provided by homologous recombination even allows the specific targeting of single members of multi-gene families [10].
The ease of its genetic manipulation, together with a high degree of conservation of morphological features, biochemical pathways and signal transduction mechanisms between Physcomitrella patens and higher plants [11][12][13][14] has made the moss an important model system for plant functional genomics. To facilitate a large-scale study of plant gene function using Physcomitrella patens as a model organism, we are developing a collection of Physcomitrella plants with insertion mutations that affect a wide variety of developmental, morphological and physiological characteristics. Transformation with constructs carrying sequences homologous to the genome typically results in 10-fold higher transformation frequencies then the use of non-homologous constructs, and among these transformants a high proportion shows integration of the construct at the homologous genomic locus [3,12]. We argued that -compared to a random mutagenesis strategy [15] -targeting insertion mutations towards expressed genes would increase the proportion of transformants displaying altered properties, and would decrease the total number of transformants to be screened to find a particular change in phenotype. We therefore developed an efficient transposon-based shuttle mutagenesis system for moss cDNA libraries, and have used pools of insertionmutagenised cDNA clones tagged with a nos-regulated nptII selection cassette for the transformation of Physcomitrella plants (Fig. 1).

cDNA library
To establish a Physcomitrella cDNA library representing most genes expressed during vegetative growth before the onset of differentiation, RNA was extracted from protonemata cultured for different time periods in liquid culture, and a cDNA library in plasmid vectors was established after normalization to decrease redundancy [16]. Mass DNA sequencing and clustering of 57,000 EST sequences yielded 12,000 non-overlapping sequence clusters, and showed a low degree of clone redundancy in the cDNA library used. Sequence analysis of these contigs, together with a large number of additional EST sequences derived from other growth stages and tissues, suggest that the total number of coding sequences for the moss Physcomitrella patens and the flowering plant Arabidopsis thaliana is similar (Rensing et al., submitted), despite a three-fold larger genome size for the moss [12].

Gene-disruption library
To create a gene-disruption library of cDNA clones carrying insertion mutations, cDNA clones were subjected to shuttle mutagenesis in E. coli. First, the normalised cDNA pool from the amplified protonema library was subcloned into the minimal vector pUCMinIV (Fig. 2). This plasmid is a 1.7 kb derivative of pUC19 from which we have deleted most non-essential DNA sequences to remove insertion targets within the vector sequence. Pooled minimal vectors with target cDNAs were introduced into a donor E. coli strain carrying an inducible transposase gene (tnpA) and a conjugative plasmid with a derivative of transposon Tn1000. This mini-transposon carries a modified nos-promoter driven nptII expression cassette encoding resistance against the antibiotic G418 as selectable plant marker gene between the border repeat sequences of Tn1000 required for transposition [17]. Induction of transposase activity by IPTG results in transposition and the formation of a cointegrate between conjugative plasmid and cDNA clone. Resolution of the cointegrates was achieved by conjugative transfer into a recipient strain overexpressing the tnpR resolvase gene, resulting in the release of a copy of the cDNA-carrying minimal vector with an insertion of the mini-Tn1000::nptII. These plasmids were isolated and retransformed into E. coli. Analysis of individual clones from this gene-disruption library, as well as model experiments using defined cDNA clone targets showed that in most of the resulting tagged plasmids the transposon had inserted into the moss cDNA, with no apparent preference for target sequence or insertion orientation (Fig. 2).

Transformation
Pools of plasmid DNA prepared from transposon-mutagenised cDNAs were used for large-scale PEG-mediated transformation of moss protoplasts grown in semi-continuous bioreactor cultures [6,18,19]. Before transformation, the plasmid DNA was linearised by digestion with a rarecutting restriction enzyme, SdaI, that cuts in the minimal vector just outside of the cDNA inserts (Fig. 2). Regeneration of protoplasts was done on supplemented Knop medium for 2 weeks without selection, followed by two cycles of G418 selection to eliminate unstable transformants [20]. More than 98% of surviving plants were stable transformants carrying the nptII selection cassette used for the cDNA disruption, as demonstrated by PCR-detection of the nptII coding sequence or a third selection step [21]. The cellular DNA content of all transformants was checked by flow cytometry [18]; 7.7% of the transformants tested (1242 of 16203) were polyploid. Currently, we have produced more than 22,000 moss transformants; the current production capacity is about 3,000 new trans-

Figure 2
Shuttle transposon mutagenesis of Physcomitrella patens cDNA clones. The structure of a representative moss cDNA clone (ID: S_PP015059353; 808 bp) in the pUC-MinIV minimal vector is shown. This defined plasmid was subjected to shuttle mutagenesis, and the transposon insertion sites for 72 resulting clones were mapped by DNA sequencing to assess the distribution of insertions. 41 insertions in "forward" orientation (nptII resistance marker on transposon and bla marker on vector transcribed in same orientation) are indicated by blue lines within the circle, 31 "reverse" insertions by red lines outside. Most of the insertions (66 / 72, corresponding to 92%) occurred throughout the cDNA, without apparent strong bias for insertion site or orientation. For production of the gene-disruption library used afterwards for the moss transformation, cDNA clones were mutagenised in pools; here about 70% of the mutagenised plasmids had insertions in the cDNA. formants per month. This will allow us to establish a large collection of Physcomitrella plants transformed with gene-disruption library constructs; our aim is to obtain a collection of plants carrying mutations in the majority of expressed moss genes.

Physiological mutants
To test for physiological alterations induced in moss plants after transformation with the gene-disruption library, we looked for changed nutritional requirements induced in the transformants (Fig. 3). 7.2% (1,163 of 16,203) of the regenerated plants showed a retarded growth during the culture on minimal medium but a normal growth during the culture on the same medium containing a combination of supplements (see Materials), whereas wild-type plants grew well on both media. This suggests a metabolic defect in transformants with a slowgrowth phenotype on minimal medium that could be rescued by supplying metabolites in the medium. 3.1% (508 of 16,203) of the transformants showed a retarded growth during the culture on minimal medium as well as on supplemented medium. Those plants might require substanc-es that are not present in our supplemented medium, or they might be defective in another way. Different nutritional mutants of Physcomitrella patens, e.g. resulting from mutagenised spores [22] have been described previously.

Morphological mutants
To screen for morphological and developmental mutations, transformants were microscopically observed after regeneration and selection on supplemented Knop medium for 11 weeks followed by growth on minimal Knop medium for 8 weeks. Under these conditions, the filamentous structure (protonema) that is initially formed by regenerating wild-type protoplasts undergoes a developmental switch. It forms a three-faced apical meristem bud, and differentiates into leafy stems, the gametophores (Fig. 4A), that will eventually carry the sex organs. 16.2% (2,631 of 16,203) of the gene-disruption library transformants showed morphological deviations in one or more of the features observed: structure and colour of the moss plant, coverage of the plant by gametophores, shape and uniformity of leaves and arrangement of cells within the leaves (Table 1). Such deviations were not observed to occur spontaneously in 350 analysed wild-type plants and in less than 1% of 400 plants derived from protoplasts mock-transformed without DNA and regenerated in the absence of antibiotic. Examples for deviating phenotypes observed in gene-disruption library transformants are shown in Fig. 4b to 4E. In addition to the phenotypic classes described above, we observed other developmental abnormalities, like the formation of outgrowths on the leaf surface, the formation of thread-like appendages on leaf tips, and an increased number of darkcoloured sectors on leaves. Therefore, a wide spectrum of morphological and developmental alterations is observed in moss plants transformed with a gene-disruption library. In addition, between the three parameters we assayed for each transformant -growth requirements, morphology and ploidy level -there did not appear to be a strict correlation, and we found various combinations of characteristics (Fig. 5).

Molecular analysis
For molecular analysis, genomic DNA was isolated from randomly chosen transformants and analysed in Southern blot experiments. To obtain an estimate for the number of gene-disruption constructs integrated in the transformed moss plants, this DNA was digested with PvuII, which has a single recognition site within the nptII coding sequence.
This results in the formation of two fragments detected by a nptII-derived probe for each copy of the selection cassette. The length of these fragments differs depending on the adjacent sequences, which can be either derived from genomic sequences or from cDNA contributed by the transforming DNA. On Southern blots of 55 transformants, 19.3 ± 3.6 (mean ± s. d.) bands were detected, suggesting an average insertion number of about 10 nptIIcassettes per transformant (data not shown). To discriminate between independent integration events at different genomic positions and co-integrations at the same locus, genomic DNA of 16 transformants was digested with several restriction enzymes that cut genomic Physcomitrella DNA frequently but do not cut within the nptII cassette, and only rarely within the cDNA sequences carried by the transforming DNA (Fig. 6). Most of the genomic DNA was cut by these enzymes into fragments ranging from 1 to 15 kb, as judged by ethidium-bromide staining after gel electrophoresis. However, the majority of fragments detected after hybridisation with the disruption-construct specific nptII-probe were more than 20 kb in size; only few distinct nptII-hybridising bands were observed. Digestion of the same DNA with PvuII, which cuts within the nptII selection cassette, resulted in several nptII-hybridising fragments of smaller size ranging between 1 and 10 kb (Fig.  6). This suggests that most of the transforming DNA molecules are co-integrated in close vicinity to each other,  without intervening genomic sequences, at few genomic loci. PCR amplification of transgene sequences between the nptII-insertions using nptII-primers and sequence analysis of the obtained products confirmed tandem integrations of different gene disruption constructs in close vicinity (data not shown). Similar integration patterns of transforming DNA molecules in Physcomitrella have been observed previously [5,15].

Conclusions
Taken together, the collection of Physcomitrella transformants that we are establishing by a cDNA based genedisruption library transformation approach shows a high degree of phenotypic diversity. The fraction of plants with altered morphological or developmental features -16.2% -exceeds that which has been previously described for a shuttle mutagenesis strategy using genomic clones of moss DNA (3.9%) [15]. This difference might be caused by preferential, homologous recombination-driven disruption of expressed moss genes after the cDNA mutagenesis approach. The frequency of altered phenotypes in our collection of moss transformants also is considerably higher than the less than 2% reported for confirmed Arabidopsis knockout mutants [23], which might be attributed to a lower degree of gene redundancy in Physcomitrella (Rensing et al., submitted). The small number of genomic loci tagged by insertion of gene-disruption library constructs in plants from our mutant collection should allow the recovery of insertion sites and the analysis of moss genes affected in transformants with altered phenotypes. Indeed, we have re-isolated genomic sequences surrounding nptII-insertion sites e. g. by PCR-based methods. The presence of introns in some of these sequences suggests that these were derived from the genomic target locus, into which transgene sequences had inserted by homologous recombination (data not shown). We therefore expect that the collection of moss mutants being established here will help to identify novel genes previously unknown in plants [7,12] and will allow to rapidly link DNA se-quence and functional information. Given the high degree of genetic and physiological conservation between moss and higher plants, this collection of gene disruption library transformants will be a valuable tool not only for gene function studies in the moss Physcomitrella, but for plant functional genomics in general.

Transposon mutagenesis
Plasmid constructions and microbiological techniques followed standard procedures [24,25] [26], a chloramphenicol resistance marker and a nos-promoter-nptII -nos terminator expression cassette flanked by the transposon Tn1000 border repeat sequences required for transposition [17]. The nptII cassette was derived from pBIN19 [27]. To increase the efficiency of subsequent plant selection, a point mutation present in the nptII coding sequence [28] was reverted to the wild-type sequence. Ampicillin-resistant R2117 donor strain transformants with moss cDNA clones were pooled, treated with IPTG to induce transposase activity, and cocultivated with R1037 recipient cells in the presence of IPTG. E. coli strain R1037 carries an IPTG-inducible tnpR resolvase gene and a streptomycin resistance locus, both encoded on the chromosome. Tn1000 transposition in the donor results in the formation of cointegrates between conjugative plasmid and moss cDNA clones, which then can be transferred by conjugation into the recipient where resolution occurs. Recipient cells with mutagenised moss cDNA clones were Deviations from wild-type for any morphological characteristic 16.2% -Plant structure (less compact than wild-type, fluffy) 13.1% -Color (darker or lighter than wild-type) 8.0% -Coverage of plant by gametophores (coverage less than 95%; wild-type: completely covered) 8.0% -Leaf shape (width/length ratio or symmetry different from wild type) 12.3% -Cell shape (irregular patterns, or cell size different than wild-type) 10.1% -Uniformity of leaves (more than 10% deviating on one plant, wild-type: all uniform) 6.3%

Figure 6
Southern blot analysis of gene-disruption library transformants. 1 µg of genomic DNA isolated from two representative Physcomitrella transformants (BC344 and BC3238) was digested to completion with PvuII, which has a recognition site within the nptII coding sequence, or one of seven other restriction enzymes that do not cut within nptII and only rarely in Physcomitrella cDNAs. Fragments carrying nptII -sequences were detected by a DIG-labelled probe after electrophoresis and transfer to a nylon membrane.
selected by their simultaneous resistance against ampicillin and streptomycin, and plasmid DNA prepared from pooled cells was used then to transform P. patens.

Molecular analysis
The presence of transgenes in G418 resistant moss transformants was confirmed by PCR with nptII specific primers, or a third selection step as described [21]. For Southern blot analyses, 1 µg of genomic DNA isolated by a modified CTAB method [31] was digested for 5-6 hours with 20 U of restriction enzymes purchased from MBI Fermentas or New England Biolabs. After electrophoresis on 0,7% agarose gels the DNA was transferred onto positively charged nylon membrane (Roche). Fragments carrying nptII -sequences were detected using hybridisation and blocking solutions as well as Anti-digoxigenin-AP conjugate from Roche but CDP-Star from Promega. The DIG-labelled nptII probe was generated by PCR using the random primed labelling mix from Roche and Taq polymerase from Promega.

Availability of materials and mutant plants
Materials, single plant lines described in this communication, as well as sequence information for Physcomitrella EST clones with similarity to defined genes of interest will be available for non-profit research after completion of an appropriate material transfer agreement with BASF Plant Science GmbH and Freiburg University. Also, inquiries for sequence information for Physcomitrella EST clones with similarity to defined genes of interest are encouraged. Requests should be directed to the corresponding author (RR).