High-efficiency stable transformation of the model fern species Ceratopteris richardii via microparticle bombardment.

A highly efficient method transforms fern callus tissue, with rapid and simple selection for stable transgenic lines through antibiotic selection. Ferns represent the most closely related extant lineage to seed plants. The aquatic fern Ceratopteris richardii has been subject to research for a considerable period of time, but analyses of the genetic programs underpinning developmental processes have been hampered by a large genome size, a lack of available mutants, and an inability to create stable transgenic lines. In this paper, we report a protocol for efficient stable genetic transformation of C. richardii and a closely related species Ceratopteris thalictroides using microparticle bombardment. Indeterminate callus was generated and maintained from the sporophytes of both species using cytokinin treatment. In proof-of-principle experiments, a 35S::β-glucuronidase (GUS) expression cassette was introduced into callus cells via tungsten microparticles, and stable transformants were selected via a linked hygromycin B resistance marker. The presence of the transgene in regenerated plants and in subsequent generations was validated using DNA-blot analysis, reverse transcription-polymerase chain reaction, and GUS staining. GUS staining patterns in most vegetative tissues corresponded with constitutive gene expression. The protocol described in this paper yields transformation efficiencies far greater than those previously published and represents a significant step toward the establishment of a tractable fern genetic model.


INTRODUCTION
Ferns represent an under-investigated group compared to many other taxa of land plants.
Ferns and horsetails together comprise the monilophytes, which diversified from the seed plant (spermatophyte) lineage approximately 400 Mya (Pryer et al., 2001). As such, monilophytes represent the closest extant sister-group to seed plants. Comparisons between ferns and seed plants should thus provide important insights into the developmental mechanisms present in the ancestral tracheophyte from which both taxa derive, and also elucidate subsequent evolutionary trajectories.
The most extensively-studied fern species is Ceratopteris richardii, a homosporous fern increasingly viewed as a viable experimental model (Hickok et al., 1995;Chatterjee and Roux, 2000;Leroux et al., 2013). The C. richardii lifecycle comprises gametophyte and sporophyte stages that are capable of growing independently of one another. Dispersal is via haploid spores, which germinate to form thalloid gametophytes. Gametophytes develop into either chordate hermaphrodites, characterized by the presence of a lateral 'meristem' (Banks, 1999), or, in the presence of a hermaphrodite-secreted antheridiogen, as males (Banks, 1997).
Sexual reproduction in this species requires the presence of water and occurs through fusion of retained egg cells and motile sperm. The resultant diploid embryo develops within the gametophyte archegonium (Johnson and Renzaglia, 2008). Subsequent growth of the sporophyte occurs indeterminately through divisions of a tetrahedral shoot apical cell (Hou and Hill, 2002), the products of which establish both frond primordia and a shoot-derived root system, each with their own associated apical cells (Hou and Hill, 2004). Frond and root development are both heteroblastic in nature, in that the morphology of newly-arising organs alters with the age of the sporophyte (Hou and Hill, 2002). Ultimately, haploid spores are generated on the lower lamina surface of reproductive fronds. The spore to spore lifecycle takes an average of 22 weeks.
The establishment of a fern genetic model has been hindered by a number of technical factors, not least large haploid genomes (Bennett and Leitch, 2001) that in the absence of a pressing incentive remain uneconomical to sequence. For example, C. richardii is estimated to have a haploid genome size of approximately 11.3 Gbp (Nakazato et al., 2006). The greatest impediment to detailed genetic analysis, however, is an inability to efficiently transform ferns. Transient transformation of fern gametophyte prothallus cells, typically for RNA interference (RNAi), has been previously demonstrated through either direct DNA 6 uptake by germinating C. richardii spores (Stout et al., 2003) or direct microparticle bombardment in Adiantum capillus-veneris (Kawai-Toyooka et al., 2004), C. richardii (Rutherford et al., 2004) and Pteris vittata (Indriolo et al., 2010). Evidence of stable transmission to the subsequent sporophyte generation was reported, but where quantified (Rutherford et al., 2004) transmission through self-fertilization was very low (7%), and a significant proportion of transmitted events ultimately reverted to a non-silenced phenotype (32%). Transmission to subsequent generations was not determined. A recently published protocol utilizing Agrobacterium-mediated transformation of spores reported stable transformation in two fern species, Pteris vittata and Ceratopteris thalictroides (Muthukumar et al., 2013), but the very low transformation efficiencies achieved (0.053% and 0.03% respectively) preclude routine adoption of this approach.
In this paper we demonstrate the genetic transformation of both C. richardii and C. thalictroides using microparticle bombardment of callus tissue and hygromycin selection of regenerating transformed plants. Transgenes were stably inherited in subsequent generations.
With transformation efficiencies of 72% (C. richardii) and 86% (C. thalictroides) this technical advance positions C. richardii as a tractable genetic model for the analysis of gene function in ferns.

Induction of callus tissue from fern sporophytes with cytokinin
Callus tissue has been generated from numerous angiosperm species and maintained in an undifferentiated state by treatment with the phytohormones auxin and cytokinin (CK) (Ikeuchi et al., 2013). A similar approach was attempted in both C. richardii and C. thalictroides, first by treating gametophytes with either CK or auxin. Although gametophyte development was visibly altered, neither hormone treatment induced callus formation in either species (Supplemental Fig. S1). To determine the effect on sporophyte development, 11 day-old C. richardii sporophytes (Fig. 1A) were incubated on Murashige and Skoog (MS) media containing CK or auxin analogues. After 14 days, new fronds and roots had emerged on untreated sporophytes (Fig. 1B), whereas sporophytes treated with the auxin analogue 1napthaleneacetic acid (NAA) produced new fronds, but new roots were replaced by disorganized, callus-like tissue (Fig. 1C). Treatment with a second auxin analogue, indole-3butyric acid (IBA) did not dramatically perturb shoot or root development (Fig. 1D); however, the tip of the embryonic root remained green for longer than in control or other 7 hormone-treated sporophytes. In contrast to the root-specific effects of auxin, treatment with two separate cytokinins, benzylaminopurine (BAP) and kinetin (KT), prevented the production of both fronds and roots on growing sporophytes. Instead, undifferentiated callus tissue was visible at the shoot apex (Fig. 1E, 1F) and, occasionally, the root apex (Fig. 1E).
CK treatment of C. thalictroides sporophytes also successfully induced callus at the shoot apex (Supplemental Fig. S2).
When grown on MS media without further hormone treatment, shoot-derived C. richardii callus completely differentiated into new shoots and roots within four weeks (Fig. 1G, 1H).
Callus treated with auxin analogues also regenerated into new shoots over the same period ( Fig. 1I-1L), with differentiation slightly delayed under NAA treatment (Fig. 1I, 1J). In contrast, continued incubation on BAP maintained callus tissue in an apparently undifferentiated state (Fig. 1M, 1N). Treatment with KT, although partially successful in preventing cell differentiation, was less effective (Fig. 1O, 1P). Combined treatment with BAP plus NAA resulted in reduced callus growth, whereas BAP and BAP plus IBA treatments were essentially indistinguishable, both yielding highly-friable callus (Supplemental Fig. S3). Fern callus can thus be maintained on BAP, or on a combination of BAP plus IBA. To determine callus longevity, both C. richardii and C. thalictroides calli were repeatedly subcultured on successive BAP plus IBA treatments at 14 day intervals.
Indeterminate cell fate was successfully maintained in this manner for over a year.

High-efficiency transformation of C. richardii and C. thalictroides using microparticle bombardment and hygromycin selection
Callus transformation was carried out using microparticle bombardment with a hygromycinselectable 35S::GUS construct (pCAMBIA1305.2: see Materials and Methods). To first determine the effectiveness of hygromycin as a selection agent, untransformed C. richardii callus was subjected to a range of antibiotic concentrations across different timeframes. A two week incubation period on 40µgml -1 hygromycin B was sufficient to prevent regeneration of shoots from untransformed callus (Supplemental Figure S4) and was therefore used in all subsequent callus selection assays. Test bombardments with either the 35S::GUS construct or with uncoated control microparticles were performed on both C. richardii and C. thalictroides callus. Callus was incubated on MS media containing CK (5µm KT) during and following bombardment to prevent premature tissue differentiation. Callus was transferred to this media two days prior to bombardment and left on the same media after https://plantphysiol.org Downloaded on December 31, 2020. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved. 8 bombardment for a three day recovery period without antibiotic selection. After that time, GUS staining analysis was performed on samples of bombarded C. richardii callus tissue. Figure 2A shows that callus bombarded with the 35S::GUS construct exhibited numerous spots of GUS staining, whereas callus bombarded with uncoated microparticles showed none ( Fig. 2B). Multiple transformation events had thus taken place.
After the three day recovery period, the remaining callus was transferred to antibiotic selection media and a week later GUS assays were repeated. Once more, spots of GUS expression were visible within the population of 35S::GUS-bombarded calli ( Fig. 2C) but not on control calli (Fig. 2D). The frequency of spots visible on callus under selection was visibly reduced compared to callus stained immediately after bombardment (compare Fig. 2A and 2C). This difference most likely reflects stable versus transient transformation events.
Antibiotic selection was maintained for 14 days, after which time callus was transferred to non-selective MS media to regenerate. At this stage CK treatment was stopped. After a further seven days, small protruding regions of green tissue were visible on 35S::GUSbombarded callus (Fig. 2E) whereas control callus had turned dark brown and stopped growing (Fig. 2F). Regenerating tissue subsequently went on to differentiate discrete organs and continued GUS expression in these regenerating tissues was confirmed by GUS staining (Fig. 2G). Eight weeks after bombardment, regenerated sporophyte shoots had successfully established an indeterminate growth pattern (Fig. 2I), with most individual calli regenerating more than one shoot.
To assess transformation and regeneration frequencies, 18 bombardments were performed on C. richardii callus over a period of 10 weeks, each replicate comprising 45-60 calli. On average, 81.21% ± 2.45 of calli from each bombardment regenerated at least one shoot after selection, whereas corresponding control calli showed no regeneration (Table 1). Of these regenerating calli, 88.06% ± 1.60 exhibited GUS expression when stained, resulting in a final transformation efficiency of 71.58% ± 2.56. A parallel experiment using C. thalictroides yielded efficiencies of 96.76% ± 0.82, 88.58% ± 1.18 and 85.79% ± 1.62, respectively.
Once transplanted to soil, maturation of transgenic C. richardii sporophytes took 10-14 weeks (harvesting of first spores to harvesting of final spores), with a total minimum regeneration period of 18 weeks from bombardment to harvesting of earliest T 1 spores (compared to 22 weeks spore to spore for untransformed plants). In comparison, maturation Not all of the T 0 shoots that regenerated after hygromycin selection displayed GUS staining.
The extent of GUS expression in regenerating shoots also varied, from staining across entire shoot clusters originating from a single callus (e.g. Fig. 3A), staining of individual shoots within a shoot cluster (Fig. 3B), to staining of sectors within a single shoot or frond (Fig. 3C).
Staining across entire shoot clusters was observed at mean frequencies of 18.93% ± 2.92 (whole tissue staining) and 45.54% ± 2.91 (vasculature only) of GUS-staining shoots in C.
The occasional non-coincidence of GUS staining and hygromycin resistance was further observed in mature T 0 sporophytes: 28.57% (C. richardii) and 25.58% (C. thalictroides) of https://plantphysiol.org Downloaded on December 31, 2020. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved. mature individuals sampled within the regenerated T 0 populations showed no GUS staining.
T-DNA expression was analysed in greater detail in six T 0 C. richardii transformants through RT-PCR (Fig. 3). Of these six, only two (plants 2 and 10) showed GUS staining (Fig. 3G. 3H) in conjunction with amplification of both 35S::GUS and Hyg R gene products (Fig. 3M,   3N). Two further individuals (plants 16 and 17) showed no evidence of transgene expression except hygromycin resistance during tissue regeneration, and the remaining two (plants 18 and 23) each produced conflicting results: plant 18 was positive for GUS staining but negative for both 35S::GUS and Hyg R amplification, whereas plant 23 was positive for GUS staining and Hyg R amplification but negative for GUS amplification. Different frond tissues were necessarily sampled for GUS and RT-PCR assays, which might explain these discrepencies (see Discussion).

Transgene inheritance and stable expression in T 1 transformants
To assess transgene inheritance, T 1 spores that were harvested from 25 C. richardii T 0 individuals were screened for hygromycin resistance (Fig. 4). The possibility that T 0 transformants are chimeric necessarily creates the hypothesis that T 1 progeny will comprise a mix of transformed and untransformed individuals, requiring an efficient method to identify transgenics. Empirical testing determined that 20µgml-1 hygromycin is sufficient to kill untransformed, germinating C. richardii spores (Supplemental Fig. S4; Fig. 4A, 4B) and untransformed sporophytes (Supplemental Fig. S4). Under this selection regime, spores harvested from 18 T 0 individuals (72%) produced hygromycin-resistant gametophytes.
thalictroides, under similar screening conditions, 94% of lines produced resistant individuals, with frequencies within each line over a similar range (Supplemental Fig. S6). GUS staining of hygromycin-resistant T 1 individuals revealed that gametophytes in 90% (C. richardii) and 68% (C. thalictroides) of transgenic lines also expressed GUS. Hygromycin selection of T 1 germinating spores is thus a highly efficient method for identifying lines that carry intact transgenes.
Differences in GUS expression were occasionally found between the gametophyte and sporophyte stages of some T 1 lines, in that constitutive expression was not seen in transgenic gametophytes but was seen in sporophytes. No variation in expression pattern was observed and 17 (Fig. 4L), the latter displaying no GUS expression in the T 0 parent (Fig. 3J, 3M). Two lines (16 and 18), which had low frequencies of resistant gametophytes (Fig. 4M, 4N, 4P, 4Q), showed GUS expression only in basal thallus tissues and rhizoids (Fig. 4O, 4R), again inconsistent with the expression observed in the T 0 parents (Fig. 3I, 3K). A single T 1 line (23) displayed no GUS expression within the gametophyte despite a high frequency of hygromycin resistance (Fig. 4S-4U). The absence of GUS expression from the gametophyte persisted in four out of five T 2 lines descended from this line (Supplemental Fig. S7

DNA blot analysis of transformed C. richardii lines
To determine transgene copy number in transformed C. richardii T 0 plants, genomic DNA blots were hybridized to fragments of both the Hyg R and GUS genes ( Figure 6A). Figure 6B and 6C show that the fragments hybridized to multiple copies of each transgene in all T 0 https://plantphysiol.org Downloaded on December 31, 2020. -Published by Copyright (c) 2020 American Society of Plant Biologists. All rights reserved. transformants tested. The fewest insertions were found in T 0 plant 2, with two Hyg R and two GUS fragments hybridized. The remaining individuals all contained in excess of eight copies of each transgene. Hybridization patterns in individuals 10, 16, 17 and 18 were very similar, raising the possibility that these lines were derived from a single transformation event.
Importantly, in each T 0 individual, both Hyg R and GUS probes hybridized to genomic fragments greater than the size of the introduced plasmid (11.9 kb, Fig. 6A), with at least one instance per individual of a hybridized fragment being shared between the two probes (Supplemental Fig. S9), supporting linked insertion of the Hyg R and GUS genes.
To assess the inheritance of transgene insertions, DNA blot analysis was performed on T 1 progeny from four of the T 0 transformants analyzed above. Similar numbers of hybridized fragments were identified in T 1 individuals (Fig. 6D, 6E) as in the T 0 parents (Fig. 6B, 6C).
Of the two progeny tested from line 2, one (plant 2) demonstrated a hybridization pattern very similar to the T 0 parent for both Hyg R (Fig. 6B, 6D) and GUS (Fig. 6C, 6E) probes, but the second (plant 1) had apparently lost the insertion carrying the linked transgenes (Supplemental Fig. S9). The T 1 individuals tested from other lines also demonstrated very similar hybridization patterns to their T 0 parents, including the presence of linked 35S::GUS and Hyg R cassettes (Supplemental Fig. S9). Transgene insertions thus remained stably integrated within the C. richardii genome between the T 0 and T 1 generations, and linkage was maintained through meiosis.

DISCUSSION
Stable transformation of Ceratopteris richardii and its sister species C. thalictroides has been achieved using a combination of tissue culture, microparticle bombardment and antibiotic selection. Callus was initiated from the apical region of developing sporophytes of both species by application of the phytohormone cytokinin, and was maintained in an indeterminate state in vitro under the same treatment ( Fig. 1; Supplemental Fig. S2, S3). A 35S::GUS cassette linked to a hygromycin B resistance (Hyg R ) selectable marker cassette was introduced into callus cells by microparticle bombardment, and the regeneration of transformant sporophytes was successfully selected through hygromycin selection in tissue culture ( Fig. 2-3). GUS expression was detected in bombarded C. richardii calli before (transient expression), during, and after (stable expression) antibiotic selection. The expression of the binary cassette in regenerated C. richardii T 0 tissues was confirmed by RT-PCR and GUS staining (Fig. 3-5 gametophyte and sporophyte stages of the T 1 generation of both species (Fig. 4-5), suggesting stable integration into the Ceratopteris genomes. DNA blot analysis of individual C.
richardii T 0 sporophytes and T 1 progeny confirmed inheritance of the two cassettes, with multiple insertion events present in each of the lines examined (Fig. 6). This approach achieved transformation efficiencies of 72% for C. richardii and 86% for C. thalictroides.
Although past attempts to transform C. richardii via Agrobacterium reportedly failed (Hickok et al., 1987), a protocol for successful Agrobacterium-mediated transformation of C. thalictroides has recently been published (Muthukumar et al., 2013), with transformation achieved through infection of geminating spores. Interestingly, Muthukumar et al. (2013) report that their attempts to use hygromycin as a selectable marker for transformation were unsuccessful. This is potentially explained through our observation that hygromycin killed untransformed T 1 gametophytes on germination (Supplemental Figure S4). Transformed T 0 spores may not have sufficient time to express the introduced Hyg R gene at levels sufficient to confer resistance. In contrast, hygromycin selection was found to be a very efficient screening mechanism when identifying transformants regenerating from callus tissue.
Regardless of the selection technique, Agrobacterium-mediated transformation of both C. thalictroides and P. vittata spores occurred at very low frequencies (0.03% and 0.053% respectively; Muthukumar et al., 2013). Attempts in the same report to directly transform a population of P. vittata spores by microparticle bombardment similarly yielded a very low transformation efficiency of only 0.012%.
The available data thus suggest that transformation of spores is an inherently less efficient method than transformation of callus tissue.
Using the protocol described in this paper, the time from callus transformation to recovery of T 1 spores was 18 and 16 weeks for C. richardii and C. thalictroides respectively, in contrast to an 11-13 week period for C. thalictroides via Agrobacterium-mediated transformation (Muthukumar et al., 2013). Despite the longer time frame, transformation of the sporophyte generation confers the advantage of direct production of T 1 spores without an intervening recombination/outcrossing event (sexual reproduction of T 0 gametophytes). Outcrossing of T 0 transformant gametophytes could conceivably reduce initial transgene copy number, however, cross-fertilization of two independent transformants is also possible. Notably, Muthukumar et al. (2013)  transformation offered by microparticle bombardment and the concomitant ability to screen effectively using hygromycin offset the slightly longer T 0 generation time required, and provide significant benefits over the Agrobacterium-mediated protocol.

Evidence for a conserved role of cytokinin in the shoot meristem of ferns and angiosperms
Treatment with CK was sufficient to induce the formation of callus tissue in place of new fronds at the shoot apex of C. richardii and C. thalictroides sporophytes, and to maintain the callus in an undifferentiated state ( Fig. 1; Supplemental Fig. S2). Application of auxin failed to induce shoot callus but did perturb cellular activity specifically at the root apex. It was recently reported that auxin treatment of the lycophyte Selaginella kraussiana similarly disturbs root organization (Sanders and Langdale, 2013). Organogenesis in the C. richardii shoot arises from divisions of a single tetrahedral apical cell (Hou and Hill, 2002) instead of a multicellular meristem as found in angiosperms (Sussex, 1989). The observed formation of callus suggests that CK treatment can block the differentiation of cells derived from the apical initial cell. In Arabidopsis, CK acts downstream of knotted1-like homeobox (KNOX) genes to maintain undifferentiated cell fate in the shoot apex, in part through antagonism of the gibberellin signaling pathway (Jasinski et al., 2005;Bartrina et al. 2011). Given the results observed here it is possible that CK acts in both ferns and angiosperms to promote indeterminate cell fate at the shoot apex. Interestingly, CK treatment did not induce callus formation from C. richardii or C. thalictroides gametophytes, although morphology at the notch meristem was slightly affected (Supplemental Fig. S1). This could indicate that the mechanisms regulating initial cell specification differ between the notch meristem and shoot apex, and could reflect an important distinction between 2D (gametophyte) and 3D (sporophyte) growth and patterning.

Developmental trajectories explain chimeric versus stable transgene expression in T 0 and T 1 generations
The transmission of transgenes from regenerated T 0 plants to stable T 1 lines was assessed in C. richardii by comparing the transgenic status of T 0 parents (as assessed by GUS staining) ( Fig.3; Supplemental Fig. S5) to their T 1 progeny (as assessed by gametophyte hygromycin resistance) ( Fig. 4; Supplemental Fig. S6). 60% of lines demonstrated consistent transgenic status between the T 0 and T 1 generations, i.e. both parent and offspring were transgenic (52%) or neither parent nor offspring were transgenic (8%). The remaining 40% of lines showed inconsistent inheritance patterns, with transgenic T 1 individuals identified from apparently non-transgenic T 0 parents (20%) or non-transgenic T 1 progeny descending from apparently transgenic T 0 parents (20%). In C. thalictroides, corresponding frequencies of 67% consistent (64% plus 3%) and 33% inconsistent (3% and 30%) lines were recorded. A distinction in GUS expression patterns between the T 0 and T 1 generations was also observed, with frequent examples of partial or tissue-specific (vasculature) GUS staining in T 0 regenerated sporophytes (young and mature) but constitutive GUS expression in subsequent T 1 progeny. A low incidence (11-12%) of regenerated shoot clusters that lacked any GUS staining could reflect stable transformation with only the Hyg R cassette or subsequent loss of the 35S::GUS cassette after hygromycin selection. Non-congruence of GUS staining patterns between the T 0 and T 1 generations is unlikely to reflect technical issues of GUS substrate penetration given the examples of constitutive staining observed. Instead these results are most simply explained through the regeneration of chimeric T 0 transformants as previously seen in transformed gametophytes (Rutherford et al., 2004). This interpretation would also explain the observed discrepancies between GUS analysis, antibiotic selection and transgene expression data in T 0 plants (Fig. 3), as different fronds, potentially not all of them transgenic, were sampled for each analysis.
The emergence of chimeric shoots does not correspond with our current understanding of fern shoot development from single apical initial cells (Hou and Hill, 2002;Sanders et al., 2011). However, in accordance with the observations made above regarding callus induction, we hypothesize that induction of callus through CK treatment prevents the mitotic derivatives of the apical initial from differentiating into frond and root initials, and thus artificially expands the population of undifferentiated shoot apical initial cells. When this constraint is removed after bombardment, normal developmental patterning and gradients are presumably re-imposed onto an abnormally large population of undifferentiated cells. This could theoretically result in the incorporation of multiple apical initials into single regenerating shoots, allowing the formation of different tissue types from distinct subpopulations of initials. This scenario could theoretically explain our observations of vasculature-specific GUS staining patterns. However, given that regenerated plants appear morphologically similar to untreated controls, such a hypothesis would also imply that fern shoots and organs can successfully organize both from a founding population of multiple cells and from single initials. In contrast to the scenario proposed above for T 0 transformants, T 1 individuals must necessarily arise from a single progenitor cell (spore and zygote), and thus all cells within the individual have a common genetic ancestry. This is evidenced by the stable and constitutive GUS expression observed in the T 1 generation (Fig. 4, 5; Supplemental Fig. S6, S8). In the case of T 1 gametophytes, variations in the frequency of transgenics within each line (Fig. 4) most likely reflect the frequency of transformed sporangia on chimeric fronds: each sporangium arises from a single separate initial cell (Hill, 2001). This conclusion is supported by observations of T 0 pinnae containing both GUS-stained and unstained sporangia ( Fig. 3H, Supplemental Fig. S5). Importantly, the recovery of constitutively-expressing T 1 progeny from apparently chimeric T 0 transformants demonstrates that chimeric expression in the T 0 generation does not represent a barrier to establishing stable and pure-breeding transgenic lines, and argues that screening for stable transformants should occur in the T 1 generation instead of in regenerated T 0 shoots.

Transgene copy number
Microparticle bombardment resulted in the incorporation of multiple T-DNA fragments into the genomes of individual C. richardii T 0 transformants, including fragmented copies of individual expression cassettes. The complex insertion of multiple transgene copies is a known factor in biolistic-based transformation techniques (Hansen and Wright, 1999), and can in part be mitigated through bombardment with linearised DNA (Lowe et al., 2009). In all of the T 0 plants tested, the presence of linked Hyg R and 35S::GUS cassettes was revealed by shared hybridization to large genomic DNA fragments. Importantly, very little rearrangement of hybridization fragments was observed between the T 0 and T 1 generations, suggesting that transgene insertions remain essentially stable after initial integration. Within the relatively small sample of nine transgenic lines selected for DNA blot analysis in this study, one (line 2) was found to carry an entire transgene containing both the Hyg R and 35S::GUS cassettes, plus a single unlinked copy of both the Hyg R and 35S::GUS cassettes.
This observation suggests that the isolation of single insertion lines is feasible, especially if coupled with outcrossing to untransformed individuals.
Although microparticle bombardment typically results in a higher T-DNA copy number in transgenics than Agrobacterium-mediated transformation, as demonstrated by side-by-side transformations into barley (Hordeum vulgare; Travella et al., 2005), recent Agrobacterium-mediated transformation of C. thalictroides resulted in similar copy numbers to those reported here (Muthukumar et al., 2013).

C. richardii as a fern genetic model
The protocol described in this paper was able to successfully generate stable transformants in C. richardii and C. thalictroides at high efficiencies, and may be more generally applicable to other fern species. As stated in the introduction, C. richardii has previously been proposed as a candidate model fern: examples from a number of important classes of transcription factor have already been identified (Hasebe et al., 1998;Aso et al., 1999;Sano et al., 2005;Himi et al., 2001), and phylogenetic and cross-species complementation analyses have been published, for example with CrLFY (Maizel et al., 2005). With its smaller size, more rapid lifecycle and smaller genome (3.7 Gbp; Bennett and Leitch, 2001), C. thalictroides must now also be seriously considered as a candidate model. Although, the advantage of a smaller genome in this species is offset by polyploidy in relation to C. richardii (McGrath et al., 1994), the power of C. richardii as a genetic tool is most likely to be further enhanced by the availability of C. thalictroides as a closely-related, comparable, genetically-tractable species.
The advent of a high-efficiency stable transformation system in C. richardii and C. thalictroides removes one of the final technical barriers for the adoption of ferns as genetic models.
Sporophytes were subsequently transplanted to Sinclair potting growing medium (William Sinclair Horticulture Ltd., Lincoln, U.K.), typically when the third frond had expanded.
Callus tissue was cultured on 0.7% agar media (pH5.8) containing 1x Murashige and Skoog nutrients (Duchefa Biochemie, Haarlem, NL) and 2% sucrose, supplemented with hormone (5µM) and/or hygromycin B (40µgml -1 ) treatments (Sigma Aldrich, St. Louis, U.S.A.) as specified in the text. Attempts to grow callus on C-fern media were not successful (Supplemental Fig. S3). All hormone stocks were prepared to 1000x working concentration in 1N NaOH. 1000x hygromycin B was prepared in dH 2 O. Spores were sterilized by incubating for 10 minutes at room temperature in 2% sodium hypochlorite solution, 0.1% Tween, which was subsequently removed by six sequential rinses in sterile dH 2 O. Sterile spores were imbibed in dH 2 O and incubated for 48 hrs at room temperature in darkness before sowing. Gametophytes were fertilized between 9 and 11 days after germination by the application of sterile dH 2 O. Fertilization of transgenic T 1 gametophytes was not successful if grown under hygromycin selection. T 1 sporophytes were recovered through fertilization within T 1 gametophyte populations grown without hygromycin selection, and transgenic individuals subsequently identified though hygromycin selection on C-fern media. 20ugml -1 hygromycin is sufficient to kill untransformed young sporophytes within seven days (Supplemental Fig. S4). Transgenic sporophytes were removed from selection after 14 days and transplanted to soil.
Transgenic plants were regenerated from bombardment of two-week old callus tissue at 900 psi under vacuum conditions of 28 psi, with callus tissue placed at a distance of 6cm from the firing disc. Bombarded callus tissue was allowed to recover for three days before transfer to hygromycin selection. All mean efficiency values are expressed as a percentage of the bombarded population ± S.E.
Chloroform:IAA extraction was performed three times, and NaCl/EtOH precipitation twice.