MERISTEM-DEFECTIVE, an RS domain protein, is required for the correct meristem patterning and function in Arabidopsis


(fax +44 191 334 1201; e-mail


Plant growth and development is dependent on the specification and maintenance of pools of stem cells found in the meristems. Mutations in the Arabidopsis MERISTEM-DEFECTIVE (MDF) gene lead to a loss of stem cell and meristematic activity in the root and vegetative shoot. MDF encodes a putative RS domain protein with a predicted role in transcription or RNA processing control. mdf mutants exhibit decreased levels of PINFORMED2 (PIN2) and PIN4 mRNAs, which is associated with a reduction in PIN:GFP levels, and with a defective auxin maximum in the basal region of the developing mdf embryo and seedling root meristem. Seedling roots also exhibit reduced PLETHORA (PLT), SCARECROW and SHORTROOT gene expression, a loss of stem cell activity, terminal differentiation of the root meristem and defective cell patterning. MDF expression is not defective in the bodenlos, pin1 or eir1/pin2 auxin mutants, and is not modulated by exogenous auxin. plt1 plt2 double mutants have unaffected levels of MDF RNA, indicating that MDF acts upstream of PIN and PLT gene expression. Differentiation of the shoot stem cell pool also occurs in mdf mutants, associated with a reduced WUSCHEL (WUS) expression domain and expanded CLAVATA3 (CLV3) domain. Overexpression of MDF leads to the activation of markers of embryonic identity and ectopic meristem activity in vegetative tissues. These results demonstrate a requirement for the MDF-dependent pathway in regulating PIN/PLT- and WUS/CLV-mediated meristem activity.


Animals grow by cell division in all organs, which contain stem cell populations for tissue renewal (e.g. Fuchs, 2008; Sambasivan and Tajbakhsh, 2007). In contrast, post-embryonic plant growth is driven by the expansion of cells derived from localized meristems, which contain stem cell-like cells that, as in animals, are the precursors of all differentiated cell types (Mayer and Jürgens, 1998). These meristems can remain active throughout the life of the plant, which in extreme examples can extend to centuries. Understanding the processes controlling meristem identity and activity is a major goal for developmental biologists.

In the model dicot Arabidopsis thaliana (L.) Heynh, the primary root and shoot meristems are established during embryogenesis (Laux and Jürgens, 1997). Auxin has been shown to be intricately involved in the embryonic patterning process that establishes the position and pattern of these meristems, and in particular the root meristem (Blilou et al., 2005; Friml et al., 2003). This patterning process is dependent on the establishment of a gradient of auxin concentration, mediated by its active transport, and requires the PIN family of auxin efflux facilitators/transporters. It is the directed localization of PIN proteins to the plasma membrane that determines the direction of auxin flow (Friml et al., 2003; Wisniewska et al., 2006). A localized auxin maximum develops in the region of the presumptive root meristem, and induces and restricts the expression of the PLETHORA gene family (PLT1, PLT2, PLT3 and BABY BOOM, BBM), which encode APETALA 2 domain transcription factors (Aida et al., 2004; Galinha et al., 2007). Together with the transcription factors SCARECROW (SCR) and SHORTROOT (SHR), which also establish radial patterns (Di Laurenzio et al., 1996; Helariutta et al., 2000), the PLT family establishes the position of the quiescent centre (QC): a small group of rarely dividing cells surrounded by the stem cells that give rise to the differentiated cells of the root (Aida et al., 2004; Galinha et al., 2007; Sabatini et al., 2003).

Following its establishment during embryogenesis, the position and size of the seedling root meristem, and the transit of cells from it, are maintained by a combination of an auxin concentration maximum and PLT gene action, maintained by the PIN proteins (Blilou et al., 2005; Galinha et al., 2007; Grieneisen et al., 2007). Subtle differences in the expression pattern of the PLT family members result in individual expression gradients across the root meristem, and PLT gene dosages define different outputs, from stem-cell identity to mitotic activity and cellular differentiation (Galinha et al., 2007). The PLT genes also regulate PIN gene transcript distribution, which further acts to stabilize the auxin maximum (Blilou et al., 2005). As a result of functional redundancy, mutations in individual PLT or PIN genes often display only very minor defects in root meristem organization and activity. However, double, triple and quadruple PLT mutants show various defects in stem cell identity and meristem maintenance (Aida et al., 2004; Galinha et al., 2007). Similar defects are also observed in single and multiple mutants in PIN family members, further demonstrating their role in patterning and maintenance of the root meristem (Blilou et al., 2005; Friml et al., 2003).

Multiple factors are therefore required for both patterning and maintenance of meristem activity. However, there is limited knowledge of the processes required to control the auxin–PLT regulatory loop. Here, we describe the analysis of the MERISTEM-DEFECTIVE (MDF) gene of Arabidopsis, encoding a novel putative serine–arginine (SR)-related arginine–serine (RS) domain protein. RS domain proteins in animals regulate diverse cellular processes, including cell division and cell structure, and in particular have molecular functions in transcriptional control, such as RNA splicing or transport (Boucher et al., 2001). MDF expression initiates early in embryogenesis (Casson et al., 2005), and is regulated independently of PLT gene activity and auxin. Our results indicate an essential role for MDF in regulating PIN and meristem transcription factor gene expression, and in establishing the correct auxin distribution, meristem pattern and function in Arabidopsis.


MDF is a nuclear protein expressed in the root meristem

We have previously used laser capture microdissection (LCM) and transcriptomics to analyse patterns of temporal and spatial gene expression during embryogenesis in Arabidopsis, with the aim of identifying genes required for meristem organization and identity (Casson et al., 2005; Spencer et al., 2007). The MDF gene (At5g16780) was found to be preferentially expressed in the basal region of globular- and heart-stage embryos (Casson et al., 2005). MDF encodes a predicted polypeptide of 820 amino acids, and in Arabidopsis only At3g14700 encodes a predicted protein (204 amino acids) sharing homology with MDF (45.9% identity at the amino acid level, limited to the C-terminal region of MDF). Putative orthologues of MDF exist in rice, brassica, medicago, vine and Physcomitrella.

BLAST searches revealed homology with the animal SART-1 family of proteins, with 41% identity between MDF and human SART-1 (hSART-1) in the 250-amino-acid C-terminal domains, but with only limited homology elsewhere. hSART-1 has been shown to be an SR-related protein of the U4/U6.U5 tri small nuclear ribonucleoprotein (tri-snRNP) complex of the spliceosome, and is required for correct spliceosome assembly (Makarova et al., 2001). As with hSART-1, the N terminus of MDF contains a putative RS domain characterized by a number of arginine residues alternating with serine, glutamate or aspartate dipeptides, indicating that MDF is a putative SR-related protein (Figure S1; Neugebauer et al., 1995; Blencowe et al., 1999). The MDF polypeptide is predicted to contain a C-terminal nuclear localization motif and an N-terminal RNA-binding domain, but no other conserved domains were identified (Figure S1; Prosite,; RNABindR, Terribilini et al., 2007).

To determine the spatial expression pattern of MDF in roots, we examined proMDF::GUS expression in seedlings. GUS activity was observed in the lateral root cap, columella and meristem, and was strongly active in the QC (Figure 1a). Expression was also found in young lateral root primordia, with weaker expression through the rest of the meristem and vasculature (Figure 1b). To determine the subellular localization of the MDF protein, proMDF::GFP:MDF protein fusion expression was determined in transgenic plants. Analysis of these plants by confocal laser scanning microsopy confirmed that the MDF polypeptide is preferentially localized to the nucleus in cells of the root meristem (Figure 1c,e): a control pro35S::GFP (lacking the targeting sequence) is localized to the cytoplasm (Figure 1d,f). A nuclear localization for MDF is also predicted by the Plant-PLoc protein localization software (Chou and Shen, 2007).

Figure 1.

 Expression of the MDF gene and protein in roots.
(a) proMDF::GUS expression in the primary root meristem at 3 days post-germination (d.p.g.), showing strong expression in the quiescent centre (QC; arrow).
(b) proMDF::GUS expression in a lateral root primordium and vasculature at 7 d.p.g.
(c, d) proMDF::GFP:MDF expression and localization to nuclei in the primary root meristem at 3 d.p.g.
(e, f) Expression and localization of pro35S::GFP to the cytoplasm in the primary root at 3 d.p.g. Scale bars: 50 μm (a–d); 25 μm (e, f).

MDF is required for organization and maintenance of the root meristem

To characterize the developmental role of the MDF gene, two T-DNA insertion alleles were identified. The mdf-1 allele contains an insertion within the ninth predicted intron (Casson et al., 2005), whereas the mdf-2 allele has an insertion within the ninth exon; exon 1, 5′ to the ATG start codon, is non-coding (Figure 2a). A wild-type MDF transcript could not be detected in plants homozygous for either mdf-1 or mdf-2, indicating that these are putative null alleles (Figure 2b). Homozygous mdf-1 and mdf-2 seedlings displayed similar phenotypes, being severely dwarfed, with typically three cotyledons, and a reduced root system (Figure 2c). The mutation was seedling-lethal, with 100% lethality by ca. 20–25 days post-germination (d.p.g.). This phenotype segregated with the T-DNA insertions in the MDF gene, and was complemented by the introduction of an MDF cDNA (proMDF::MDF), confirming this function.

Figure 2.

MDF gene structure and phenotypes of T-DNA insertion mutants.
(a) Schematic of the MDF gene organization showing the T-DNA insertions in mdf-1 (SALK_040710) and mdf-2 (SAIL_775_F10). ATG and STOP denote the putative start and stop codons, respectively, and boxes indicate exons. Triangles indicate the T-DNA insertion sites. Arrows indicate the position of oligonucleotides used for the verification of T-DNA inserts (see Experimental procedures).
(b) Semiquantitative RT-PCR analysis of MDF expression in wild-type seedlings (Col-0) and mdf-1 and mdf-2 seedlings at 7 d.p.g. ACT1 is the loading control.
(c) Phenotypes of mdf-1 and mdf-2 seedlings at 7 d.p.g. Scale bar: 5 mm.

To further characterize the function of the MDF gene we first investigated its role in root development. The primary root of the mdf-1 mutant is significantly shorter than that of wild-type seedlings (3.4 ± 0.2 mm versus 21.6 ± 1.2 mm; 7 d.p.g., n = 20, = 1.3 × 10−9). We examined roots at 2 and 7 d.p.g. to determine if this resulted from a reduction in either meristematic activity or cell length. At both time points both the size of the proximal meristem and the length of mature cortical cells were significantly shorter in mdf-1 mutants, indicating that both reduced meristematic activity and cell expansion contribute to the shorter roots of mdf mutants (Figure 3a). The reduction in size of the proximal meristem is also demonstrated by an analysis of the expression of the CYCAT1::CDB:GUS marker (Hauser and Bauer, 2000). The number of dividing cells, as detected by the CYCAT1::CDB:GUS marker, was highly variable in mdf-1 mutants (Figure 3c,d), compared with the wild type (Figure 3b). Moreover, the dividing cells occupied a smaller area in mdf-1, indicating a reduction in the population of dividing cells in the mdf-1 root meristem. It was also found that between the two time points the proximal meristem in mdf-1 mutants was reduced in size. No reduction in size of the proximal meristem was observed in wild-type roots or in the size of mature cortical cells in both the wild-type and mdf-1 mutant between 2 and 7 d.p.g. (Figure 3a).

Figure 3.

mdf mutants are defective in meristem maintenance and cell specification.
(a) Proximal root meristem (n = 10) and mature cortical cell (n > 50) measurements for Col-0 and mdf-1 at 2 days post-germination (d.p.g.) (meristem, P = 2.2 × 10−7; cortex, P = 8.7 × 10−11) and 7 d.p.g. (meristem, P = 1.3 × 10−9; cortex, P = 1.8 × 10−15). Error bars indicate SEMs.
(b–d) CYCAT1::CDB:GUS marker expression in wild-type (b) and mdf-1 (c, d) primary roots at 7 d.p.g. Brackets indicate the proximal meristem. Scale bars: 50 μm.
(e–j) GFP expression at 3 d.p.g. in primary roots of wild-type (e, g, i) and mdf-1 (f, h, j) carrying the enhancer traps J2672 (e, f), J2341 (g, h) and J1092 (i, j). Scale bars: 50 μm.
(k–n) QC25 expression in the roots of wild-type (k, m) and mdf-1 (l, n) seedlings at 5 d.p.g. (k, l) and 9 d.p.g. (m, n). Scale bars: 25 μm.
(o–t) Staining of amyloplasts in columella cells in roots of wild type (o, q, s) and mdf-1 (p, r, t) at 2 (o, p), 4 (q, r) and 10 d.p.g. (s, t). Scale bars: 25 μm.

To determine whether MDF is also required for correct root meristem patterning, we examined the expression of different tissue-specific markers in the mdf-1 mutant. Enhancer trap J2672 marks the endodermis, and was typically observed in isolated cells and in non-contiguous files of cells in mdf-1 roots (Figure 3e,f). Similar observations were made for enhancer traps J2341 and J1092, which mark the ground tissue and vascular initials (J2341; Figure 3g,h), and the lateral root cap and QC (J1092; Figure 3i,j), where again mdf-1 roots showed expression in isolated cells or a reduced expression domain. These results indicate an inability either to specify or maintain the root radial pattern and cell identity in mdf-1.

To determine whether these defects were the result of an inability to correctly specify QC cells, which act as a central organizer in the root meristem, we examined the expression of the marker QC25 (Sabatini et al., 2003) in mdf-1. At 5 d.p.g. QC25 activity was observed in both wild-type and mdf-1 root meristems (Figure 3k,l). However, in mdf-1 often only a single cell showed expression of the QC25 marker (Figure 3l). By 9 d.p.g. cells marked by QC25 expression were more distally positioned in the mdf-1 mutant compared with the wild type (Figure 3m,n), which appeared to result from a loss of root cap cells. To determine if this was the case, we examined the columella cells over a developmental time course in the mdf-1 mutant. As early as 2 d.p.g., the columella of mdf-1 lacks organization, and is not arranged in the distinctive tiers displayed by the wild type (Figure 3o,p). At 5 d.p.g. fewer cells have columella cell identity, as determined by the presence of starch granules (Figure 3q,r), and no columella cells are evident by 10 d.p.g., indicating a failure to specify columella cells associated with a terminal differentiation of the root meristem (Figure 3s,t).

MDF is required for the correct expression of PIN genes, auxin distribution and PLT-family expression

The establishment of an auxin gradient and maximum in the root is critical for correct meristem organization, and this is determined by correct PIN protein expression and localization (Blilou et al., 2005; Grieneisen et al., 2007). PIN4 and PIN2 in particular have significant roles in directional auxin flow through the root cap and epidermis, and in maintaining the auxin reflux loop (Friml et al., 2002; Wisniewska et al., 2006). To determine if the defects in cell specification and QC size in mdf-1 mutants are associated with defective PIN gene expression, transcription analysis and PIN:GFP localization was carried out.

Quantitative real-time RT-PCR showed that transcript levels of PIN2 and PIN4 were significantly reduced in mdf-1 mutants, although other members of the PIN family showed only minor changes in expression levels (Figure 4a). Transcriptional changes for these PIN genes were also reflected in defective protein fusion levels. proPIN2::PIN4:GFP expression was found, as expected, in the basal cells of the wild-type heart-stage embryo, but it was absent in heart-stage mdf-1 embryos (Figure 4b). Wild-type seedlings containing proPIN2::PIN2:GFP and proPIN4::PIN4:GFP constructs showed the expected localization of PIN2:GFP in the cortical and epidermal layers of the root (Figure 4c), and of PIN4:GFP in the meristem and columella (Figure 4d). In mdf-1 mutants, however, although the cellular polarity of the PIN2:GFP and PIN4:GFP appeared essentially correct, PIN2:GFP levels were much reduced in the epidermal layer (Figure 4c), and the PIN4:GFP expression domain was much reduced in the columella (Figure 4d).

Figure 4.

mdf mutants are defective in formation of an auxin maximum in the embryo and root meristem.
(a) Quantitative real-time RT-PCR analysis of PIN gene expression in wild-type and mdf-1 mutants, expressed relative to UBIQUITIN. These data are representative of two independent experiments using biological replicate samples. The error bars represent SEMs of three technical replicates.
(b) proPIN4::PIN4:GFP localization in the basal region of heart-stage embryos of wild-type (left panel) and mdf-1 (right panel). Scale bars: 25 μm.
(c) proPIN2::PIN2:GFP localization in roots of wild-type (left panel) and mdf-1 (right panel) seedlings at 3 days post-germination (d.p.g.). Scale bars: 50 μm.
(d) proPIN4::PIN4:GFP localization in roots of wild-type (left panel) and mdf-1 (right panel) seedlings at 3 d.p.g. Scale bars: 25 μm.
(e) DR5::GFP expression in the basal region of heart-stage embryos of wild-type (left panel) and mdf-1 (right panel) plants. Scale bars: 25 μm.
(f) DR5::GFP expression in the root meristem of wild-type (left panel) and mdf-1 (right panel) seedlings at 12 d.p.g. Scale bars: 25 μm.
(g) Semiquantitative RT-PCR analysis of MDF expression and ACT1 control following hormone treatment: Con, untreated control; NAA, treatment with 1 μm naphthalene acetic acid; BA, treatment with 1 μm benzyladenine; GA, treatment with 1 μm gibberellic acid.
(h) Semiquantitative RT-PCR analysis of MDF expression and ACT1 control in the auxin transport mutants pin1 (EN-2 background) and eir1 (Col-0 background).
(i) Histochemical localization of proMDF::GUS activity in roots of the auxin-insensitive bdl mutant and Col-0 wild type (WT). Scale bar: 100 μm.

To determine whether auxin distribution was also disrupted in the mutant, we examined expression of the auxin-responsive DR5::GFP reporter (Friml et al., 2003). In the mdf-1 heart-stage embryo, the typical expression of DR5::GFP in the incipient root meristem is absent (Figure 4e). DR5::GFP expression in mdf-1 seedling root tips was significantly reduced, and was restricted to the QC and central columella, with none detected in the lateral root cap as observed in the wild type (Figure 4f). Together, these data show that MDF function is required for the correct PIN2 and PIN4 gene expression and auxin distribution in the embryo, and in the seedling.

Given the role of phytohormones in root meristem organization and maintenance, and the defects observed in mdf mutants, we examined whether hormones regulate MDF expression. Exogenous treatment of seedlings with auxin, cytokinin and gibberellic acid did not significantly affect transcript levels or the spatial expression pattern of MDF (Figure 4g and data not shown), suggesting that MDF is not transcriptionally regulated by these hormones. Furthermore, MDF expression is not affected in the auxin transport mutants pin1 or eir1/pin2 (Figure 4h), nor is the mdf-1 root phenotype rescued by growth in the presence of auxins (data not shown). The auxin-insensitive bodenlos (bdl) mutant fails to establish a QC during embryogenesis (Hamann et al., 1999, 2002). Although the root of bdl mutants is highly disorganized, MDF is still expressed (Figure 4i). Together, these data show that MDF expression is regulated independently of these auxin transport and signalling pathways.

The establishment of an auxin peak around the QC and root cap, as well as correct auxin signalling, is required for the expression of PLT1 and PLT2 (Aida et al., 2004). Interestingly, roots of mdf mutants resemble those of plt1 plt2 double mutants. It is therefore possible that MDF acts upstream of the auxin distribution pathway, and of the PLT genes. To investigate this, we examined the expression of the PLT genes, and of other transcriptional regulators required for meristem specification and positioning, in mdf-1 mutants.

PLT1, PLT2 and the related AP2-like transcription factor BABY BOOM (BBM) all showed significant reductions in transcript levels in the mdf-1 mutant (Figure 5a). Similarly, the transcript levels of SCR and SHR, which are independent of the PLT genes and auxin, are reduced in mdf-1, consistent with the defective radial patterning and QC specification (Figure 5a). We also determined whether the failure to maintain the root meristem in plt1 plt2 double mutants is caused by a reduction in MDF expression. However, MDF transcript levels are not detectably affected in plt1 plt2 seedlings (Figure 5b). Furthermore, MDF expression was not dependent on SHR and SCR activity (Figure 5c), indicating that MDF acts independently of these meristem regulators.

Figure 5.

 Root meristem specification genes are downregulated in mdf-1 mutants.
(a) Quantitative real-time RT-PCR analysis of the expression of PLT1, PLT2, BBM, SCR and SHR relative to UBIQUITIN, in wild-type and mdf-1 seedlings at 7 days post-germination (d.p.g.). These data are representative of two independent experiments using biological replicate samples. The error bars represent SEMs of three technical replicates.
(b) MDF expression as determined by semiquantitative RT-PCR in plt single (plt1 and plt2) and double (plt1 plt2) mutants at 7 d.p.g. Col-0 is the wild-type control, and ACT1 is the loading control.
(c) MDF expression as determined by semiquantitative RT-PCR in shr and scr mutants at 7 d.p.g. Col-0 is the wild-type control, and ACT1 is the loading control.

MDF is required for regulating the shoot stem cell niche

Given the root meristem defects in mdf mutants, we asked whether MDF is also required to maintain the stem cell population of the shoot. Consistent with such a role, proMDF::GUS expression was found in the shoot apex (Figure 6a), and the mdf seedling shoot typically exhibits three cotyledons. WUS expression occurs below the stem cell population in the shoot apical meristem (SAM), and positively controls the size of the stem cell population in a non-cell autonomous manner (Brand et al., 2000; Schoof et al., 2000; Sharma et al., 2003). In mdf-1 mutants, the domain of WUS expression, determined using proWUS::nlSGUS, was consistently reduced compared with that found in wild-type seedlings (Figure 6b,c), which is consistent with a role for MDF in the SAM, and with the gradual loss of meristem activity observed in mdf mutants.

Figure 6.

MDF is required to maintain the correct meristem gene expression in the shoot.
(a) proMDF::GUS expression in the shoot apex at 7 days post-germination (d.p.g.).
(b, c) proWUS::nlsGUS expression in (b) wild-type (WT) and (c) mdf-1 shoots, showing a reduced expression in mdf-1 (arrows) at 7 d.p.g.
(d) proCLV3::nlsGUS expression (arrows) in the WT (left two seedlings) and in mdf-1 (right two seedlings) at 7 d.p.g.

WUS acts in a feedback loop with the CLAVATA genes CLV1, CLV2 and CLV3. They negatively regulate the shoot stem cell population, and a loss of WUS activity leads to an expansion of the CLV3 expression domain, which marks the stem cell niche in the shoot. Consistent with reduced WUS expression, CLV3 expression was upregulated in mdf-1 (Figure 6d). clv3-2 mdf-1 double mutants show a shoot phenotype intermediate between the two parental mutants, suggesting that MDF does not function directly in the same genetic pathway as CLV3 to regulate meristem activity (Figure S2).

MDF overexpression promotes stem cell-like identity

As mdf mutants are unable to maintain meristem activity, we investigated whether MDF expression is sufficient to promote stem cell identity or meristem activity. The approach was to overexpress ectopically the MDF cDNA under the control of the CaMV35S RNA gene promoter (Figure 7a). Seven independent pro35S::MDF lines were generated, and T1 and T2 lines of each were found to contain morphologically abnormal seedlings. Predominantly, these included seedlings that resemble those previously described following ectopic expression of the embryogenic identity gene LEC1, resulting in a turnip (tnp)-like mutant phenotype exhibiting ectopic accumulation of storage products (Figure 7b; Casson and Lindsey, 2006). Similarly, we found ectopic accumulation of oils at the root–hypocotyl junction, one characteristic of embryonic tissues (Ogas et al., 1997; Figure 7c–f). Other lines displayed ectopic shoot formation (Figure 7g), resembled mutants defective in the CUP-SHAPED COTYLEDON genes (Figure 7h; Aida et al., 1997) or completely lacked a root meristem (Figure 7i).

Figure 7.

pro35S::MDF seedling phenotypes showing activation of embryonic and stem cell-like phenotypes.
(a) Semiquantitative RT-PCR analysis of expression of MDF and (as loading control) ACT1 genes in the wild type (Col-0), and four examples of independent transgenic lines transformed with pro35S::MDF (1–4), showing variable levels of upregulation of MDF.
(b) pro35S::MDF overexpressing seedlings at 5 days post-germination (d.p.g.), showing a hypocotyls with swollen tnp-like phenotypes (arrows). Scale bar: 5 mm.
(c) tnp-like pro35S::MDF overexpressing seedlings at 8 d.p.g. stained with Fat Red to reveal lipid accumulation. The three seedlings on the left show lipid accumulation at the root–hypocotyl junction (arrows); the seedling on the right shows the wild-type phenotype. Scale bar: 5 mm.
(d) Higher magnification image of a root–hypocotyl junction of a phenotypically wild-type seedling (8 d.p.g.). Scale bar: 500 μm.
(e, f). Higher magnification images of a root–hypocotyl junction of pro35S::MDF overexpressors showing lipid accumulation (8 d.p.g.). Scale bars: (e) 1000 μm; (f) 500 μm.
(g, h) pro35S::MDF overexpressing seedlings at 5 d.p.g. showing ectopic shoot formation (g, arrow) and defective shoot apical meristem organization (h). Scale bars: 5 mm.
(i) pro35S::MDF overexpressing seedling at 5 d.p.g. showing an absence of the root meristem (arrow). Scale bar: 5 mm.

The phenotypes of these pro35S::MDF seedlings did not correlate with changes in the mRNA levels of the root meristem regulators PLT1, PLT2, SCR, SHR or the embryo-specific LEC1 transcription factor, nor was the expression of these genes influenced following inducible overexpression of MDF (data not shown). Furthermore, the penetrance of the phenotypes of the pro35S::MDF lines was not influenced by growth in the presence of phytohormones (data not shown), as has been demonstrated for pickle (Ogas et al., 1997) and tnp mutants (Casson and Lindsey, 2006). Taken together, this suggests that MDF is able to confer embryonic or meristematic identity on vegetative tissue.


Plant growth and development is dependent on the activity of meristematic groups of undifferentiated cells that provide the tissues for new organ growth. We have previously examined global transcriptional expression patterns in spatial domains of the developing Arabidopsis embryo in order to identify genes required for meristematic activity and organization (Casson et al., 2005; Spencer et al., 2007). The MDF gene was identified as being expressed from the globular stage of embryogenesis, and preferentially in the root pole (Casson et al., 2005). Here, we demonstrate that the MDF gene is required to maintain embryonic patterning and signalling, and post-embryonic meristematic activity, in both the root and the shoot.

MDF is required for root meristem organization and maintenance

DNA microarray data on microdissected embryos revealed that MDF is expressed to relatively high levels in the basal domain (Casson et al., 2005), and promoter::GUS fusion studies reveal expression is most strongly associated with meristematic tissue in both the seedling root and shoot. The relatively strong activity of the MDF promoter in the QC is seen more easily in the proMDF::GUS transgenics (rather than in the proMDF::GFP:MDF transgenics), most likely because of the ability to limit GUS histochemical staining by reducing the staining time, better revealing the relative levels of expression in different tissues. It is also possible that the GUS and MDF proteins have different stabilities. The expression pattern of MDF correlates with a role in meristem function, and the phenotype of mdf mutants confirms that MDF function is required for both embryonic and post-embryonic meristem patterning and maintenance.

The reduction in root length of mdf mutants is the result of both a reduction in the size of the proximal meristem and a reduction in mature cell length. Analysis of the cell division marker CYCAT1::CDB:GUS (Hauser and Bauer, 2000), along with the measurements of the proximal meristem, indicate that the population of dividing cells is reduced in mdf mutants. mdf mutants show reduced transcription and protein levels of PIN2 and PIN4, as determined by both qRT-PCR and by proPIN2::PIN2:GFP and proPIN4::PIN4:GFP expression. Both PIN2 and PIN4 are required for cell division control in the root meristem (Blilou et al., 2005; Friml et al., 2002), and it is likely that the reduction in size of the proximal meristem in mdf mutants is caused, at least in part, by reductions in PIN2 and PIN4 expression, resulting in a reduction in auxin levels in the meristem, as visualized by DR5::GFP expression.

The analysis of tissue-specific markers reveals that patterning is defective in mdf mutant roots. Patterning of the root meristem begins during early embryogenesis (Jenik et al., 2007), and MDF is expressed in the region of the presumptive root meristem at these stages of embryogenesis (Casson et al., 2005). A critical step in root meristem formation is the establishment of an auxin maximum in this region during embryogenesis. This involves an intricate balance between PIN and PLT gene expression, with PLT gene transcription regulated by auxin, and, in turn, PLT genes modulate PIN gene expression (Aida et al., 2004; Blilou et al., 2005; Friml et al., 2003). mdf mutants display a much reduced auxin maximum, which is first evident during embryogenesis, and later is evident in the seedling root. DR5::GFP and PIN4 expression are barely detectable during mdf embryogenesis; PIN2 and PIN4 transcript levels are reduced in the mdf seedling root, associated with reduced DR5::GFP expression. Expression of the PLT genes in the root is also downregulated.

The data are therefore consistent with mdf mutants being unable to establish and maintain an auxin peak in the root, presumably resulting from defects in the expression of PIN2 and PIN4, which are required for the auxin reflux loop (Friml et al., 2002; Blilou et al., 2005). In turn, this results in a failure to maintain the correct PLT gene activity, which is vital for both meristem patterning and maintenance (Aida et al., 2004; Galinha et al., 2007), and is consistent with the observation that mdf mutants have fewer QC cells. Given the intricate relationship between PIN and PLT gene activities, any perturbations in the levels of one may feedback and influence the other. However, plt1 plt2 double mutants do not have dramatically defective auxin distribution, placing PLT action downstream of auxin accumulation (Aida et al., 2004). The roots of mdf mutants phenocopy those of plt1 plt2 double mutants, suggesting that the consequent reduction in PLT gene activity leads to the terminal differentiation of the mdf root. The reduced expression of both SCR and SHR in mdf mutants is also consistent with the observed defective radial patterning and QC specification.

MDF function is also required for maintenance of the vegetative shoot meristem. mdf mutants typically produce several leaf primordia before the SAM arrests, at a similar time as the root meristem. The mutants exhibit a reduced WUS expression domain, coupled with an increased CLV3 expression domain. WUS is required to maintain stem cell activity, and acts in a feedback loop with CLV3 to maintain the shoot meristem (Brand et al., 2000; Schoof et al., 2000; Sharma et al., 2003). In both wus mutants and plants overexpressing CLV3, the SAM undergoes premature differentiation (Brand et al., 2000; Laux et al., 1996; Mayer et al., 1998; Schoof et al., 2000). The data are therefore consistent with MDF being required to maintain shoot meristem activity by regulating the crosstalk between WUS and CLV3 expression.

Auxin and cytokinin are intricately involved in the regulation of both meristem organization and activity (Blilou et al., 2005; Friml et al., 2003; Gordon et al., 2007). However, analysis of MDF expression by RT-PCR and proMDF::GUS seedlings, as well as plants expressing proMDF::GFP:MDF (data not shown), did not reveal any role for hormones in regulating MDF transcription, protein localization or stability. Furthermore, MDF expression was not affected in the auxin-insensitive bdl mutant, or in pin1 or eir1 mutants. This suggests that MDF acts independently of these pathways to maintain meristem activity.

MDF encodes a putative SR-related nuclear protein

MDF shares homology in its C-terminal domain with the hSART-1 protein (Makarova et al., 2001), and its SR-rich N terminus suggests that MDF is a member of the SR-related family of proteins. SR proteins function at a number of steps in the generation of mature RNAs, including roles in chromatin remodelling, transcriptional control, constitutive and regulated splicing and 3′-end processing, and cell division control (Blencowe et al., 1999; Boucher et al., 2001). They typically contain one or two N-terminal RNA recognition motifs (RRMs), and a C-terminal RS domain (Neugebauer et al., 1995) that mediates interactions with other RS domain-containing proteins, including SR-related proteins (Blencowe et al., 1999; Lorkovićet al., 2000). hSART-1 belongs to the SR-related family of proteins that are distinct from SR-family proteins, but contain RS domains (Blencowe et al., 1999). It has been demonstrated that hSART-1 is required for one of the final steps in spliceosome assembly, which is the association of the [U4/U6–U5] tri-snRNP (Makarova et al., 2001). In both humans and yeast, pre-mRNA splicing is inhibited by depletion of hSART-1 or the orthologous Snu66p (Gottschalk et al., 1999; Makarova et al., 2001). The N-terminal region of MDF contains a putative RS domain, and is also predicted to bind RNA (RNABindR, Terribilini et al., 2007).

A possible role for MDF in transcription or RNA processing would correlate with the observed nuclear localization of MDF. It is possible that quantitative changes in MDF levels alter processing events of key mRNAs regulating meristematic activity, via changes in the levels of key regulators such as the PIN, PLT, SCR or SHR genes. The reduced expression of these genes is considered unlikely to be simply an artefact of defective meristem structure or function, as the expression of MDF in the meristem-defective plt1 plt2 double, and scr and shr mutants, is unaffected (Figure 5b), and PIN4:GFP in the less severely abnormal mdf embryo is completely lost. Misregulation of the PIN, PLT, SCR or SHR genes would be expected to disrupt meristem function, but only the PIN family members are expressed in both root and shoot meristems. Furthermore, disruption of PLT expression (e.g. in plt1 plt2 double mutants; Aida et al., 2004) does not lead to the severe defects in auxin distribution seen in mdf. MDF expression is not itself regulated by meristematic patterning genes or hormones, and its ability to influence auxin-mediated patterning events in the root suggests that it acts either upstream or in parallel with these processes. This suggests that PIN genes, rather than PLT genes, are more likely to be direct targets for MDF action. However, it is also possible that MDF is a regulator of other genes that, in turn, regulate PINs and meristem-associated transcription factors.

Interestingly, pro35S::MDF plants show a range of developmental abnormalities that include phenotypic characters previously described for misregulation of embryonic and meristem regulators. In particular, pro35S::MDF seedlings exhibit a tnp-like phenotype, with a swollen hypocotyl and ectopic accumulation of storage oils, as has also been found in LEC1 overexpressors (Casson and Lindsey, 2006). These results implicate MDF as a positive regulator of embryonic identity pathways, or of stem cell activity. pickle mutants ectopically express LEC genes, and similarly accumulate oils (Dean Rider et al., 2003; Ogas et al., 1999). Overexpression of the meristem regulators WUS (Zuo et al., 2002) and the PLT-related BBM (Boutilier et al., 2002) similarly promotes embryonic/stem cell-like identity in vegetative tissues. Given that these genes are in part regulated by auxin, it is therefore possible that the phenotypes observed in pro35S::MDF lines are caused by the incorrect processing or retention in the nucleus of auxin regulatory RNAs (such as encoding PINs), mRNAs for these transcriptional regulators or their repressors. It has been previously demonstrated that mutations in the essential splicing factor SR45 protein cause defects in Arabidopsis development (Ali et al., 2007), whereas the overexpression of a truncated version of the SR-related AtSRL1 gene conferred increased tolerance to LiCl and NaCl (Forment et al., 2002). Therefore, quantitative changes in the levels of RS domain proteins are capable of changing diverse aspects of plant development. Future work will be directed towards the further elucidation of the mode of action of MDF.

Experimental procedures

Plant materials and analysis

For in vitro growth studies, seeds were vernalized, surface sterilized (Clarke et al., 1992) and plated on growth medium (half-strength MS medium; Sigma-Aldrich,, 1% sucrose and 2.5% phytagel (Sigma-Aldrich) at 22 ± 2°C, at a photon flux density of ∼150 μmol m−2 s−1. For hormone application experiments, seeds were germinated aseptically on growth medium, and at 7 d.p.g. were transferred to growth medium containing 1 μm of hormone for 16 h.

Arabidopsis seeds transgenic for QC25, pl1-4, plt2-2 and plt1-4 plt2-2 were kindly provided by Ben Scheres (Utrecht University,; DR5::GFP, proPIN2::PIN2:GFP and proPIN4::PIN4:GFP seeds were provided by Jiri Friml (VIB, University of Ghent,; proWUS::nlsGUS and proCLV3::nlsGUS seeds were provided by Thomas Laux (University of Freiburg,; bdl seed was provided by Gerd Jürgens (University of Tübingen,; and CYCAT1::CDB:GUS seed was provided by Marie-Theres Hauser (University of Agricultural Sciences, Vienna, Salk_040710, SAIL_775_F10, J1092, J2341, J2672, shr-2, scr-3, eir1 and pin1 seeds were obtained from the Nottingham Arabidopsis Stock Centre ( J1092, J2341 and J2672 are from the Haseloff enhancer trap GFP line collection ( Marker lines were crossed with MDF-1/mdf-1 heterozygous plants, and F2 progeny were examined.

T-DNA insertions in the MDF gene were confirmed using the oligonucleotides Lba1 5′-TGGTTCACGTAGTGGGCCATCG-3′ (salk_040710) or LB1 5′-GCCTTTTCAGAAATGGATAAATAGCCTTGCTTCC-3′ (SAIL_775_F10), in conjunction with the oligonucleotides 5′-CTGATGAGGCATCCAGGCTAC-3′ and 5′-CTTTCAGGCTTCCGCACATCTG-3′.


Light micrographs were taken using a CoolSNAP digital camera (Photometrics, with Openlab v3.1.1 (Improvision Ltd, on Leica MZ125 (Leica Microsystems,, Olympus SZH10 (Olympus, or Zeiss Axioskop (Carl Zeiss, microscopes. Root meristem and cell measurements were performed using Volocity v4.2 (Improvision). Confocal images were taken with a Bio-Rad Radiance 2000 microscope (Bio-Rad, after counterstaining tissues with 10 mg ml−1 propidium iodide. Staining of the columella root cap for the presence of starch was done by placing seedlings in Lugol’s solution (Sigma-Aldrich) for 5 min. Images were processed in Adobe Photoshop v7.0 (

Plasmid constructs and plant transformation

For proMDF::GFP:MDF, the MDF promoter and cDNA were inserted into the plant binary vector pGreenNtailGFP (kindly supplied by Patrick Hussey, Durham University,, and Col-0 plants were transformed. Oligonucleotides for promoter and cDNA amplification are: MDF-Prom, 5′-ACGGGCCCGAGAAACCAGATCCTTCTTC-3′ and 5′-ACGCGGCCGCGAACTTTAGTCAAGACCTAACCAC-3′; MDF-cDNA, 5′-CGGTCGACGAAGTGGAGAAGTCTAAATCAAGGC-3′ and 5′-GGGGATCCTCAAGGCTTTGGTCTCTTTGGTGG-3′. The CaMV35S promoter of pGreenNtailGFP was replaced with the MDF promoter as an Apa1Not1 fragment, and the MDF cDNA was inserted as a Sal1BamH1 fragment.

For pro35S::MDF, the MDF cDNA was amplified using the oligonucleotides 5′-CTGTCGACGAAAATGGAAGTGGAGAAGTC-3′ and 5′-CTGTCGACCAAAGACGCAAGCAAGACC-3′. A CaMV35S promoter and terminator cassette was cloned as an EcoR1 fragment into the binary vector pCIRCE (a gift from M. Bevan, John Innes Centre, The MDF cDNA was then cloned between the CaMV35S promoter and terminator as a Sal1 fragment.

For complementation of the mdf-1 mutant, the MDF promoter and cDNA were amplified using the oligonucleotides: MDF promoter, 5′-TCGGTACCGAGAAACCAGATCCTTCTTC-3′ and 5′-ACGGGCCCGAACTTTAGTCAAGACCTAACCAC-3′; and cDNA, 5′-CGGTCGACATGGAAGTGGAGAAGTCTAAATCAAGGC-3′ and 5′-GGGGATCCTCAAGGCTTTGGTCTCTTTGGTGG-3′. The MDF promoter was inserted into the pGreen 0179 vector ( as a Kpn1Apa1 fragment, and the MDF cDNA was inserted as a Sal1BamH1 fragment.

Plant transformation was performed by the floral-dip method (Clough and Bent, 1998) using the Agrobacterium tumefaciens C58C1 (Dale et al., 1989). Primary transformants were selected on growth medium supplemented with either 50 mg l−1 kanamycin (pCIRCE- and pGreenNtailGFP-based constructs) or 20 mg l−1 hygromycin B (pGreen 0179).

Gene expression analysis

Tissue localization of GUS enzyme activity was performed as described by Topping and Lindsey (1997). For semiquantitative RT-PCR, RNA was extracted from seedlings at 7 d.p.g. using the RNeasy Plant RNA Extraction kit (Qiagen, RT-PCR was performed using the OneStep RT-PCR kit (Qiagen), as detailed in the manufacturer’s instructions. Total RNA was treated with DNase according to the method of Sanyal et al. (1997), and 250 ng of RNA was used per reaction. The oligonucleotide pairs used for amplification were: ACT1, 5′-GATCCTAACCGAGCGTGGTTAC-3′ and 5′-GACCTGACTCGTCATACTCTGC-3′; MDF, 5′-CAGGCAGTTGCGCATCTTGTGG-3′ and 5′-GGAATTTGTGCGAGAGTAGTCG-3′; SCR, 5′-CACATTGCTGCTACAGTGTGC-3′ and 5′-CAGTAGAGTCGCTTGTGTAGC-3′; SHR, 5′-CACTACTCCCACCCAATACC-3′ and 5′-CATCCTCCTCGACCACTTCCTCG-3′.

Typical reaction conditions were 50°C for 30 min, 95°C for 15 min, followed by two cycles of 94°C denaturation for 30 sec, 65°C primer annealing for 30 sec, and a 72°C extension for 60 sec. This was followed by 20–40 cycles of 94°C denaturation for 30 sec, 55°C primer annealing for 30 sec, and a 72°C extension for 60 sec, with a final extension at 72°C for 7 min. Minus RT control experiments were performed by adding enzyme after the 50°C incubation.

For quantitative real-time RT-PCR, RNA was extracted from seedlings at 7 d.p.g. and was purified as described above. cDNA was synthesized using the SuperScript cDNA synthesis Kit (Invitrogen, Total RNA (5 μg) was used as the template, and oligo dT was used as the primer. The cDNA preparations were diluted 1:4 prior to using them as PCR templates. PCR reactions were carried out in a RotorGene 3000 (Corbett Research, Data was analysed using the Comparative Quantification program (Rotor-Gene v6.0.19). PCR reactions were set-up using the SYBER Green JumpStart Taq ReadyMix (Sigma-Aldrich), with 0.25 mm of each primer in a total volume of 20 μl. The PCR reaction conditions comprised an initial denaturation step of 95°C for 3 min followed by 40 repeats of 95°C for 20 sec, 55°C for 20 sec and 72°C for 20 sec. A melting curve protocol immediately followed the amplification cycles that comprised heating from 55 to 95°C in 1°C increments, holding for 5 s at each stage. Reactions were set-up in triplicate, and were repeated with two biological replicates. The abundance of each transcript is expressed relative to the UBIQUITIN transcript. The oligonucleotide pairs used for amplification were: PIN1, 5′-TATGAGATTTGTCGTTGGACCTGCC-3′ and 5′-CGCGATCAACATCCCAAATATCAC-3′; PIN2, 5′-CATGTGGAAATGGACCAAGACGG-3′ and 5′-GACCAAGCAAGGCCAAAGAGAC-3′; PIN3, 5′-GGAGCACCTGACAACGATCAAGG-3′ and 5′-GTTGACTTGCTTCGGCTCCTCCTAG-3′; PIN4, 5′-CAACGTGGCAACGGAACAATCTG-3′ and 5′-AGCCCTGCTGTAGCTTTCTCTATC-3′; PIN7, 5′-GTTTCGGACCGAGCTGGTCTTC-3′ and 5′-TTAGGCACTTCCTTTACCCTCTCC-3′; PLT1, 5′-TGTCGTTGTAGCAGCTTGTGACTC-3′ and 5′-CGTCTCAACAACGGCTAGTGCTC-3′; PLT2, 5′-CCATCAATATGGTGCAGCGAGC-3′ and 5′-CTGTGAAGTGTGACTAGAAGACTG-3′; BBM, 5′-ATTGCAGCAACAGCAGGAGAGG-3′ and 5′-CGTTGTAAGCAATCTCAGCAGCAGT-3′; SHR, 5′-GTAGAAGAAGAAGCTGATCTTGTCG-3′ and 5′-CACTATTCCTCATCCTCCTCGACC-3′; SCR, 5′-TCTTCAGGCTACAGGGAAACGTCT-3′ and 5′-GTGTGTGCATCAGAGCCAGTGACA-3′; MDF, 5′-GCGGAAGCCTGAAAGTGAAGATGT-3′ and 5′-TTCGACCTTCTCCTTTAGGGTTCCT-3′; UBQ, 5′-CACACTCCACTTGGTCTTGCGT-3′ and 5′-TGGTCTTTCCGGTGAGAGTCTTCA-3′.


The authors are grateful to the Biotechnology and Biological Sciences Research Council for financial support.