cDNA-AFLP expression profiling

First we developed a synchronization protocol to monitor the cell morphogenesis-related gene expression in M. denticulata. The protocol was based on the observation that the majority of the cells grown in a 14-h light/10-h dark regime divided during the second dark period, after the growth medium of a stationary culture (obtained after 3-4 weeks) had been refreshed and, concomitantly, the cell density reduced at the start of the light period. Replacing the dark period by a light period enhanced the amount of synchronically dividing cells (Figure 1B). The effect of cell density on synchronization was significant (GLM; F-test; P < 0.001), with an optimal cell density below 80 cells mL^-1. Following synchronization, up to 85% of the cell population divided during an 8- to 9-h period, showing a sigmoid course (Figure 1C,D; Additional file 1). By sampling this period at five consecutive time points we obtained samples with different proportions of cells at the major morphogenetic stages (Figure 1A,C,D). cDNA-AFLP expression profiling of these samples allowed the assignment of differentially expressed genes to either the onset of cell division (Figure 1A2; Figure 2 (C1a and C1b)), the bulge (Figure 1A3; Figure 2 (C2)), the lobe (Figure. 1A4-A9; Figure 2 (C3)), or the doublet stage, during which the secondary cell wall is formed (Figure 2 (C4 and C5)). In total, the relative abundance was monitored of 4574 transcript-derived fragments (TDFs) during the cell growth of M. denticulata (Figure 3, Additional file 2), for which the expression patterns were altered visibly across time in 1420 and significantly (P<0.009; Q<0.05) in 476 TDFs. According to other studies [29, 30], we estimate that two-thirds of the mRNA population was sampled, implying that the real number of genes differentially expressed during cell growth of M. denticulata could be ~2100. A high similarity (E-value < 1.E-01 and similarity >50%) to database entries with assigned identities and unknown or hypothetical genes was found for 107 and 22 TDFs, respectively, mostly with Embryophytes and not with Chlorophyta. However, the majority of the TDFs (324 or 71.5%) showed no sequence similarity to any database entry (Figure 3; Additional file 3). Plausible explanations might be sequences too short to reveal any significant identity, short sequences representing non-conserved portions of genes, TDFs originating from the 3'-untranslated region of a gene, or TDFs representing genes specific to M. denticulata or streptophytic algae.

Of the 129 annotated genes, 118 clustered into six groups (designated C1a, C1b, C2, C3, C4, and C5) (Figure 2) according to the timing of their highest expression (Figure 1C,D). Except for one cluster consisting of six genes (cluster C1b; Figure 2), the expression profiles were reproducible in the two independent sampling series. The few genes not included in one of the described clusters typically showed narrow temporal expression patterns.

Based on their annotation, the TDFs were classified into 14 functional categories, named according to the Gene Ontology terminology (http://www.geneontology.org) (Figure 3; Additional file 3). The association between the functional category and the TDF clustering was not significant (χ² test; p = 0.070). The major group with a significant hit was involved in cell wall metabolism. The second largest category corresponded to sequences sharing significant similarity to unknown or hypothetical proteins.

Of 18 TDFs with similarity to genes involved in cell wall biogenesis or cell pattern formation, the RNA samples of the second cDNA-AFLP replication series and on an independently sampled series (Additional file 1) were analyzed by real-time quantitative reverse-transcription (qRT)-PCR. In general, the expression profiles obtained by cDNA-AFLP and qRT-PCR (Additional file 4) corresponded well (Additional file 5), confirming the obtained expression results.

Genes relevant for cell pattern formation

Seven TDFs could be identified that might be relevant for cell pattern formation in M. denticulata, among which two members of the Rab GTPase cycle and two members of the SNARE cycle of membrane fusion reactions. Rab8, similar to Md1852, is known to be involved in post-Golgi transport to the plasma membrane, inducing the formation of new surface extensions and believed to be regulated by a guanine nucleotide dissociation inhibitor [31] possibly corresponding to Md0818. Both Md1852 and Md0818 belonged to cluster C1a and, thus, had increased mRNA levels before the onset of mitosis. This observation might be related to the determination of the basic symmetry of a M. denticulata cell before mitosis, indicated by the development of a three-lobed semicell after removal of the nucleus [32]. In contrast, the SNARE cycle members were highly expressed in cluster C3, pointing to a role in further differentiation during the lobe stages for Md1404 (similar to plant syntaxin 32) and Md1560 (similar to a regulatory AAA-type of ATPase).

Two TDFs were identified encoding putative glycophosphatidylinositol (GPI) anchors: Md4071 and Md4341, belonging to clusters C1a, and C4, respectively. Among other properties, the function of a GPI anchor might be its dominant targeting to a specific membrane domain [33], possibly establishing a membrane template for morphogenesis. Md4341 turned out to be a 179-amino-acid protein containing a signal peptide and a fasciclin domain (a putative cell adhesion domain) (E-value 2.9E-07), with similarity to a fasciclin-like and an AGP-like protein from Brachypodium sylvaticum [CAJ26371.1] and Arabidopsis thaliana [AAM62616.1], respectively (Additional file 6).

Md3533 (cluster C3), similar to a very-long-chain fatty acid-condensing enzyme, might be involved in morphogenesis in accordance to the essential role in cell expansion during plant morphogenesis of Arabidopsis [34].

Genes involved in cell wall metabolism

A total of 30 cell wall-related genes were identified. Six TDFs operating in the monosaccharide metabolism, evenly distributed over C1 and C3, could be identified as UDP-pyrophosphorylases (Md1739, Md2333, and Md2565), a phosphoglucomutase (Md2842), a rhamnose synthase (Md1089), and a GDP-mannose 3,5-epimerase (Md3053). Nine polysaccharide synthesis enzymes all nearly clustered in C3, among which two cellulose synthases, Md0757 (see also [35]) and Md3668, and one cellulose synthase-like (CSL) gene of the CSLC family, Md2838. The exostosin family glycosyltransferases Md0450, Md1114, Md2144, and the glycosyltransferase Md0257 might synthesize the hemicellulosic or pectinous part of the cell wall and mucilage as well that is pectic in nature [11] and secreted simultaneously with cell wall material during cell growth [36]. Md3598 was the α-1,6-xylosyltransferase, typical of the hemicellulose biosynthetic pathway, whereas Md0888 was the xyloglucan endotransglycosylase/hydrolase (XET/XTH) that is a xyloglucan-modifying enzyme. The open reading frame (ORF) of Md0888 encoded a 277-amino-acid protein with a signal peptide and a GH16-XET domain (E-value 6.10E-37) and therefore designated MdXTH1. The catalytic site DEIDFEFLG, conserved among GH16 family members [37] and most seed plant XTHs [38] was present in MdXTH1 as xExDxEFxG and immediately followed by a potential N-glycosylation site NxT/S [39] (Additional file 7). The other 15 identified TDFs were involved in wall assembly, reorganization, and selective degradation. Four of them gave significant hits with expansins: MdEXP1 (C4), MdEXP2 (C4), MdEXP3 (C3), and MdEXP4 (C3). Whereas MdEXP4 and MdEXP3 were expressed during the early morphogenetic stages (C3), MdEXP1 and MdEXP2 were up-regulated during later stages (C4) (Figure 4). Changes in the internal structure of the cell walls, required for cell expansion, might be achieved by the release of hydroxyl radicals mediated by the class-III peroxidases Md0434 and Md0493. Peroxidase-generated hydroxyl radicals could cause non-enzymatic wall loosening by cleavage of various polysaccharides [40]. The ORF of Md0434 contained a secretion signal peptide and a Pfam peroxidase domain (E-value 2.50E-97) (Additional file 8). The H₂O₂ substrate for the peroxidase activity was probably generated by the glyoxal oxidases Md0606, Md1709, and Md3495. Hydrolytic enzymes included the pectinesterase Md4415, the endo-β-1,6-galactanase Md1480, and two members of cluster C5: the polygalacturonidase Md3500 and the β-glucosidase Md0559, possibly involved in degradation of a connecting zone between the primary and the secondary cell wall, thereby enabling shedding of the primary cell wall [41].

Phylogenetic relationship of M. denticulataexpansin-resembling proteins

As the involvement of expansins in cell growth of green algae had not been examined previously, we concentrated the experiments on this class of proteins. The full length characteristics of the M. denticulata expansin-resembling proteins (MdEXPs) are given in Additional file 9. MdEXP1 and MdEXP4 exhibited the highest sequence similarity (74% identity, 84% similarity) (Figure 5).

Phylogenetic analysis of the first dataset revealed that all MdEXPs were recovered as a monophyletic group with high support (BV = 99, PP = 1.00) (Figure 6A). The Micrasterias and Spirogyra sequences fell within the plant expansins and were most closely related to the EXPA family, with which they formed a well supported clade (BV = 86, PP = 1.00). The MdEXPs are recovered sister to the EXPA clade and the Spirogyra sequences form a paraphyletic assemblage, but the relationships between the Micrasterias and Spirogyra expansins and the EXPA clade are poorly supported. The high sequence divergence of expansins within and among Micrasterias and Spirogyra is shown by the relatively longer branches than those within the EXPA clade. In the second dataset, the putative expansin sequences of Chlorophyta formed a highly divergent clade, separated from the plant expansins by a very long branch (Additional file 10). Although the relationships between the Chlorophyta clade, the Dictyostelium clade and the plant expansin families were poorly resolved, the phylogenetic position of the Micrasterias clade, closely allied to the EXPA family, was well supported.

Domain organization of the M. denticulataexpansin-resembling proteins

The structural domain organization of the different MdEXPs was compared with the characteristic structural features of plant expansins (Table 1, Figure 5, Figure 7). A secretion signal peptide was present in all of them (Figure 5, Figure 7, Table 1). While the pollen-allerg-1 domain occurred in all proteins, except MdEXP4, the GH45 domain was found in MdEXP2 and MdEXP3 only, albeit with insignificant E-values. Nevertheless, in all sequences, a DPBB-1 domain was present, a rare lipoprotein A-like double-psi beta-barrel, to which GH45 belongs, and even twice in MdEXP2 (Additional file 11). The eight cystenyl residues forming disulfide bridges in fungal GH45 enzymes and maintaining their folded structure [16] were conserved in the expansin domain 1 of some of the plant expansin groups [22] and also in the MdEXPs (Figure 5). In M. denticulata, the GGACGY motif was present as GGSCGY/F, whereas the GxxCGxCF/Y motif in the same expansin domain 1 was fully conserved. A third motif characteristic for this domain, the Y/FRRVPC motif, varied among the MdEXPs (Table 1). The key residues of the GH45 catalytic site, conserved among EXPA and EXPB proteins (see Figure 5, indicated in bold), were absent. In land plant expansins, the pollen-allergen domain contains four conserved tryptophan residues that form part of the hydrophobic core of this domain [42] (Figure 5). In the MdEXPs up to two of these residues occurred and were fully conserved, when the structurally related amino acids phenylalanine and tyrosine are taken into account (Figure 5, Table 1). Although the highly conserved HATFYG motif near the N-terminus is characteristic of EXPA proteins [22], this motif could not be found in the MdEXPs. The EXPA and EXPB proteins were distinguished by the presence or absence of short stretches of amino acids at conserved positions at either side of the HFDL motif in the GH45 active site (α- and β-insertions) [16, 43]. According to the phylogeny, the MdEXPs contained an α-insertion characteristic of EXPAs, but they lacked the four highly conserved N-terminal residues 'GWCN' found in other EXPAs [16]. Of the HFDL motif, only the leucine residue was conserved (Figure 5). However, the long C-terminal extension of MdEXP2 was typical for EXLA proteins [22]. Although MdEXPs were heterogeneous and divergent, they clearly shared several characteristics of the EXPA protein domains, supporting our phylogenetic results.

Characteristic	MdEXP1	MdEXP2	MdEXP3	MdEXP4
Signal peptide	1-24	1-23	1-19	1-24
GH45 domain	No	39-180	52-211	No
Eight conserved cysteines	Yes	Yes	Yes	Yes
GGACGY motif	GGsCGY	GGsCGf	GGsCGY	GGsCGY
GxxCGxCF/Y motif	Yes	Yes	Yes	Yes
Y/FRRVPC motif	IYAVSC	FTQVPC	VTRVPC	VYAAGC
Catalytic site key residues	No	1 A	1 A	No
DPBB_1 domain	51-132	56-136; 302-385	82-161	51-132
Pollen_allerg_1 domain	144-223	147-226	172-253	No
Four conserved tryptophan (W) residues (* structurally related residues)	2(W) 1(F*) 1(Y*)	2(W) 1(F*) 1(Y*)	2(W) 2(Y*)	No
HATFYG motif (A)	No	No	No	No
α-insertion (A)	Yes	Yes	Yes	Yes
β-insertion (B)	No	No	No	No
HFDL motif (A, B)	Only L	Only L	Only L	Only L
CDRC motif (LA)	No	No	No	No
Long carboxy terminal extension (LA)	No	Yes	No	No

When a domain is present, its position is given (starting from the first methionine). A, unique characteristic of the EXPA family; B, unique characteristic of the EXPB family; LA, unique characteristic of the EXLA family; LB, unique characteristic of the EXLB family

Subcellular localization of the expansin-resembling MdEXP2and phenotypic changes due to its overexpression

The ORF of the M. denticulata expansin-resembling protein with the highest mRNA levels during cell growth, namely MdEXP2, was cloned into an overexpression vector to allow C-terminal fusions to the green fluorescence protein (GFP) [35]. As observed by confocal laser scanning microscopy of transiently MdEXP2-GFP-overexpressing interphase cells, the MdEXP2-GFP fluorescence occurred as motile cytoplasmic dots (Figure 8; Additional file 12) but could not be observed in the secondary cell wall itself, probably because of quenching due to a low apoplast pH [44]. Therefore, MdEXP2-GFP-overexpressing interphase cells were processed for transmission electron microscopy (TEM) and stained with GFP antibodies and protein A-gold to investigate whether the MdEXP2-GFP protein localizes into the secondary cell wall. Indeed, a positive signal was observed in the secondary cell wall, albeit not abundantly (Figure 9A,B), probably due to the instability of the GFP protein in this acid compartment [44]. In addition, mucilage vesicles still attached to distal Golgi cisternae (Figure 10A) and some released from the dictyosome (Figure 10C,D) were stained. This immunogold labelling indicated that the punctate pattern of the GFP fluorescence (Figure 8A) could correspond to Golgi-derived mucilage vesicles and that the fusion protein was directed to the wall via the endoplasmic reticulum-Golgi secretory pathway. No staining was observed in experiments for specificity control consisting of sections treated with protein A-gold alone (Figure 10B). In control sections of transgenic cells producing the free GFP, labelling occurred in the cytoplasm and was absent from the cell wall and cell organelles (Figure 9C,D).

Next, 26 independent transient transgenic cells were isolated and further analysed (Additional file 13). A group of cells lost the GFP-fluorescence within a few days and divided, resulting in normal daughter cells, while the majority of the cells died, possibly because of strong MdEXP2 overexpression as indicated by their bright GFP fluorescence. However, in eight independent cell lines, a range of phenotypes related to MdEXP2 overexpression during cell division and growth could be observed. Line 11 exhibited strong lobe elongation without loss of growth polarity after the first cell division (Figure 8B). The lobes were stretched and rounded instead of flattened at their tips. After the second cell division of line 11 and in all other cases (lines 6, 7, 8, 12, 13, 18, 19), the growth polarity was altered. Line 13 lost its planarity upon cell division and, thus, had the most severe phenotype. New semicells, without the characteristically lobed morphology, but almost without indentations, grew out three-dimensionally. Upon a new cell division of one of the daughter cells, the same phenotype was observed, whereas the newly formed semicells were also fused with each other (Figure 8F-I). In lines 6, 7, 8, 11 (from the second cell division onwards), 12, 18, and 19 axial but not radial elongation was impaired, resulting in semicells with a stunted polar lobe and fused lateral lobes (Figure 8C-E). Sometimes, the second division gave rise to a similar morphology (Figure 8D), but in most cases the phenotype was lost over one to two subsequent generations (Figure 8C). That all phenotypes still had the GFP signal and none of them resulted from control experiments with transgenic cells expressing only the GFP [35] suggests that they were related to the expression of the transgene.

Culture conditions, synchronization and sampling

A clonal Micrasterias denticulata culture was grown in twofold diluted Desmidiaceae medium [69] at 23°C and 120-140 μmol photons.m^-2.s^-1 under a 14-h light/10-h dark regime.

Two independent cultures were synchronized by replacing the growth medium of a stationary culture, diluting the density, and extending the light period to 24 h. Cells were sampled from synchronized cultures for RNA extraction at five consecutive time points during growth (T1 to T5) that were chosen to include for each time point a different proportion of cells at different morphogenetic stages (bulge, lobe, and doublet stage) (Figure 1C,D). Two independent stationary cultures served as control samples. Cells were concentrated by centrifugation for 1 min at 4°C and 4000 rpm and washed with phosphate buffered saline (PBS). Cell pellets were snap-frozen in liquid nitrogen and stored at -70°C.

RNA extraction and cDNA-AFLP analysis

Total RNA was isolated from approximately 80,000 frozen cells at each time point as described [70] with slight modifications. Instead of β-mercaptoethanol, 2 M stock solution of the anti-RNase agent dithiothreitol was added to the extraction buffer to a final concentration of 50 mM [71]. Cells were disrupted and homogenized in a bead mill (Retsch) (5 min at frequency 30 s^-1) with silicone-carbide sharp particles (Biospec Products) after the cell pellet had been thawed and suspended in the extraction buffer. Phytopure resin (GE-Healthcare) was added during the first chloroform:isoamylalcohol extraction to eliminate mucilage contamination of the RNA [72]. RNA samples were controlled qualitatively with the RNA 6000 Nano kit of the Bioanalyzer 2100 (Agilent Technologies) and quantified with the ND-1000 UV-Vis Spectrophotometer (Nanodrop).

Starting from 2 μg of total RNA, cDNA synthesis and cDNA-AFLP analysis with BstYI and MseI as restriction enzymes were done according to the procedures as described [28]. For the selective amplification, BstYI + C/T + 1/MseI + 2 primer pairs were used, resulting in 128 primer combinations. The cDNA-AFLP fingerprints were visualized with an autoradiography platform (PhosphorImager 445 SI; Molecular Dynamics).

Scanned gel images were quantitatively analyzed with the AFLP-QuantarPro image analysis software (Keygene N.V.). Expression values per gene were normalized for replicate effects by subtracting the replicate mean value (Additional file 2). Average linkage hierarchical clustering with the TMeV v4 software (http://www.tm4.org) and adaptive quality-based clustering (minimum two tags in a cluster, 0.95 significance level) [73] of the normalized expression data were carried out.

To assess the effect of the various cell division stages (T1-T5) on the gene expression during synchronized growth, a linear regression model of the form y = μ + rep + time + ε was fitted to the data, where y represents the raw expression values, rep and time the fixed replicate and time effects, respectively, and ε the random error. For all TDFs, a F-statistic was calculated, P-values were assigned to the main term time and subsequently transformed into false discovery rates, and Q-values [74] (Additional file 2). Besides the TDFs with a significant (Q<0.05) differential expression across the five time points, TDFs that were clearly absent in the stationary cultures but present during the synchronized growth, were excised from the dried gels, irrespective the significance of their differential expression across the five stages, followed by amplification and subsequent sequencing [28].

The TDFs were designated by Md (for M. denticulata) followed by a number corresponding to the AFLP fragment. After mutual alignment of the sequences, only the longest one of a group of identical sequences was retained, except when the TDFs displayed different expression profiles. Each sequence was identified by a similarity search against the public databases with the Blast2GO v1.7.2 program (http://www.Blast2GO.de) [75]. In addition to hits displaying an E-value < 1.E-03, hits with E-values between 1.E + 00 and 1.E-03 and a similarity > 50% were retained for further analysis.

Real-time qRT-PCR assay

Primers (Additional file 14) were designed with the Beacon Designer 7.0 (PREMIER Biosoft International) and the Oligo PerfectTM Designer (Invitrogen). Isolated RNA was treated with DNaseI (GE-Healthcare). An aliquot of 1 μg of total RNA from each sample was used for cDNA synthesis. The reverse transcription was carried out with oligo-dT primers and the SuperscriptTM II reverse transcriptase (Invitrogen) according to the manufacturer's instructions.

A set of reference genes was selected, based on their constitutive expression pattern during morphogenesis, to serve as a normalization factor in quantitative reverse-transcription-(qRT)-PCR analysis. Their expression stability (M) was analyzed with the geNormTM program [76]. Among 10 constitutively expressed TDFs, Md0789 (similar to a reticulon of Arabidopsis thaliana) and Md1473 (similar to peroxiredoxin 6 of Norway rat [Rattus norvegicus]) (M value = 0.421 and 0.489, respectively) were the two most stably expressed genes, followed by Md0386 (similar to the unknown gene of A. thaliana at5g13390 t22n19_40). To determine the number of internal control genes necessary for reliable data normalization, the pairwise variation value between two sequential normalization factors V_2/3 was calculated with geNormTM and turned out to be 0.151 under our experimental conditions, slightly higher than the cut-off value of 0.15. The inclusion of a fourth internal control gene resulted in an increase of the pairwise variation, yielding a V_3/4 value of 0.129. As a result, the use of the two or three most stably expressed genes was considered to be sufficient for reliable data normalization [76]. PCR fragments were amplified in triplicate on the LightcyclerTM 480 (Roche Applied Science) platform with SYBRTM Green QPCR Master Mix (Stratagene), according to the manufacturer's instructions with cycling conditions of 10 min preincubation at 95°C and 45 cycles at 95°C for 10 s, 58°C for 15 s, and 72°C for 15 s. Amplicon dissociation curves were recorded by heating from 65°C to 95°C. qBaseTM [77] was used for relative quantification.

RACE PCRs, protein domain identification, sequence alignment and phylogenetic analyses

The ends of the cDNAs were obtained by rapid amplification of cDNA ends (RACE) PCRs with plasmid DNA from a cDNA library of growth-synchronized M. denticulata (purchased from Invitrogen) as template. For MdEXP1, MdEXP3, and Md0434 only 5' RACE PCR was done, because the TDF contained the stopcodon and a part of the 3' untranslated region. For Md4341, MdXTH1, MdEXP2, and MdEXP4 both 5' RACE and 3' RACE were necessary. Gene-specific primers were designed with the eprimer3 program [78] and used in combination with vector-specific (pDONR222.1) primers (Additional file 15) in a PCR consisting of 1 min pre-incubation at 95°C and 30 cycles at 95°C for 30 s, 54°C for 30 s and 72°C for 2 min 30 s, followed by 1 cycle at 72°C for 5 min.

Protein domains in the ORF sequences were identified with the SMART tool (http://smart.embl-heidelberg.de/) [79, 80]. Signal sequences were confirmed with the SignalP 3.0 Server (http://www.cbs.dtu.dk/services/SignalP) [81, 82] and iPSORT prediction (http://ipsort.hgc.jp/) [83].

Similar sequences were retrieved from GenBank (http://www.ncbi.nlm.nih.gov) using protein BLAST and tblastx [84] (Additional file 16). The sequences were aligned using MUSCLE [85]. To remove signal peptides and C-terminal extensions, the alignment was trimmed from a conserved tryptophan near the N-terminus to a conserved phenylalanine near the C-terminus [17].

Two sets of alignments were considered for the phylogenetic analyses. The first dataset consisted of the four M. denticulata expansin-resembling proteins (MdEXPs), 26 land plant expansins representing the 17 orthologous clades within the four land plant expansin families [22], and six EST sequences of the streptophyte green alga Spirogyra pratensis that showed significant similarity to land plant expansins [27]. Five expansin-like sequences of the social amoeba, Dictyostelium discoideum, were selected as outgroup based on their inferred relationship with land plant expansins [16, 86]. This alignment was 227 amino acids long (Additional file 17). The second dataset included all sequences of the first alignment plus nine putative expansin genes found in four species of Chlorophyta (the sister clade of the Streptophyta) (322 amino acids long; Additional file 18) and was used to assess the phylogenetic position of other putative expansin sequences of green algae.

Models of protein evolution were selected with ProtTest 1.4 [87]. Maximum likelihood (ML) and Bayesian phylogenetic inference (BI) were analyzed under a WAG model of amino acid substitution with among site rate heterogeneity (gamma distribution with eight categories) for all datasets with PhyML v2.4.4 [88] and non-parametric bootstrapping (1000 replicates) to assess statistical support of internal branches with MrBayes 3.1.2 [89], respectively. Two parallel runs, each consisting of four incrementally heated chains were run for 2 x10⁶ generations, sampling every 1000 generations. Convergence of log-likelihoods was assessed in Tracer v1.4 [90]. A burn-in sample of 500 trees (well beyond the point at which convergence of parameter estimates had taken place) was removed before the majority rule consensus trees were constructed.

Overexpression of MdEXP2and microscopy

The ORF of MdEXP2 was cloned into the SpeI restriction site (ACTAGT) of the vector pSA405A under the control of the chlorophyll a/b-binding protein encoding gene of the desmid Closterium and was C-terminally fused to the green fluorescence protein gene (GFP) [35]. Primers were: forward primer 5'-ATGACTAGTATGAAAATCGGCATAATCCA-3' and reverse primer 5'-GGAACTAGTTAGGCACCCATTAACGGC-3'. The PCR was 2 min preincubation at 94°C and 5 cycles at 94°C for 45 s, 45°C for 45 s, and 68°C for 3 min, followed by 30 cycles at 94°C for 45 s, 55°C for 45 s, and 68°C for 2 min, and by 1 cycle at 72°C for 5 min. The recombined plasmid was introduced into M. denticulata by microparticle bombardment [35].

For confocal microscopy, a 100M microscope (Zeiss) was used, equipped with the LSM510 software version 3.2. Samples were scanned with a 20x (numerical aperture of 0.5) and a 63x water corrected objective (numerical aperture of 1.2). GFP fluorescence was visualized with argon laser illumination at 488 nm and a 500 to 530 nm band emission filter.

For transmission electron microscopy (TEM), a GFP antibody (Rb, (ab6556) Abcam) compatible protocol was followed to prepare the samples. MdEXP2-GFP-overexpressing M. denticulata cells were embedded in a yeast paste in a membrane carrier (Leica) and frozen immediately in a high-pressure freezer (EM PACT; Leica Microsystems). Freeze substitution was carried out in a Leica EM AFS instrument. Samples were infiltrated at 4°C stepwise in LR-White, hard grade (London Resin Company Ltd.) and embedded in capsules. Ultrathin sections of gold interference color were cut with an EM UC6 ultramicrotome (Leica) and collected on formvar-coated copper mesh grids. Grids were floated on blocking solution followed by incubation in a 1:25 and 1:10 dilution (in 1% bovine serum albumin in PBS) of primary antibodies (GFP antibody, (Rb, (ab6556) Abcam) for 60 min. The grids were labelled with protein A-10-nm gold (PAG10nm) (Cell Biology Department, Utrecht University). Control experiments consisted of treating sections with PAG10nm alone. Sections were post-stained in an automatic contrasting instrument (EM AC20; Leica Microsystems GmbH) for 30 min in uranyl acetate at 20°C and for 7 min in lead stain at 20°C. Grids were viewed with a 1010 transmission electron microscope (JEOL) operating at 80 kV.

Newly obtained sequence data were deposited in GenBank; the transcript derived fragments under accession numbers HE578289 to HE578716, the reference genes under accession numbers HE580226 to HE580228, and the open reading frames under accession numbers HE578717 to HE578726.