Arabinogalactan proteins 6 and 11 are required for stamen and pollen function in Arabidopsis


(fax +972 8 6472992; e-mail


Successful male reproductive function in plants is dependent on the correct development and functioning of stamens and pollen. AGP6 and AGP11 are two homologous Arabidopsis genes encoding cell wall-associated arabinogalactan glycoproteins (AGPs). Both genes were found to be specifically expressed in stamens, pollen grains and pollen tubes, suggesting that these genes may play a role in male organ development and function. RNAi lines with reduced AGP6 and AGP11 expression were generated. These, together with lines harboring point mutations in the coding region of AGP6, were used to show that loss of function in AGP6 and AGP11 led to reduced fertility, at least partly as a result of inhibition of pollen tube growth. Our results also suggest that AGP6 and AGP11 play an additional role in the release of pollen grains from the mature anther. Thus, our study demonstrates the involvement of specific AGPs in pollen tube growth and stamen function.


Stamens, consisting of a simple structured filament bearing an anther in which pollen is produced, are the male reproductive organs of flowering plants. The anther includes both somatic and reproductive tissues whose development is connected (Goldberg et al., 1993; Ma, 2005; Sanders et al., 1999). The sporogenous cells, found in the center of each of the four lobes of the anther, develop into microsporocytes, which, in turn, undergo meiosis to form tetrads of haploid microspores. These are then released into the anther locule to begin male gametophyte development (Goldberg et al., 1993; Ma, 2005; Sanders et al., 1999; Scott et al., 2004). The haploid microspores undergo asymmetric mitosis, resulting in the appearance of a large vegetative cell that completely engulfs a smaller generative cell. The generative cell undergoes mitosis to form two sperm cells (Ma, 2005; McCormick, 2004). When the flower opens, anther dehiscence occurs to release the pollen (Sanders et al., 1999; Scott et al., 2004). Once in contact with the stigma of the carpel, the pollen grains are hydrated, and the vegetative cell expands by polar growth, forming the pollen tube. As the tube extends, the vegetative cell nucleus and the two sperm cells move into the tube. The pollen tube grows along the female style and enters the ovule, with the sperm cells being released into the embryo sac (McCormick, 2004).

Pollen tube extension is enabled by the continuous fusion of Golgi-derived vesicles, providing membrane and wall components to the tip of the cell (Hepler et al., 2001; Lord, 2000). The vesicles are transported via directional cytoplasmic streaming, which is dependent on the organization of the F-actin cytoskeleton (Hepler et al., 2001; Lord, 2000). The primary pollen tube wall consists primarily of pectin, together with hemicellulose and cellulose, while the secondary wall is composed of callose. However, the callose lining is absent from the tube tip. Other components of the pollen tube cell wall include extensin, pollen extensin-like proteins and arabinogalactan proteins (AGPs; Hepler et al., 2001; Lord, 2000; Taylor and Hepler, 1997). AGPs are hydroxyproline-rich proteoglycans, usually containing 90-99% carbohydrate and only 1-10% protein. In AGPs, large branched polysaccharide chains (arabinogalactans), as well as short chains of arabinose (oligoarabinosides), are attached to the protein backbone, resulting in a highly glycosylated, complex macromolecule (Gaspar et al., 2001; Majewska-Sawka and Nothnagel, 2000; Rumyantseva, 2005; Seifert and Roberts, 2007; Showalter, 2001).

AGPs are widely distributed within the plant kingdom, and are expressed in a developmentally regulated manner in various cell types and tissues (Majewska-Sawka and Nothnagel, 2000; Serpe and Nothnagel, 1999; Showalter, 2001). AGPs are found at the cell surface, attached to the outer surface of the plasma membrane or bound to the cell wall. Accordingly, AGPs may contain a C-terminal glycosylphosphatidylinositol (GPI) anchor signal sequence (Gaspar et al., 2001; Schultz et al., 1998). GPI anchors provide an alternative to transmembrane domains for anchoring proteins to the plasma membrane, and possibly enable AGPs to participate in signal transduction pathways (Gaspar et al., 2001; Schultz et al., 1998). Alternatively, AGPs can be secreted from the cell (Serpe and Nothnagel, 1999). The complex structure, distribution and localization of AGPs indicate that these compounds probably serve important functions in plants. Indeed, there is evidence indicating that AGPs are involved in cell-to-cell signaling and extracellular matrix interactions during cell differentiation, tissue development and somatic embryogenesis (reviewed by Majewska-Sawka and Nothnagel, 2000; Rumyantseva, 2005; Schultz et al., 1998; Seifert and Roberts, 2007; Showalter, 2001).

In recent years, studies have begun to link various functions to specific AGP genes. Arabidopsis thaliana xylogen protein 1 (AtXYP1) and AtXYP2 act redundantly as mediators of cell–cell interaction in vascular development (Motose et al., 2004). Arabidopsis AGP17 has been shown to play a role in Agrobacterium-mediated transformation, possibly through involvement in a signaling pathway(s) (Gaspar et al., 2004). Overexpression of tomato LeAGP1 resulted in shorter, highly branched plants producing fewer fruit (Sun et al., 2004). Interestingly, several AGP genes from various plant species have been shown to encode proteins required for cell division and/or expansion in various developmental processes (Acosta-Garcia and Vielle-Calzada, 2004; Lee et al., 2005; Park et al., 2003; Shi et al., 2003; Yang et al., 2007).

AGPs are associated with several tissues of the developing anther (Coimbra et al., 2007; Otegui and Staehelin, 2004; Peng et al., 2005; Qin et al., 2007) and the pollen tube cell walls (Abreu and Oliveira, 2004; Ferguson et al., 1999; Hepler et al., 2001; Mollet et al., 2002; Qin et al., 2007). In addition, microarray studies have demonstrated elevated and specific expression of AGP genes in stamens and pollen (Amagai et al., 2003; Lalanne et al., 2004; Wellmer et al., 2004; Zik and Irish, 2003). A function for AGPs in cell-wall deposition during pollen tube growth is supported by the results of experiments performed with several plant species showing that perturbation of AGPs interferes with wall assembly at the pollen tube tip and inhibition of pollen tube elongation (Abreu and Oliveira, 2004; Qin et al., 2007; Roy et al., 1998). In addition, Mollet et al. (2002) have shown that the site of new emerging pollen tips in lily can be predicted by the location of AGP secretion. However, the involvement of AGPs in pollen tube growth has so far been demonstrated only indirectly, using antibodies to carbohydrate epitopes on AGPs or by perturbing AGPs with β-glucosyl Yariv reagent [Y(β-Glc)3], a dye that specifically interacts with AGPs (Schultz et al., 1998).

In addition to indications for participation of AGPs in pollen function and pollen tube growth in particular, it has also been shown that glycosylphosphatidylinositol (GPI) anchor biosynthesis is required for normal pollen germination and tube growth. Mutations in SETH1 and SETH2, which encode proteins involved in GPI biosynthesis, specifically block male transmission and pollen function (Lalanne et al., 2004). However, SETH1 and SETH2 are not specifically expressed in pollen. Thus, it was suggested that the observed defects in pollen function are probably due to loss of the GPI anchor from target proteins fulfilling specific roles in pollen growth and development (Lalanne et al., 2004). However, it is still unknown which of the 47 predicted GPI-anchored pollen proteins are important in these processes (Lalanne et al., 2004; McCormick, 2004).

AGP6 (At5g14380) and AGP11 (At3g01700) are two highly similar classical AGP genes. The two genes each comprise a single exon, 452 and 410 bp in length, respectively (Figure 1a), and are highly similar, sharing 67% identity at the nucleotide level and 65% identity at the amino acid level (Figure 1b). The two encoded proteins include three features of classical AGPs, i.e. an N-terminal secretory signal peptide, a region rich in the amino acids proline, hydroxyproline, alanine, serine and threonine, and a signal for addition of a GPI anchor at the C-terminus (Schultz et al., 2000). The tri-amino acid sequence that is the signal for addition of a GPI anchor is identical in the two genes (Figure 1b). Several studies relying on Northern blot analysis, RT-PCR and microarrays have shown the expression of AGP6 and AGP11 to be floral-specific, with the two genes being expressed in stamens, pollen and pollen tubes (Amagai et al., 2003; Lalanne et al., 2004; Pereira et al., 2006; Zik and Irish, 2003).

Figure 1.

 Genomic structure and the encoded protein sequences of AGP6 and AGP11.
(a) Genomic structure of AGP6 and AGP11. The grey boxes represent the coding regions, the white boxes correspond to the 5′ and 3′ UTRs, as indicated, and the grey line on the left represents the upstream region of AGP6. The black bars indicate regions that were used in generating the AGP6 AGP11 RNAi construct. The ET5767 enhancer line contains a DsE element inserted 740 bp upstream of the first ATG of AGP6, and the ET3978 enhancer line contains a DsE element inserted in the coding region of AGP11 97 bp downstream of the first ATG.
(b) Local alignment of AGP6 and AGP11. The amino acid sequences of AGP6 and AGP11 were aligned using clustal w (Thompson et al., 1994). Amino acids that are changed in the TILLING lines are highlighted in grey, and the new amino acids are indicated above them. Numbers indicate the position of the amino acid in the AGP6 protein sequence. The GPI anchor addition sequences are boxed.

In this study, we show that AGP6 or AGP11 are required for pollen function. Pollen of plants harboring mutated AGP6 or AGP11 failed to develop elongated tubes, both in vitro and on intact carpels, resulting in reduced male fertility. In addition, we report that severe reduction in AGP6 or AGP11 expression leads to defects in pollen release.


AGP6 and AGP11 are expressed in stamens, pollen and pollen tubes in a developmentally regulated manner

To follow AGP6 and AGP11 temporal and spatial expression in detail, plants expressing the β-glucoronidase reporter gene (GUS) or red fluorescent protein (RFP) under the control of the AGP6 and AGP11 promoters (AGP6pro:GUS/RFP and AGP11pro:GUS/RFP, respectively) and the AGP6-encoded protein fused to GUS under the control of the AGP6 promoter (AGP6pro:AGP6-GUS) were produced and analyzed. As encoded by these constructs, AGP6 and AGP11 showed the same expression pattern (Figure 2a–e and f–j, respectively). GUS expression was first detected in stamens at floral stage 10 (Figure 2a,b and f,g; stages defined according to Smyth et al., 1990). At this stage, individual microspores were released and generated an exine wall, but the anther continued to grow (Sanders et al., 1999). AGP6 and AGP11 expression at this stage was confined to pollen sacs. GUS expression became stronger at later stages of stamen development (Figure 2c,h), and was detected throughout the intact anther tissue, as well as in floral sections (Figure 2d,i). Such expanded expression was not a result of GUS diffusion, as RFP expression was also evident in anthers bearing constructs encoding RFP under the control of the AGP11 promoter (Figure 2k–n). Expression was strongest in mature pollen and the degenerating tapetum (Figure 2d,i,m). Expression of AGP6 and AGP11 in pollen tubes was apparent, as revealed by the GUS staining of AGP6pro:AGP6-GUS and AGP11pro:GUS pollen tubes growing through the stigma and transmitting tract of wild-type gynoecium (Figure 2e,j). Strong expression was also seen in the pollen grains and pollen tubes of in vitro-germinated AGP11pro:RFP (Figure 2o,p) and AGP6pro:RFP-expressing pollen (data not shown). As in previous studies (Lalanne et al., 2004; Pereira et al., 2006; Zik and Irish, 2003), AGP6 or AGP11 expression was not detected in any other part of the flower or plant (data not shown).

Figure 2.

AGP6 and AGP11 expression patterns.
(a–d, f–i) Flowers expressing the GUS protein or the AGP6–GUS fusion protein (d) under the control of the AGP6 (a–d) or AGP11 (f–i) promoter.
(b, g),(c, h) Close-up views of young (stage 10; Smyth et al., 1990) (b, g) and mature flowers (c, h).
(d, i) Longitudinal sections of mature flowers.
(e, j) GUS staining in pollen tubes of AGP6pro:AGP6-GUS (e) and AGP11pro:GUS (j) pollen germinating on stigmas of wild-type plants.
(k–p) RFP expression in AGP11pro:RFP stamens and pollen tubes. (k) Stamens of AGP11pro:RFP flowers, compared to the wild-type (l), under red fluorescence (insets show the stamens in white light).
(m, n) Longitudinal sections of AGP11pro:RFP (m) and wild-type (n) flowers, under red fluorescence. AGP11pro:RFP (o) and wild-type (p) in vitro-germinated pollen, as visualized by confocal microscopy, using red fluorescence (insets show the same view under white light).
St, stamen; Ca, Carpel; pt, pollen tube.

Generation of AGP6 and AGP11 mutant lines

To date, to the best of our knowledge, no insertion plant lines harboring a T-DNA insert in the coding region of AGP6 are publicly available, although a line containing an insertion 740 bp upstream of the first ATG of AGP6 (ET5767; Sundaresan et al., 1995; Figure 1a) does exist. However, as detected by semi-quantitative RT-PCR, no significant reduction in AGP6 expression was detected in plants homozygous for the insertion, compared to the expression in wild-type plants (data not shown). By contrast, in the case of AGP11, an insertion line harboring an insertion 97 bp downstream of the first ATG (ET3978; Sundaresan et al., 1995; Figure 1b) showed no expression of AGP11 by semi-quantitative RT-PCR analysis (data not shown). However, this loss-of-function allele has no apparent morphological phenotype.

To obtain plants with reduced AGP6 and AGP11 expression, plants expressing an AGP6 AGP11 RNA interference (RNAi) construct were produced. Real-time RT-PCR analysis revealed that four independent AGP6 AGP11 RNAi lines showed significant reduction in expression of these genes (between 70% and 96% reduction in the various compared to the wild-type; Figure 3a). The modified plants developed siliques that were shorter than in the wild-type and contained far fewer seeds (Figure 3b). To quantify the reduction in plant fertility, the seed content of the siliques was determined. This analysis revealed that the AGP6 AGP11 RNAi lines developed siliques with variable number of seeds, probably as a result of the somatic variability of the RNAi silencing (Kerschen et al., 2004; Meyer and Saedler, 1996). Nonetheless, although 37% of the siliques in wild-type plants contained 30–60 seeds, 56% contained between 16 and 30 seeds, and only 7% included fewer than 16 seeds (Figure 3c), the AGP6 AGP11 RNAi lines had a much higher percentage of siliques with fewer than 16 seeds (21%, 36% and 57% in lines 4.1.1, 14.2 and 4.1.4, respectively; Figure 3c). Line 4.3.2, which showed the most significant reduction in both AGP6 and AGP11 transcript levels (96% reduction; Figure 3a), developed siliques that all contained fewer than 16 seeds (Figure 3c). Indeed, many of the siliques in plants from this line did not develop at all. Other than reduced fertility, however, plants from the RNAi lines presented normal morphologies.

Figure 3.

AGP6 AGP11 RNAi lines.
(a) Real-time RT-PCR analysis of AGP6 and AGP11 expression in AGP6 AGP11 RNAi lines. Relative levels of AGP6 and AGP11 expression in RNAi lines are shown as percentages of the expression in wild-type (WT) plants, defined as 100%. The numbers above the bars indicate the mean percentage of expression. Standard errors for three independent experiments are also shown.
(b) Siliques of AGP6 AGP11 RNAi plants, compared to the wild-type strain. Representative mature siliques from each RNAi line were collected and cleared to assess seed content. A typical wild-type silique is shown on the left of each panel. Siliques of AGP6 AGP11 RNAi plants are shorter and contain fewer seeds. The severity of the phenotype is correlated with the reduction in AGP6 and AGP11 transcript levels (see text for details).
(c) Histogram showing the distribution of siliques containing 0–15, 16–30 and 31–60 seeds per silique in wild-type (WT) and AGP6 AGP11 RNAi lines. Standard errors are shown (= 100 siliques per line).

The AGP6 AGP11 RNAi construct may have affected the expression of other pollen-expressed AGPs. To test this possibility, a list of all Arabidopsis AGP genes was downloaded from TAIR website ( The sequences of genes reported to be expressed in stamens and pollen were analyzed to identify those that contained stretches of at least 21 bp (Wang and Waterhouse, 2002) that completely matched the AGP6 and/or AGP11 sequences used for the RNAi construct (Figure 1). Two genes, namely AGP1 and AGP2, fit these criteria, and thus their mRNA levels in the various AGP6 AGP11 RNAi lines were assessed by real-time RT-PCR (data not shown). AGP2 mRNA levels were not significantly reduced in most lines except line 4.1.1, which showed the least severe phenotype among the various RNAi lines. AGP1 mRNA levels were more significantly reduced (between 25% and 68%). Nonetheless, it is difficult to determine whether this reduction contributed to the degree of phenotype severity, as it correlates withAGP6 and AGP11 mRNA levels in the various lines.

In addition to the AGP6 AGP11 RNAi lines, stable agp6 loss-of-function alleles were obtained. Towards this end, the services of the Seattle Tilling Project ( were used to screen for lines with point mutations in the coding region of AGP6. Eight TILLING alleles with point mutations in AGP6 were obtained (Table 1 and Figure 1b). Three of the mutations were silent. The other five alleles harbored mis-sense mutations, resulting in changes to amino acids (Table 1 and Figure 1b). The severity of the effect on protein function caused by the various mis-sense mutations was predicted to differ. Based on the ‘position specific scoring matrix’ (PSSM; and the ‘sorting intolerant from tolerant’ (SIFT) score (Ng and Henikoff, 2003; Table 1), mutations P124L, P80L and P28L (named according to the amino acid changed) were all predicted to be significantly damaging to AGP6 function. Although the mutation in line P78L was not expected to be highly damaging based on these predictions, such plants showed a clear mutant phenotype (see below). The mutation in line D117N, on the other hand, was not likely to cause a significant change in protein function. As such, lines harboring the silent mutations and line D117N were not further analyzed. Plants of line P80L presented an abnormal dwarf phenotype that was not seen in any of the other TILLING lines. As this additional phenotype was probably due to other point mutations in the background of the line, plants from this line were not further characterized either.

Table 1. agp6 TILLING lines
Nucleotide changeEffectaRestriction enzyme differences from REBASEbPSSM differencecSIFT scored
Gained in variantLost from reference
  1. aAn equals sign (=) indicates a silent mutation.

  2. bA restriction site that was gained or lost as a result of the nucleotide change.

  3. cThe PSSM (position specific scoring matrix) score corresponds to the severity of the change, based on matrices with particular values for a specific base change. A PSSM score >10 indicates a significant change.

  4. dThe SIFT score determines the effect of the amino acid change on the protein function, using the homology between sequences and physical characteristics of the amino acid. A SIFT score <0.05 indicates a significant change in protein function. Based on these scores, the lines with the most significant changes were chosen: P80L and P124L (marked with an asterisk) and two other lines: P28L and P78L.

  5. eThis restriction site was gained in the mutant during PCR amplification using a forward primer, designed according to the dCAPS method (Neff et al., 1998).

  6. Enzymes printed in bold were used in genotype analysis of the lines.

C245TP28LAlw26Ie  0.05
C395TP78LHinfI, PleIAsuI, AvaII, DraII, NlaIV, PpuMI7.50.43
C401TP80LHin4I, MnlI, PspXI, SmlI, TaqI, XhoIFinI, SecI13.6*0.13
C456TV98= BseRI, BsmAI, Esp3I  
G511AD117NBsrDIBccI, BtgZI 0.28
C533TP124LBce83I, Hpy178III, SmlI, TspRIBsiYI, HphI, TaqII14.8*0.00
C612TF150=MboII, MseIDdeI  

The P124L, P28L and P78L plants were genotyped using a PCR-based method that distinguishes the wild-type strain from the mutant alleles (see Experimental procedures for details). Lines P124L, P28L and P78L were self-crossed to isolate both heterozygous and homozygous plants. However, even after several rounds of self-crossing, homozygous plants for line P124L could not be isolated, although the P124L mutant gametes were shown to be transmitted to the progeny (data not shown). These TILLING mutants might carry additional mutations and possibly another unknown mutation in a gene linked to the agp6 mutation that could cause homozygous lethality.

Similar to the AGP6 AGP11 RNAi lines, the agp6 mis-sense mutants developed short siliques with far fewer seeds than in the wild-type (Figure 4). While the mean number of seeds in wild-type (Big Mama; Torii et al., 1996) siliques was 36, this was reduced to 5 and 13 seeds in homozygous lines P28L and P78L, respectively (Figure 4g). As expected, the phenotype of the homozygous plants of line P78L was more severe than that of the heterozygous plants (Figure 4a,b). The silique phenotype of heterozygous P124L plants was even more severe (a mean of three seeds per silique; Figure 4d,g), suggesting that the mutation in amino acid 124 was more disruptive to AGP6 function. Representative wild-type and mutant siliques (Figure 4e,f, respectively) are shown at a larger magnification to illustrate the failure of seeds to develop in the mutant.

Figure 4.

 Siliques of agp6 TILLING lines, compared to wild-type plants.
(a–e) Representative mature siliques from each agp6 TILLING line were collected and cleared to show seed content. A typical wild-type silique is shown on the left of each panel.
(a) Plants homozygous for mutation P78L.
(b) Plants heterozygous for mutation P78L.
(c) Plants homozygous for mutation P28L.
(d) Plants heterozygous for mutation P124L.
(e) Wild-type silique mechanically opened to show the developed seeds within the silique.
(f) An open, representative silique from plants heterozygous for the P78L mutation. Arrows indicate aborted ovules.
(g) Histogram showing the mean number of seeds per silique in wild-type (WT) and agp6 lines. The mean numbers of seeds are indicated above the bars. Standard errors are also shown (= 100 siliques per line).

The fact that mutations in agp6 alone result in a mutant phenotype while loss of function of AGP11 does not, demonstrates that AGP6 assumes a dominant function over AGP11. However, AGP11 may play a redundant role, such that loss of both gene functions might lead to a stronger mutant phenotype.

Loss of function of AGP6/AGP11 results in reduced pollen tube growth

To examine the source of reduced fertility in the agp6 mutants and AGP6 AGP11 RNAi lines, pollen function was tested. The percentage pollen germination in vitro, as well as pollen tube growth, was compared between mutant and wild-type (Landsberg erecta (Ler) pollen for the RNAi lines and Big Mama pollen (Torii et al., 1996) for the TILLING lines. Germination (i.e. emergence of a pollen tube) was not significantly different between wild-type and mutant pollen grains (Figure 5). However, mutant pollen developed significantly shorter pollen tubes. In wild-type Ler, 91% of the pollen tubes were longer than 100 μm, while in RNAi lines 4.1.1 and 4.3.2, only 77% and 30% of the tubes, respectively, were longer than 100 μm (Figure 5b). The mean length of the wild-type tubes was 412 μm, as compared to 350 and 92 μm in RNAi lines 4.1.1 and 4.3.2, reductions of 15% and 78%, respectively.

Figure 5.

 Pollen tube elongation is inhibited in agp6 agp11 mutants.
(a) In vitro germination of pollen from wild-type (WT) and AGP6 AGP11 RNAi line 4.1.1 and 4.3.2 plants. Arrows indicate very short pollen tubes.
(b) Histogram showing the distribution of pollen tube lengths for wild-type (WT) and AGP6 AGP11 RNAi line 4.1.1 and 4.3.2 plants (> 200 per each genotype). Standard errors for at least five independent experiments are shown.
(c) In vitro germination of pollen from wild-type (WT), P28L and P78L homozygous (HM) and P124L heterozygous (HT) plants. Arrows indicate very short pollen tubes.
(d) Histogram showing the distribution of pollen tube lengths for wild-type (WT), P28L and P78L homozygous (HM) and P78L and P124L heterozygous (HT) plants (> 200 per each genotype). Standard errors for at least eight independent experiments are shown.

In the experiments studying the pollen tubes of agp6 TILLING lines and the wild-type background Big Mama, 69% of the tubes were found to be longer than 100 μm in the latter (Figure 4d). However, mutations in agp6 reduced the percentage of long tubes to 51, 27, 45 and 37% in pollen from P28L and P78L homozygous plants and P78L and P124L heterozygous plants, respectively. In the heterozygous lines, half of the pollen grains are wild-type. Therefore, the reduction in pollen tube length, as compared to length in the wild-type, was the result of reduction in the length of tubes in the 50% population of mutated pollen. Thus, with respect to the number of progeny seeds (Figure 4g), the P124L mutation was more severe than the P28L and P78L mutations. Indeed, the mean pollen tube length was reduced from 252 to 176, 100, 147 and 88 μm for P28L and P78L homozygous plants and P78L and P124L heterozygous plants, respectively.

To test pollen tube growth under in vivo germination conditions, an assay developed by Lalanne et al. (2004) was used, in which excised pistils from male-sterile ms1-1 plants were pollinated and subsequently dyed with Alexander stain (Alexander, 1969), reflecting the density of the cytoplasm in the pollen cell body. Strong staining indicated that the pollen tube had not elongated sufficiently and that the cytoplasm had not relocated from the cell body to the pollen tube. Under our experimental conditions, significantly fewer pollen grains from RNAi line 4.3.2 and the various AGP6 mutant lines were empty 4 h after pollination, compared to wild-type pollen (Ler and Big Mama, respectively; Figure 6), indicating that less pollen had developed elongated tubes. Overall, these pollen tube growth assays indicate that the reduced male fertility of the agp6 mutant and AGP6 AGP11 RNAi plants was at least partly the result of defects in pollen tube growth.

Figure 6.

In vivo germination of wild-type, P28L, P78L and RNAi 4.3.2 pollen grains.
Boxplot showing the percentage of empty pollen grains (i.e. germinated pollen, see text for details) from Big Mama wild-type (WT BM), P28L, P78L, P124L, WT Ler and RNAi 4.3.2 plants, 4 h after pollination (> 200 pollen grains for each genotype). Significant differences in the proportion of empty pollen between the mutant lines and the wild-type were observed (anova, F3,67 = 17.944, < 0.001). Mutant lines P124L, P28L and P78L differed significantly from the wild-type BM (Bonferroni corrected post-hoc comparisons, < 0.001, = 0.01 and = 0.001, respectively). RNAi line 4.3.2 significantly differed from the wild-type Ler (anova, F1,35 = 35.148, < 0.001).

Strong reduction in AGP6/AGP11 expression reveals an additional role for these genes in pollen release

During collection of pollen for the germination assays, it was observed that the stamens of AGP6 AGP11 RNAi line 4.3.2 released almost no pollen, unless mechanically manipulated. To expand upon this observation, stamens from line 4.3.2 and wild-type flowers after anthesis were dissected and examined after Alexander staining (Alexander, 1969) to better visualize the pollen. While wild-type stamens at this stage were almost empty and many pollen grains were seen in the surrounding medium (Figure 7a), pollen grains in AGP6 AGP11 RNAi line 4.3.2 plants were contained within the mature anther locules (Figure 7b). As line 4.3.2 had the most significant reductions in AGP6 and AGP11 expression (97.6 and 95.9%, respectively; Figure 3a), the associated phenotype suggests that AGP6 and AGP11 play roles not only in pollen tube growth but also in release of pollen from the anther. A similar phenotype was also observed in P124L heterozygous plants, albeit to a lesser extent (data not shown). The possibility that AGP6 and AGP11 fulfil additional sporophytic functions is supported by the fact that AGP6 and AGP11 were expressed in stamen sporophytic tissues, in addition to their expression in pollen and the pollen tube (Figure 2).

Figure 7.

 Pollen grains are not released from mature stamens in AGP6 AGP11 RNAi line 4.3.2.
Mature stamens at anthesis were dissected from wild-type (a) and AGP6 AGP11 RNAi line 4.3.2 (b) flowers, and labeled using Alexander stain.


AGP6 and AGP11 are required for pollen function

AGPs have been linked to pollen and pollen function for over a decade. Studies of diverse plant species have indicated the presence of AGPs in the anther, pollen and pollen tube (Abreu and Oliveira, 2004; Coimbra et al., 2007; Gerster et al., 1996; Kawaguchi et al., 1996; Mollet et al., 2002; Peng et al., 2005; Qin et al., 2007; Roy et al., 1998; Wisniewska and Majewska-Sawka, 2006). More recently, expression of AGP-encoding genes in these tissues has been demonstrated (Amagai et al., 2003; Lalanne et al., 2004; Pereira et al., 2006; Wellmer et al., 2004; Zik and Irish, 2003). Several studies have employed antibodies directed against sugar epitopes in AGPs or have used Y(β-Glc)3, which binds specifically to AGPs, to show that perturbation of AGPs leads to inhibition of pollen tube growth (Abreu and Oliveira, 2004; Qin et al., 2007; Roy et al., 1998). In addition, a requirement for GPI-anchored proteins in pollen tube growth has also been demonstrated (Lalanne et al., 2004).

Our study reveals the involvement of specific AGPs in pollen tube growth. Using RNAi lines with reduced levels of AGP6 and AGP11 expression or lines harboring mutations in the coding region of AGP6, we showed that defects in these genes lead to reduced fertility (Figures 3 and 4). The specific expression of AGP6 and AGP11 in anthers and pollen (Figure 2), together with the absence of AGP6 and AGP11 expression in the gynoecium (Figure 2), suggests that these genes are required for male function in fertilization. Indeed, our results show that mutations in AGP6 or reduction of AGP6 and AGP11 expression by RNAi caused inhibition of pollen tube growth (Figures 5 and 6) and hampered pollen release (Figure 7). However, when genetic transmission through the male gametes was assessed by applying pollen from heterozygous agp6 mutants directly on wild-type stigma, the mutant allele was equally transmitted (data not shown). These results suggest that the mutations in the agp6 mis-sense mutants do not completely abolish AGP6 function, such that its disruption is not apparent when pollination conditions are optimized by introducing the pollen directly onto the stigma. Alternatively, the disruption to pollen tube elongation in agp6 mutants might not interrupt gamete transfer under optimal conditions as sufficient pollen tubes succeeded in reaching the ovules. These results also showed that the male gametes are not affected by the mutations in AGP6.

The role of AGPs in cell elongation

Interestingly, it had been previously shown that AGPs are secreted at, and thus mark, the site of emergence of the new pollen tube (Mollet et al., 2002) and that binding of Y(β-Glc)3, a dye that specifically interacts with AGPs, leads to the arrest of apical growth of lily pollen tubes (Jauh and Lord, 1996). Treatment with Y(β-Glc)3 also provided indication of involvement of AGPs in other cases of cell elongation, including elongation in Arabidopsis seedling roots associated with a bulging of epidermal cells (Ding and Zhu, 1997; Willats and Knox, 1996), elongation of cultured tobacco cells (Vissenberg et al., 2001), hypocotyl elongation in cucumber seedlings (Park et al., 2003) and apical extension in moss cells (Lee et al., 2005).

The role of AGPs in cell elongation was further demonstrated using transgenic plants. Tobacco plants overexpressing the cucumber CsAGP1 gene were taller than wild-type plants, due to promoted internode elongation (Park et al., 2003). A mutation in the Arabidopsis SOS5 gene, encoding a protein with two AGP-like domains, led to swelling of the root tip and root growth arrest (Shi et al., 2003). Moreover, a loss-of-function mutant of AtAGP19 had shorter hypocotyls and inflorescence stems than wild-type (Yang et al., 2007). Finally, a knockout of AGP1 in moss resulted in reduced cell length of the protonemal filaments (Lee et al., 2005).

The mechanism of AGP function in cell elongation

Several mechanisms have been proposed to explain the function of AGPs in cell elongation. One mechanism proposes that AGPs serve as plasticizers of the cell wall (Lamport, 2001), acting in one of three possible ways: (i) release from between cellulose fibrils leading to microfibril separation and a decrease in cell-wall rigidity (Rumyantseva, 2005), (ii) binding to ferulic acid residues, thereby disrupting the di-ferulic bridges that contribute to cell-wall rigidity (Rumyantseva, 2005), or (iii) increasing the porosity of the pectin gel (Lamport et al., 2006).

A second explanation for the role of AGPs in cell elongation comes from studies showing that AGPs are involved in cell-wall assembly. Y(β-Glc)3-inhibited pollen tubes show abnormal deposition of cell-wall components (Roy et al., 1998), and treatment with Y(β-Glc)3 inhibited cellulose deposition onto tobacco-cultured protoplasts (Vissenberg et al., 2001). As seen with emerging pollen tubes (Mollet et al., 2002), AGPs are localized to nascent wall in-growths in epidermal transfer cells of Vicia faba (Vaughn et al., 2007). Exposure of the developing transfer cells to Y(β-Glc)3 caused a significant reduction in the density of wall in-growth deposition (Vaughn et al., 2007).

However, treating cells with Y(β-Glc)3 does not provide unequivocal proof for an in vivo role of AGPs in cell-wall assembly. For instance, the abnormal cell-wall deposition and structure observed after Y(β-Glc)3 treatment might be an artifact resulting from a gain-of-function action of the inhibitor, for example AGP/Y(β-Glc)3 aggregates forming a rigid barrier, thereby blocking further growth (Seifert and Roberts, 2007) and disrupting the deposition of cell-wall material. The mutants isolated in this study can be used to address the question of whether arrest of pollen tube growth, as a result of agp loss of function, is accompanied by changes in cell-wall composition and structure. Using pollen tubes that were germinated in vitro, stained with aniline blue and examined using fluorescent light microscopy, we could not detect differences between wild-type and the arrested pollen tubes of agp mutants (data not shown). Higher resolution, as achieved by electron microscopy (Lennon and Lord, 2000; Roy et al., 1998), may be required to detect changes in the cell wall.

It has been shown that disrupting GPI anchoring leads to abnormal cell-wall synthesis. Plants that are mutant in COBRA, which encodes a GPI-anchored protein, show mis-oriented root cell expansion, in correlation with reduced amounts of crystalline cellulose in their cell walls (Schindelman et al., 2001). Mutations in the PEANUT1 gene, encoding a homolog of mammalian PIG-M, which is required for GPI anchor biosynthesis, resulted in radially swollen embryos whose cell walls showed decreased crystalline cellulose content and irregular and ectopic deposition of pectins, xyloglucan and callose (Gillmor et al., 2005). Similar to the seth mutants, peanut1 mutants showed reduced transmission through pollen, supporting the premise that GPI anchoring is required for pollen function. AGP6 and AGP11 encode classical arabinogalactan proteins to which a GPI anchor is predicted to be added (Figure 1; Schultz et al., 2000). Among the agp6 TILLING mutants we obtained, the most severe phenotype was observed in line P124L, where the mutation is located just two amino acids upstream of the sequence motif that directs addition of the GPI anchor (Figure 1b). We thus hypothesize that a significant amino acid change in the proximity of the GPI anchor addition sequence, which disrupts splicing of the protein C-terminus and/or linkage of the GPI anchor, affects cell-wall biogenesis in agp6 and agp11 mutants.

agp6 and agp11 mutants as tools to study the mode of action of AGPs

The high complexity of AGPs makes their analysis using biochemical tools very complicated, limiting studies on the structure–function relationships of these proteins. Use of the stable agp6 and agp11 mutants obtained in this study for mutant rescue assays enables assessment of the significance of various amino acids or protein domains for protein function, such as those that direct the addition of a GPI anchor or the proline/hydroxyproline-rich domains to which the polysaccharide chains of AGPs are attached (Riou-Khamlichi et al., 1999; Rumyantseva, 2005; Schultz et al., 2000).

The biological importance of pollen function and the uniqueness of pollen tube tip growth have made these processes the focus of substantial research (Becker and Feijo, 2007). One of the challenges facing current efforts in the field relates to defining the functions of pollen-specific proteins (Becker and Feijo, 2007; McCormick, 2004). Our study assigns functions to AGP6 and AGP11– two stamen- and pollen-specific genes – in pollen tube growth and pollen release. In addition, our study provides tools for study of the mechanism of action of the proteins encoded by these genes, thereby contributing to the understanding of the genetic basis of pollen tube tip growth and AGP activity.

Experimental procedures

Plant materials and growth conditions

RNAi lines were produced as described below in a Landsberg erecta (Ler) background. The TILLING mutants are in a Col er105 (Big Mama) background (Torii et al., 1996). TILLING line seeds were obtained from the Nottingham Arabidopsis Stock Centre (University of Nottingham, UK). Seeds of lines CSHL_ET5767 and CSHL_ET3978 were obtained from the Cold Spring Harbor Laboratory Gene Trap collection ( Seeds were plated on a mixture of soil (sphagnum peat and tuff) and perlite (2:1), and grown under a 16 h light/8 h dark cycle at 22°C. Seeds from transformations were plated onto MS plates comprising 0.5× MS medium, 0.5% sucrose, 0.7% agar and the appropriate selection agent [i.e. 50 μg ml−1 kanamycin or 15 μg ml−1 gluphosinate ammonium (DL-phosphinotricin,].

GUS staining and RFP fluorescence microscopy

For the AGP6:AGP6-GUS construct, a fragment of 1010 bp upstream from the first ATG of gene At5g14380 to the end of the coding region (450 bp) was PCR-amplified using specific primers 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTTCACCGGAGAAAATCTACCACACGT-3′ and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTCGAAAGAAAAGAAGAAGAAGCCGAC-3′, containing attB1 and attB2 sites, respectively, and cloned into the Gateway pENTR plasmid (Invitrogen, according to the manufacturer’s guidelines. The fragment was then recombined into the Gateway destination binary vector, pGUSGate1, in-frame with the GUS gene. The pGUSGate1 vector confers spectinomycin resistance to bacteria and contains the BAR gene for gluphosinate ammonium resistance in plants.

The AGP6pro:GUS and AGP11pro:GUS constructs were generated by amplifying fragments of 810 or 600 bp upstream from the first ATG genes At5g14380 (AGP6) and At3g01700 (AGP11), respectively, using gene-specific primers AGP6PF (5′-AACCTGCTACGTAATTTATTTGAC-3′) and AGP6PR (5′-TTTCTGTTTTAATAGTTTTTTTATTG-3′), and AGP11PF (5′-ACACGAATTCAAACCACACACTCACACTCAC-3′) and AGP11PR (5′- GAATGGATCCACTAGTATTTTCTACTTTTTAAATATAGTTATA-3′). The fragments were inserted upstream of the GUS gene in a plasmid derived from the plant binary vector pMD1, a derivative of the pBIN19 plasmid (Bevan, 1984), containing the NPTII kanamycin resistance gene.

The AGP6pro:RFP and AGP11pro:RFP constructs were generated as follows: the promoter sequences were PCR-amplified (as described above) using specific primers containing EcoRI/SpeI sites and cloned into plasmid BJ36-10XOP-RFP (a gift from Yuval Eshed, Weizmann Institute of Science, Rehovot, Israel), replacing the 10XOP sequence. The AGP6/AGP11:RFP-ocs fragment was cleaved using NotI and ligated into binary plasmid pMLBART (Gleave, 1992).

GUS staining and tissue preparation for sectioning were performed according to the protocol of the EMBO Practical Course on Genetic and Molecular Analysis of Arabidopsis, downloaded from the TAIR website ( For whole-mount observation, tissues were cleared after staining in 70% ethanol and images were collected using a Nikon DXM1200F digital camera ( connected to a Nikon stereo SMZ1000 dissecting microscope. For sectioning, samples were dehydrated in a series of ethanol step dilutions (up to 50% ethanol) and fixed in FAA (5% formaldehyde, 50% ethanol and 10% acetic acid) for 30 min. The samples were further dehydrated in a second series of ethanol step dilutions up to 100% ethanol, passed through a series of increasing concentrations of tert-butanol (TBA) and finally incubated overnight in 100% TBA at 60°C. Samples were embedded overnight in 50% paraplast (ParaplastPlus, and then transferred to 100% paraplast. The paraplast solution was changed twice for two additional days. Sections (8 μm thick) were mounted on slides, and the paraplast was removed by washing the slides in Histoclear. Images were collected using a Nikon DXM1200F digital camera connected to a stereo dissecting microscope or a Nikon Eclipse 80i light microscope. RFP-expressing tissue was visualized using the light microscope with a fluorescent light source through an RFP filter set (λex 545/30 nm; λem = 620/60 nm barrier filter).

AGP6 AGP11 RNAi lines

The AGP6 AGP11 RNAi construct was generated as follows: 231 and 236 bp fragments from AGP6 and AGP11, respectively, were PCR-amplified using primers containing a SpeI recognition site (RNAi6F, 5′-CTCGAGTCTAGACATCTTCGGCCTCTTCTCCA-3′; RNAi6R, 5′-ACTAGTTAGCAGTTGTCACGGCACCAC-3′; RNAi11F, 5′-ACTAGTCCGATGGTCCCTCCTCCG-3′; RNAi11R, 5′-GAATTCGGATCCAGTTAGTTAAAGAGATACGGCC-3′, respectively). The resulting AGP6 and AGP11 fragments were ligated and cloned in the sense and antisense orientations on either side of an intron found within the pHANNIBAL plasmid (Wesley et al., 2001) using XhoI/EcoRI and XbaI/BamHI sites, respectively. The construct created in plasmid pHANNIBAL was subcloned as a NotI fragment into binary plasmid pMLBART (Gleave, 1992).


The various transformation constructs were introduced into Agrobacterium strain GV3101 by electroporation. Arabidopsis plants (Ler) were transformed using the floral dip method (Clough and Bent, 1998).

DNA and RNA extraction and RT-PCR

Plant DNA was extracted using the Cetyl Trimethyl Ammonium Bromide (CTAB) extraction method (McDonald and Martinez, 1990). Total RNA was extracted using an EZ total RNA isolation kit (Biological Industries, cDNA was synthesized using a Reverse-iT 1st strand synthesis kit (ABgene,

Expression of AGP6 and AGP11 in AGP6 AGP11 RNAi lines was analyzed by real-time quantitative RT-PCR using a 7300 real-time PCR system (Applied Biosystems,, with 18S rRNA serving as an internal control and wild-type cDNA serving as calibrator. Real-time PCR was performed using Absolute Blue SYBR Green ROX mix (ABgene), according to the manufacturer’s instructions. The following primer sets were used: AGP6F-RT (5′-GACCAAGCGATGGACCCA-3′) and AGP6R-RT (5′-CGACTGTGCCGACAACAGAG-3′), AGP11F-RT (5′-TTTGTCGTAGTTGCCCTTTTGG-3′) and AGP11R-RT (5′-TTGGTGAGGCGGTGG GT-3′), and 18S-F (5′-AAGCAAGCCTACGCTCTGGA-3′) and 18S-R (5′-AGGCCAACACAATAGGATCGA-3′). Relative quantification of the target AGP6 and AGP11 expression levels was performed using the comparative ΔCt method in the 7300 system software. To achieve maximum amplification, PCR conditions for each primer combination were optimized for concentration.


The following primers were employed to screen for the TILLING mutants: forward primer 5′-TTCACACCCCCATGTCTTATTCTTACATCC-3′ and reverse primer 5′-GACCACACAAATCATAAGCATTTCCAAAA-3′, yielding an 827 bp fragment starting 162 bp upstream from the first ATG. Genotyping was performed by PCR amplification, followed by digestion with restriction enzymes whose recognition site was modified by the mutation (Table 1). The forward primer 5′-CTCGAGTCTAGACATCTTCGGCCTCTTCTCCA-3′ was used for lines P78L, P80L and P124L, while primer 5-CCGCCGACGCTCCCTCAGCTTGTC-3′ (designed using the dCAPS method; Neff et al., 1998) was used as the forward primer for line P28L. Used together with the reverse primer, 5′-ACTAGTTAGCAGTTGTCACGGCACCAC-3′, 231 bp (P78L, P80L and P124L) or 345 bp (P28L) fragments were amplified. The following restriction enzymes were used: Alw26I for P28L; AvaII for P78L; XhoI for P80L; HphI for P124L.

Phenotypic and cytological analyses of pollen

In vitro germination of mature pollen grains was performed on solid medium containing 0.01% H3BO3, 0.07% CaCl2, 20% sucrose and 0.5% agarose, as described previously (Johnson-Brousseau and McCormick, 2004). Pollen tubes were scored by examining images collected from slides mounted in a light microscope connected to a digital camera. NIH image software (Abramoff et al., 2004) was used to measure pollen tube lengths from all captured images. Mature pollen grains were stained for viability as described by Alexander (1969).

In vivo germination efficiency was determined by pollinating ms1-1 (Wilson et al., 2001) stigmas, and subjecting them to Alexander staining 4 h after pollination, as described previously (Lalanne et al., 2004). At least 200 pollen grains where used to pollinate 20 stigmas of each line from the mutated and wild-type lines. The percentage empty pollen (i.e. pollen grains that germinated on the stigma and developed elongated tubes) was determined for each stigma separately, and the mean for all 20 stigmas was calculated for each line. Differences in the proportion of empty pollen tubes between the wild-type and the mutant lines were tested using a one-way analysis of variance (anova) followed by Bonferroni corrected post-hoc comparisons. As accepted for proportions, data were arcsine-square root-transformed (Sokal and Rohlf, 1995). All statistical analyses were performed using SYSTAT version 11 (SYSTAT Software,


We thank CSIRO Plant Industry for the pHANNIBAL plasmid, the Seattle TILLING Project (STP) for providing the agp6 mutant lines, the Arabidopsis Research Center (ABRC) for providing the TILLING line seeds, the Arabidopsis gene trap project at Cold Spring Harbor Laboratory for seeds of enhancer trap lines CSHL_ET5767 and CSHL_ET3978, Yuval Eshed (Weizmann Institute of Science, Rehovot, Israel) for the BJ36-10XOP-RFP plasmid, Ofer Ovadia (Ben Gurion University, Beer Sheva, Israel) for advice on statistical analyses) and Michele Zaccai, Khalil Kashkush, Jerry Eichler (Ben Gurion University, Beer Sheva, Israel) and the two anonymous reviewers for critical reading of the manuscript.