Plant growth and fertility requires functional interactions between specific PABP and eIF4G gene family members

The initiation of protein synthesis requires the involvement of the eukaryotic translation initiation factor (eIF) 4G to promote assembly of the factors needed to recruit a 40S ribosomal subunit to an mRNA. Although many eukaryotes express two eIF4G isoforms that are highly similar, those in plants, referred to as eIF4G and eIFiso4G, are highly divergent in size, sequence, and domain organization. Species of the Brassicaceae and the Cleomaceae also express a divergent eIFiso4G isoform, referred to as eIFiso4G2, not found elsewhere in the plant kingdom. Despite their divergence, eIF4G and eIFiso4G interact with eIF4A, eIF4B, and eIF4E isoforms needed for binding an mRNA. eIF4G and eIFiso4G also interact with the poly(A)-binding protein (PABP) which promotes ribosome recruitment to an mRNA. Increasing the complexity of such an interaction, however, Arabidopsis also expresses three PABP isoforms (PAB2, PAB4, and PAB8) in vegetative and reproductive tissues. In this study, the functional interactions among the eIF4G and the widely-expressed PABP isoforms were examined. Loss of PAB2 or PAB8 in combination with loss of eIF4G or eIFiso4G had little to no effect on growth or fertility whereas pab2 pab8 eif4g or pab2 pab8 eifiso4g1/2 mutants exhibited smaller stature and reduced fertility. Although the pab4 eifiso4g1 mutant grows normally and is fertile, pab4 eif4g or pab4 eifiso4g2 mutants could not be isolated. Even pab4/PAB4 eif4g/eIF4G heterozygous plants exhibited growth defects and low fertility. Mutant co-inheritance analysis in reciprocal crosses with wild-type plants revealed that most ovaries and pollen from pab4/PAB4 eif4g/eIF4G plants were PAB4 eif4g. Similarly, co-inheritance studies with pab4/PAB4 eifiso4g2/eIFiso4G2 plants suggested most ovaries were PAB4 eifiso4g2. These results suggest that a functional interaction between PAB4 and eIF4G and between PAB4 and eIFiso4G2 is required for growth and normal fertility.


Introduction
Protein synthesis in eukaryotes requires multiple translation initiation factors that promote 40 S ribosomal subunit recruitment to an mRNA, initiation codon recognition, and 80 S ribosome assembly [1,2,3]. One such eukaryotic initiation factor (eIF) is eIF4F composed of eIF4E, which binds the 5 0 -cap structure; eIF4A which is an RNA helicase involved in a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 unwinding secondary structure in a 5 0 -leader that might impede 40 S subunit scanning; and eIF4G, which interacts with eIF4E, eIF4A, eIF4B (which stimulates the RNA helicase activity of eIF4A), eIF3 (required for 40 S binding to an mRNA), and the poly(A)-binding protein (PABP) [2,4,5,6]. The eIF4E-eIF4G interaction and the eIF4G-PABP interaction bridge the termini of an mRNA to stimulate 40 S subunit recruitment, an interaction observed in yeast, plants, and animals [7,8]. An interaction between eIF4B and PABP, observed in plants and animals, serves to increase the RNA-binding affinity of PABP [9,10,11].
Two eIF4G isoforms are present in plants as in many eukaryotes [12]. However, unlike other eukaryotes, the plant eIF4G isoforms, referred to as eIF4G and eIFiso4G, differ substantially in size and sequence and each interacts with distinct eIF4E isoforms [13]. Nevertheless, both eIF4G isoforms from wheat interact with eIF4B, eIF4A, and PABP. Wheat eIF4G contains two interaction domains for these three partners whereas eIFiso4G has a single domain for eIF4B and PABP [14,15]. Such divergence has suggested functional specialization in mRNA selection, a conclusion supported by the observation that alfalfa mosaic virus RNA 4 and Arabidopsis HSP21 use wheat eIF4G or eIFiso4G equally well whereas barley α-amylase, oat globulin mRNAs, and the tobacco etch virus 5 0 -leader use eIF4G preferentially [16,17].
Translation from mRNAs containing the 5 0 -leader sequence from tobacco mosaic virus (TMV) (known as O) preferentially uses eIF4G over eIFisoG1 in wheat germ lysate [18]. More recent work revealed that in addition to eIF4G, O-mediated translation in Arabidopsis was also facilitated by the eIFiso4G2 isoform [19]. Arabidopsis expresses two eIFiso4G isoforms (eIFi-so4G1 and eIFiso4G2) as do other species in the Brassicaceae and the Cleomaceae whereas species outside these species express only eIFiso4G1 isoforms [19]. eIFiso4G1 and eIFiso4G2 exhibit substantial sequence divergence, supporting their functional difference in facilitating O-mediated translation. The presence of eIFiso4G1 as the only eIFiso4G isoform in wheat explains the preferential use of eIF4G by O in wheat germ translation lysate [18].
PABP is highly conserved throughout eukaryotes and is involved in translation, mRNA biogenesis, mRNA transport, and mRNA stability [20]. Although its physical interaction with eIF4G serves to promote translation, PABP also increases mRNA stability by inhibiting DCP1/ 2-mediated decapping in yeast [21,22]. Binding of eIF4E to the 5'-cap may be destabilized following loss of PABP during mRNA deadenylation, making the cap more susceptible to DCP1/ 2. Inhibition of decapping by PABP is also observed in mammals [23]. PABP promotes poly (A) nuclease (PAN)-mediated deadenylation but inhibits Caf1/Not1-mediated deadenylation in yeast [21,22].
While yeast expresses just a single and essential PABP [24], PABP in Arabidopsis is expressed by a highly divergent, multigene family composed of eight members that can be grouped into three classes based on sequence similarity and gene structure [25]. Three members of the family, i.e., PAB2, PAB4, and PAB8, comprise class II PABP genes which are widely expressed, particularly in vegetative tissues [26,27,28,29]. Expression of class I genes (PAB3 and PAB5) is restricted to reproductive tissues whereas class III genes (PAB6 and PAB7) exhibit weak and restricted expression. PAB1, originally considered an "orphan gene" is actually a class I gene expressed weakly in roots, pollen, and during late embryo development [25,30,31].
Analysis of the evolution of the PABP gene family in plants revealed that PABP expanded from a single gene in marine algae to two genes in fresh water algae which underwent further expansion to four highly similar progenitor genes during the evolution of non-vascular plants. These progenitor genes are most structurally similar to class I reproductive members of the Arabidopsis PABP gene family but all three classes of PABP genes present in higher plants evolved from these progenitor genes [25]. By the appearance of Selaginella moellendorffii, an early vascular plant, the first class III gene (i.e., PAB7) had appeared and a class II family member appears first in gymnosperms such as pine [25]. The three class II family members of Arabidopsis arose from a single class II gene in Amborella trichopoda, the most recent common ancestor of all extant flowering plants. However, PAB8 appears largely only in the Brassicaceae and its sequence similarity suggests that it is a recent duplication of PAB2 [25]. Similarly, PAB3 is likely a duplication of PAB5 that is present largely in the Brassicaceae and PAB1 also appears limited to the Brassicaceae, demonstrating a recent expansion of this gene family in these species [25]. Finally, PAB6 is present in rosids but not in asterid species.
The presence of a large gene family containing divergent members, suggests possible functional specialization that is supported by studies of the ability of members of the Arabidopsis gene family to compliment various PABP functions in pab1 mutant yeast. Although expression of Arabidopsis PAB2, PAB3, and PAB5 in pab1 yeast rescued viability, PAB2 and PAB5 promoted translation initiation and poly(A) shortening in yeast; PAB2, but not PAB5, protected the mRNA 5'-cap against decapping; and PAB2 interacted with eIF4G [27,28,32]. The function of other PABP family members, such as PAB1, PAB4, PAB6, and PAB7, has not been examined.
Given the importance of the interaction between eIF4G and PABP and its conserved nature throughout eukaryotes, how the isoforms of eIF4G and PABP in plants interact functionally to control specific aspects of plant growth and development is an important question that has remained unanswered. In this study, the functional interactions among eIF4G isoforms and those members of the PABP gene family that are widely-expressed were examined. Loss of PAB2 or PAB8 expression in combination with loss of eIF4G or eIFiso4G expression had little to no effect on growth or fertility whereas pab2 pab8 eif4g triple mutants or pab2 pab8 eifi-so4g1/2 quadruple mutants exhibited smaller stature than did eif4g or eifiso4g1/2, respectively. pab4 eifiso4g1 mutant plants grew normally and were fertile but pab4 eif4g or pab4 eifiso4g2 mutants could not be isolated. Moreover, pab4/PAB4 eif4g/eIF4G heterozygous plants exhibited low fertility. Analysis of mutant co-inheritance in reciprocal crosses with wild-type plants indicated that most ovaries and pollen from pab4/PAB4 eif4g/eIF4G plants were PAB4 eif4g. Similar results were obtained for co-inheritance studies with pab4/PAB4 eifiso4g2/eIFiso4G2 plants. Co-inheritance of pab4 eif4g or pab4 eifiso4g2 through the female germ line was observed in crosses with wild-type plants, albeit at low frequencies, demonstrating that the functional interaction between PAB4 and eIF4G or between PAB4 and eIFiso4G2 was not absolutely essential for ovary development. In contrast, co-inheritance of pab4 eif4g through the male germ line was not observed although inheritance of pab4 eifiso4g2 was at least under conditions of eIFiso4G1 expression. These results suggest that the functional interactions between PAB4 and eIF4G and between PAB4 and eIFiso4G2 are essential for aspects of plant development including embryogenesis and/or vegetative growth.

Plant growth
Arabidopsis seeds were imbibed and cold-treated for four days prior to germination in soil. Plants were grown in a growth room supplemented with Sylvania Gro-Lite fluorescent bulbs (Sylvania, Danvers, MA) at a photon flux density (PFD) of 100 μmol photons m −2 s −1 .

Chlorophyll measurements
Chlorophyll a and b were measured spectrophotometrically. Leaf samples were ground in liquid nitrogen and extracted with 90% (v/v) acetone. The absorbance at 664 and 647 nm was determined and used to calculated chlorophyll a and b content by the equations: Chl a = 11.93A 664 -1.93A 647 and Chl b = 20.36A 647 -5.50A 664 , respectively. Measurements were repeated 2-3 times and representative results presented.

The functional interaction between PAB4 and eIF4G is necessary for fertility and vegetative growth
Class II members of the PABP gene family (PAB2, PAB4, and PAB4) are widely expressed and predominate in vegetative tissues [26,27,28,29]. T-DNA insertion mutants of these gene members result in loss of expression but loss of any one class II member has little to no effect on growth or fertility [33]. Although the pab4 pab8 double mutant also grows similar to wildtype (WT) plants, the pab2 pab8 double mutant is slightly more compact than wild-type as a result of a shorter petiole (i.e., leaf stem) and reduced internode length best seen in the inflorescence (i.e., flower head) in the distance between siliques (i.e., seed pods) and the leaves of the pab2 pab4 double mutant are substantially smaller and fewer in number than WT [33]. However, loss of any one eIF4G isoform (eIF4G, eIFiso4G1, or eIFiso4G2) fails to affect growth or fertility [34]. To examine whether the loss of expression of any one class II PABP gene member in combination with eIF4G would affect growth not observed for the individual mutants, single and double pab mutant plants were crossed with eif4g plants to generate each combinatorial mutant.
The combination of pab2 or pab8 with eif4g had little effect on plant stature (Fig 1) or on leaf size of young adult plants (Fig 2A). However, the inflorescence of each was slightly reduced in size relative to the single mutants or to WT plants ( Fig 2C). Combining the pab2 pab8 double mutant with eif4g resulted in a more compact stature compared to the eif4g mutant but was similar to that observed for the pab2 pab8 double mutant prior to flowering (Figs 1 and 2A) and a more compact inflorescence due to a shorter internode distance similar to that observed for the pab2 pab8 double mutant ( Fig 2D). The pab2 eif4g double mutant and pab2 pab8 eif4g triple mutant developed fewer ovaries per silique and shorter siliques than WT or any single mutant of these loci (Table 1) [33].
Despite at least 15 attempts, isolation of a pab4 eif4g double mutant failed suggesting that loss of PAB4 and eIF4G results in synthetic lethality. Similarly, pab2 pab4 eif4g or pab4 pab8 eif4g triple mutants could not be isolated, supporting the conclusion that the combination of PAB4 and eIF4G is essential. Even attempts to isolate pab4/PAB4 eif4g/eif4g or pab4/pab4 eif4g/eIF4G mutants were unsuccessful. Isolation of a pab4/PAB4 eif4g/eIF4G double heterozygous mutant, however, was possible. pab4/PAB4 eif4g/eIF4G plants grew relatively normally although they exhibited reduced chlorophyll content ( Table 2). pab4/PAB4 eif4g/eIF4G flowered normally (S1 Fig), however the plants exhibited extremely poor fertility with fewer ovaries per silique than WT (Fig 3 and Table 3). The failure of most ovaries to develop into embryos could be seen during embryogenesis (S1 Fig). The plants continued to flower for longer than WT, resulting in a substantial increase in silique number although most siliques were devoid of seed (Fig 3). Those siliques that did contain seed were much reduced in size, likely as a result of the presence of a single or few seed ( higher than in pab4/PAB4 eif4g/eIF4G plants and, as a consequence, the siliques in eif4g/eIF4G heterozygote plants were longer to accommodate the increased number of seed present. Interestingly, despite pab4/PAB4 eif4g/eIF4G plants having far more siliques per plant relative to eif4g/eIF4G heterozygote plants, the extremely low seed set in the former resulted in fewer seed. Thus, as the reduction in fertility in pab4/PAB4 eif4g/eIF4G was greater than that observed in eif4g/eIF4G heterozygote plants, the addition of PAB4 in a heterozygous state contributed to the reduction in fertility associated with the eif4g/eIF4G heterozygous mutation. Ovary and seed development in pab2/PAB2 eif4g/eIF4G and pab8/PAB8 eif4g/eIF4G plants was similar or slightly more impaired than that observed in eif4g/eIF4G heterozygote plants (Table 3 and S1 Fig). As pab2 eif4g and pab8 eif4g double homozygotes were viable whereas pab4 eif4g could not be isolated, these results suggest that the combination of PAB4 and eIF4G is essential for plant viability and/or fertility. Although neither PAB2 or PAB8 was essential for plant viability and fertility, loss of expression of either partially ameliorated the growth defects and reduction in fertility observed in pab4/PAB4 eif4g/eIF4G plants through an increase in ovaries per silique and siliques with seed to a level similar to that observed for eif4g/eIF4G heterozygote plants (Fig 4  and Table 3), suggesting possible genetic interactions between PAB4 and either PAB2 or PAB8. An analysis of variance (one way ANOVA) with post hoc comparisons using the Tukey HSD test supported those differences that were determined significant by t-test (Table 3, S1 Table, S2  Table, and S3 Table).

The functional interaction between PAB4 and eIF4G is required for seed development
The inability to isolate a pab4 eif4g double mutant and the poor fertility of pab4/PAB4 eif4g/ eIF4G heterozygote plants suggested the possibility of altered Mendelian inheritance of the mutant loci in progeny following self-pollination. Genotypic analysis of such progeny revealed 54 out of 80 plants (67.5% versus expected of 6.25%) were PAB4/PAB4 eif4g/eif4g with 19 out of 80 (23.8% versus expected of 12.5%) being PAB4/PAB4 eif4g/eIF4G (Table 4). Four (5% versus expected of 6.25%) were pab4/pab4 eIF4G/eIF4G and 3 (3.75% versus expected of 25.0%) were pab4/PAB4 eif4g/eIF4G heterozygotes. These results deviate substantially from the expected inherence from a double heterozygous parent. All seed were viable suggesting that the combinatorial loss of PAB4 and eIF4G expression affected germ line development or embryogenesis, consistent with the poor seed set exhibited by the heterozygous parent and the absence of double mutants among the progeny. To investigate whether loss of PAB4 and eIF4G expression affected germ line development, pab4/PAB4 eif4g/eIF4G plants were crossed reciprocally with WT plants to examine co-inheritance distributions in ovaries and pollen that resulted in successful fertilization. Genotypic analysis of progeny from crosses using pab4/ PAB4 eif4g/eIF4G plants as the recipient of WT pollen indicated that almost all (93.0%) ovaries resulting in successful fertilization events were PAB4 eif4g with only 2.3% being pab4 eIF4G and 4.7% being pab4 eif4g (Table 5). Interestingly, no PAB4 eIF4G ovaries were detected. Genotypic analysis of progeny from crosses using pab4/PAB4 eif4g/eIF4G plants as the pollen donor onto WT plants indicated that almost all (93.6%) pollen resulting in successful Specific functional interactions between PABP, eIF4G, and eIFiso4G isoforms fertilization events were PAB4 eif4g with 2.1% being pab4 eIF4G and 4.3% being PAB4 eIF4G ( Table 5). The predominance of the PAB4 eif4g combination in ovaries and pollen of pab4/ PAB4 eif4g/eIF4G plants is consistent with the predominance of PAB4/PAB4 eif4g/eif4g in progeny of self-pollinated pab4/PAB4 eif4g/eIF4G heterozygote plants ( Table 4). The detection of two pab4 eif4g ovaries in crosses with WT as the pollen donor (Table 5) demonstrates that expression of PAB4 and eIF4G is not essential for female gametophyte development although pab4 eif4g ovaries involved in successful fertilization events were under represented. No evidence supporting the presence of pab4 eif4g pollen was observed. This and the failure to isolate pab4 eif4g plants suggests synthetic lethality between pab4 and eif4g during male or female gametophyte development/function and/or subsequent stages during embryo development in ways that the combination of pab2 and eif4g or pab8 and eif4g is not. Examination of pollen viability [35] as described (S1 Text) in flowers of pab4/PAB4 eif4g/eIF4G plants indicated that those pollen that developed to maturity were largely viable (S4 Fig).
The combinatorial loss of expression of either PAB2 or PAB8 with pab4/PAB4 eif4g/eIF4G substantially increased the number of pab4 eIF4G ovaries and pollen in reciprocal crosses with WT plants (Table 5). Thus, in the absence of PAB2 or PAB8 expression, pab4 eIF4G ovaries and pollen participate in successful fertilization events whereas they are largely absent from such events in the presence of PAB2 or PAB8 expression, suggesting possible genetic interactions between PAB4 and either PAB2 or PAB8 during male and female gametophyte development or function. pab2 pab4 eif4g and pab8 pab4 eif4g ovaries resulting in successful fertilization events were observed but not pab2 pab4 eif4g or pab8 pab4 eif4g pollen. Together with the failure to detect pab4 eif4g pollen in crosses using pab4/PAB4 eif4g/eIF4G plants (Table 5), this represents a total of 161 progeny from these three crosses in which co-inheritance of pab4 and eif4g through the male germ line failed. The lack of pab4 eif4g pollen in any of these three crosses would prevent the generation of pab4/pab4 eif4g/eif4g embryos, regardless of whether such embryos would be viable or not.
Reciprocal crosses between pab4/PAB4 plants and WT demonstrated the inheritance rate of PAB4 and pab4 was approximately equal through the male and female germ lines (Table 7). Reciprocal crosses between eif4g/eIF4G plants and WT suggested the inheritance rate of eIF4G and eif4g was approximately equal through either germ line ( Table 7).

The functional interaction between PAB4 and eIFiso4G2 is required for ovary and pollen function
The synthetic lethality between pab4 and eif4g raised the question of whether a similar interaction exists between PAB4 and the other eIF4G isoform, eIFiso4G. eIFiso4G is encoded by two genes in A. thaliana: eIFiso4G1 (At5g57870) and eIFiso4G2 (At2g24050) which are substantially divergent (55.4% identity and 64.9% similarity). No visible phenotype is observed in eifi-so4g1 or eifiso4g2 mutants other than the slightly reduced stature of eifiso4g1 plants [19]. The eIFiso4G double null mutant, i.e. eifiso4g1/2, is considerably smaller with reduced chlorophyll levels and slower growth rate than WT plants [34,36]. Although this suggests that the eIFiso4G isoforms are functionally similar in contributing to vegetative growth, they also exhibit functional differences in the mRNAs they support translationally [19].
To examine whether the combinatorial loss of any of the widely expressed PABP gene members with eIFiso4G would affect growth, single and double pab mutant plants were crossed with eifiso4g1 and eifiso4g2 plants to generate each combination of mutants. pab2 pab8 Specific functional interactions between PABP, eIF4G, and eIFiso4G isoforms eifiso4g1 or pab2 pab8 eifiso4g2 triple mutants were similar in stature to pab2 pab8 double mutant and WT plants although these particular mutants are more compact than WT due to shorter petioles (Figs 5 and 6). The inflorescence of pab2 pab8 eifiso4g1 or pab2 pab8 eifiso4g2 triple mutants was also more compact than WT as it was in the pab2 pab8 double mutant ( Fig  7). The eifiso4g1/2 double mutant is substantially reduced in stature and leaf size relative to WT plants or either single mutant (Figs 5 and 6). The growth rate of pab2 eifiso4g1/2 and pab8 Specific functional interactions between PABP, eIF4G, and eIFiso4G isoforms eifiso4g1/2 triple mutants was slower than eifiso4g1/2 plants and the leaf size of the pab8 eifi-so4g1/2 triple mutant was smaller than fully flowering eifiso4g1/2 plants (Figs 5, 6 and 7). The growth rate of the pab2 pab8 eifiso4g1/2 quadruple mutant was even slower with smaller stature and leaf size relative to either triple mutant (Figs 5, 6 and 7) and it exhibited delayed flowering even with respect to the eifiso4g1/2 double mutant (Fig 7). These data suggest that PAB2 and PAB8 contribute combinatorially to plant stature which is observed whether eIFiso4G (or eIF4G) is expressed or not. These data also demonstrate that pab2 and pab8, either together or alone, do not exhibit synthetic lethality with eifiso4g.
As with PAB4 and eIF4G, over 10 attempts to isolate a pab4 eifiso4g1/2 triple mutant failed suggesting that loss of PAB4 and eIFiso4G expression results in synthetic lethality. Similarly, pab2 pab4 eifiso4g1/2 or pab4 pab8 eifiso4g1/2 quadruple mutants could not be isolated, supporting the conclusion that the combination of PAB4 and eIFiso4G is essential. As eIFiso4G is expressed as two divergent isoforms, the isolation of pab4 eifiso4g1 or pab4 eifiso4g2 double mutants was attempted. The pab4 eifiso4g1 double mutant was viable, flowered normally, and exhibited fertility similar to WT (S5 Fig). The pab4 eifiso4g2 double mutant, however, could not be isolated. Isolation of a pab4/PAB4 eif4iso4g2/eIFiso4G2 double heterozygous mutant was possible. Its growth characteristics were similar to WT plants (Fig 8 and S5 Fig) and did not exhibit the reduced chlorophyll content (D. Gallie, personal observation) or poor fertility observed with pab4/PAB4 eif4g/eIF4G heterozygous plants. Interestingly, a cross between pab4/pab4 eifiso4g1/eifiso4g1 and eifiso4g1/eifiso4g1 eifiso4g2/eifiso4g2 plants to generate pab4/ PAB4 eifiso4g1/eifiso4g1 eif4iso4g2/eIFiso4G2 which were then selfed, resulted in a subset of  S5 Fig). In contrast, the eif4iso4g1 eif4iso4g2 double mutant has substantially fewer ovaries per silique, reduced seed per silique, and fewer and shorter siliques. eif4iso4g1/eIFiso4G1 eif4iso4g2/eIFiso4G2 double heterozygous mutant plants also had fewer ovaries per silique, reduced seed per silique, and slightly shorter siliques although not fewer siliques (Table 8). An analysis of variance (one way ANOVA) with post hoc comparisons using the Tukey HSD test supported those differences that were determined significant by t-test (Table 8, S4 Table, S5 Table, S6 Table, S7 Table, S8 Table, and S9 Table). These results indicate a synthetic lethal relationship between PAB4 and eIFiso4G2 similar to that observed between PAB4 and eIF4G.
To determine whether certain mutant combinations affect ovary or pollen development, pab4/PAB4 eif4iso4g2/eIFiso4G2 plants were reciprocally crossed with WT plants to examine which gene combinations predominated in ovaries and pollen participating in successful fertilization events. Genotypic analysis of progeny from crosses using pab4/PAB4 eif4iso4g2/eIFi-so4G2 plants as the recipient of WT pollen indicated that 3.5% ovaries were PAB4 eIFiso4G2, 31.6% ovaries were pab4 eIFiso4G2, 61.4% ovaries were PAB4 eifiso4g2, and 3.5% ovaries were pab4 eifiso4g2 (Table 10). Genotypic analysis of progeny from crosses using pab4/PAB4 Specific functional interactions between PABP, eIF4G, and eIFiso4G isoforms  Specific functional interactions between PABP, eIF4G, and eIFiso4G isoforms eifiso4g2/eIFiso4G2 plants as the pollen donor onto WT plants indicated that 51.6% of successful fertilization events resulted from pab4 eIFiso4G2 pollen, 41.9% resulted from PAB4 eifiso4g2 pollen, 6.5% resulted from pab4 eifiso4g2 pollen, and none from PAB4 eIFiso4G2 pollen (Table 10). The presence of pab4 eifiso4g2 ovaries and pollen, although few in number, demonstrates that the combination of PAB4 and eIFiso4G2 is not essential for male or female gametophyte development (although they are underrepresented in each germ line) and thus would not prevent the generation of pab4/pab4 eifiso4g2/eifiso4g2 embryos. The failure to isolate pab4 eifiso4g2 plants therefore suggests synthetic lethality between pab4 and eifiso4g2 during embryo development.

Discussion
In this study, pab4 and eif4g exhibited synthetic lethality as did pab4 and eifiso4g2. In contrast, no synthetic lethality was observed between pab4 and eifiso4g1 or between other pab and eif4g or eifiso4g mutants. The synthetic lethality exhibited between either eif4g or eifiso4g2 with pab4 suggests that eIF4G and eIFiso4G2 are functionally similar but that their combined contribution is necessary. Interestingly, these results are consistent with the previous finding that eIF4G and eIFiso4G2 exhibit similar translational preferences in that translation from mRNAs containing the TMV O 5'-leader sequence preferentially uses eIF4G and eIFiso4G2 over eIFi-soG1 in Arabidopsis [19].
PABP and eIF4G isoforms are expressed from multigene families. The PAB gene family in Arabidopsis (and in species of the Brassicaceae) is unusual in the number and diversity of its isoforms. From a single gene in the marine algae Chlamydomonas reinhardtii to two genes in fresh water algae (Klebsormidium flaccidum) to four genes in Physcomitrella patens, the PAB gene family expanded with the colonization of land [25]. By the appearance of gymnosperms and the basal angiosperm, Amborella trichopoda, the PAB gene family had differentiated into the three classes that characterize PABP proteins in higher plants with further expansion of each class, especially within the Brassicaceae [25].
Of the three class II PAB genes in Arabidopsis, PAB2 and PAB4 contribute most to vegetative growth and normal vegetative-to-floral transition [33]. PAB2 contributes significantly to inflorescence architecture, silique morphology, and fertility [33]. PAB8 also contributed to silique and inflorescence development and to fertility as revealed in double class II pab mutants. The contribution that class II PABP proteins make to flowering and fertility demonstrated that, despite the specific expression from the class I PAB3 and PAB5 genes and the class III PAB7 gene in reproductive tissues [33], class II PAB genes are also important for reproduction. Of the three class II PABP isoforms in Arabidopsis, PAB4 is less similar to PABP from other species such as wheat (56.7%/63.6% identity/similarity) than is PAB2 (62.8%/68.2% identity/similarity) or PAB8 (66.4%/70.5% identity/similarity) suggesting potential for greater Specific functional interactions between PABP, eIF4G, and eIFiso4G isoforms functional divergence. Interestingly, in knockouts of individual class II PAB members in Arabidopsis, expression from luciferase constructs with or without the TMV O 5'-leader sequence was reduced to a greater degree in pab4 mutants than in pab2 or pab8 mutants, suggesting that PAB4 contributes more to reporter gene expression than does either PAB2 or PAB8 [33]. These observations support the conclusion that class II PAB genes exhibit some degree of functional specialization.
Although loss of expression of PAB4 or any eIF4G isoform individually had little to no observable effect on plant growth or fertility [33,36], repeated attempts to isolate a pab4 eif4g double mutant failed. As the pab2 eif4g, pab8 eif4g, and pab2 pab8 eif4g mutants were viable, this suggesting synthetic lethality between pab4 and eif4g. The requirement for co-expression of PAB4 and eIF4G was sensitive to gene copy number in that neither pab4/PAB4 eif4g/eif4g nor pab4/pab4 eif4g/eIF4G could be isolated. Isolation of a pab4/PAB4 eif4g/eIF4G double heterozygous mutant was possible although the plants were characterized by reduced chlorophyll content and very poor fertility. Reciprocal crosses between pab4/PAB4 eif4g/eIF4G and WT plants revealed that most ovaries and pollen from the pab4/PAB4 eif4g/eIF4G parent involved in successful fertilization events were genotypically PAB4 eif4g. The presence of pab4 eif4g ovaries, although few in number, demonstrated that ovary development can occur in the absence of PAB4 and eIF4G co-expression. However, out of a combined total of 161 progeny examined, no successful fertilization events resulted from pab4 eif4g pollen, suggesting that their coexpression may be needed for male gametophyte development or function. The lack of pab4 eif4g pollen prevents the generation of pab4 eif4g double mutant embryos. Consequently, it is not possible to conclude that PAB4 and eIF4G co-expression is essential for embryo development or for vegetative growth. However, the poor fertility and reduced chlorophyll content exhibited by pab4/PAB4 eif4g/eIF4G plants does suggest that co-expression of PAB4 and eIF4G may be necessary to support other aspects of reproduction and normal vegetative growth.
As observed for eIF4G, repeated attempts to isolate a pab4 eifiso4g2 combinatorial mutant failed, suggesting that loss of PAB4 and eIFiso4G2 expression results in synthetic lethality. The observed synthetic lethality was specific for eIFiso4G2 as the pab4 eifiso4g1 double mutant was viable and its growth was similar to WT plants. Isolation of a pab4/PAB4 eifiso4g2/eIFiso4G2 double heterozygous mutant was possible and these plants appeared similar to WT plants in their growth. Crosses between pab4/PAB4 eifiso4g2/eIFiso4G2 and WT plants revealed that most ovaries from the pab4/PAB4 eifiso4g2/eIFiso4G2 parent involved in successful fertilization events were predominantly PAB4 eifiso4g2 genotypically but that ovaries that were pab4 eIFi-so4G2 were well represented. Similar results were obtained when pab4/PAB4 eifiso4g2/eIFi-so4G2 plants were used as the pollen donor in a cross with WT plants although PAB4 eifiso4g2 and pab4 eIFiso4G2 progeny were more equally represented. The presence of pab4 eifiso4g2 ovaries and pollen, although few in number, suggests that co-expression of PAB4 and eIFi-so4G2 is not essential for male or female gametophyte development, at least in the presence of eIFiso4G1 expression. Therefore, the failure to isolate pab4 eifiso4g2 plants suggests that coexpression of PAB4 and eIFiso4G2 is needed to support either embryo development and/or vegetative growth.
Although eIFiso4G1 and eIFiso4G2 might be expected to exhibit substantial functional overlap as they are isoforms, eIFiso4G2 differs from eIFiso4G1 in that it evolved recently within the Brassicaceae and the Cleomaceae [19] but exhibits substantial sequence divergence from eIFiso4G1, particularly within those regions that function as binding sites for partner proteins and RNA [19]. eIFiso4G1 and eIFiso4G2 also exhibit distinct differences in translational preferences in that eIFiso4G2 is more functionally similar to eIF4G during translation than it is to eIFiso4G1 [19], supporting the synthetic lethality exhibited between pab4 and eif4g and between pab4 and eifiso4g2.
The combinatorial loss of expression from eIF4G and either PAB2 or PAB8 resulted in little to no decrease in fertility in the homozygous mutant state and the modestly reduced fertility of pab2 pab8 eif4g triple mutant plants was not greater than that observed for pab2 pab8 plants [33]. pab2/PAB2 eif4g/eIF4G and pab8/PAB8 eif4g/eIF4G double heterozygous plants exhibited only a small reduction in ovaries and seed per silique in relation to that observed for eif4g/ eIF4G plants, suggesting that the genetic interaction between eIF4G and PAB2 and/or PAB8 contributes modestly at best to fertility. Interestingly, loss of expression from either PAB2 or PAB8 in pab4/PAB4 eif4g/eIF4G plants partially reversed the defects in fertility observed for the latter, i.e., few successful and shorter siliques, fewer ovaries per silique, and reduced seed set. Loss of expression from either PAB2 or PAB8 also reversed the growth defects of pab4/ PAB4 eif4g/eIF4G plants. This suggests a genetic interaction in which expression from either PAB2 or PAB8 exacerbates the effects of PAB4 mutations in pab4/PAB4 eif4g/eIF4G plants which is relieved in the absence of their expression.
PAB8 is the most recently evolved member of class II PAB genes and is largely limited to species of the Brassicaceae. As PAB8 protein is most similar to PAB2, it likely arose from a gene duplication of PAB2 [25]. The combinatorial loss of expression from PAB2 and PAB8 results in a modest reduction in growth, a reduction in inflorescence internode length, and modest reductions in ovary and seed per silique [33]. The inflorescence of pab2 pab8 eif4g triple mutant plants were similar to those of pab2 pab8 double mutants and its vegetative growth was similar, data suggesting little genetic interaction between these class II proteins and eIF4G. In contrast, pab2 pab8 eifiso4g1/2 quadruple mutant plants were smaller than eifiso4g1/2 double mutant plants, the latter of which are characterized by a substantial reduction in stature, reduced chlorophyll content, delayed vegetative-to-floral transition, and poor fertility [33,34,36]. In addition to a further reduction in stature, the delay in the vegetative-to-floral transition in pab2 pab8 eifiso4g1/2 mutants was more pronounced and fertility was extremely poor compared to eifiso4g1/2 plants, resulting in only a few seed per plant. These results suggest that the effects of loss of expression from PAB2 and PAB8 and the loss of expression from eIFiso4G1 and eIFiso4G2 may be additive.
In addition to the synthetic lethality between pab4 and eif4g, pab4/PAB4 eif4g/eIF4G plants exhibited non-Mendelian inheritance of the mutant loci. Progeny from pab4/PAB4 eif4g/ eIF4G plants were largely PAB4/PAB4 eif4g/eif4g (67.5% versus 6.25% expected) or PAB4/ PAB4 eif4g/eIF4G (23.8% versus 12.5% expected) along with pab4/pab4 eIF4G/eIF4G (5.0% versus 6.25% expected) and pab4/PAB4 eif4g/eIF4G (3.75% versus 25.0% expected). Thus, progeny containing the two PAB4 alleles as wild-type in the presence of eif4g in either a homozygous or heterozygous state was preferred. No pab4/pab4 eif4g/eif4g plants were obtained consistent with the lack of pab4 eif4g pollen participating in successful fertilization events. Interestingly, no PAB4/PAB4 eIF4G/eIF4G plants were obtained. Reciprocal crosses with WT plants revealed that 93.0% of the eggs or 93.6% of the pollen from pab4/PAB4 eif4g/eIF4G plants involved in successful fertilization and seed development were genotypically PAB4 eif4g, consistent with the over-representation (67.5% versus 6.25% expected) of PAB4/PAB4 eif4g/eif4g progeny from selfed pab4/PAB4 eif4g/eIF4G plants. The low-frequency occurrence of pab4 eif4g eggs involved in successful fertilization and seed development demonstrated that co-expression of PAB4 and eIF4G was not required for female gametophyte development and/or function but that the absence of pab4 eif4g pollen involved in successful fertilization and seed development suggests co-expression of PAB4 and eIF4G may be necessary for male gametophyte development and/or function. Combining null mutations affecting eIF4E1 and eIFiso4E in Arabidopsis was lethal in that transmission through the male gametophyte failed although transmission through the female gametophyte was successful [37], supporting the conclusion that pollen development or function is affected to a greater extent by loss of critical components of the translational machinery than is egg development/function.
The predominance of PAB4 eif4g eggs and pollen and the lack of PAB4 eIF4G eggs or the very few PAB4 eIF4G pollen in a pab4/PAB4 eif4g/eIF4G parent participating in successful fertilization events suggests possible defects including alterations to selection of the megaspore (which gives rise to the female gametophyte), pollen tube guidance, or rate of pollen tube growth. The observation that the eggs and pollen from a pab4/PAB4 or an eif4g/eIF4G parent that successfully participated in fertilization events were not biased to the same extent as that observed in a pab4/PAB4 eif4g/eIF4G parent suggests that it is the interaction of the mutant or wild-type alleles of PAB4 and eIF4G that is responsible for altered genotypic distribution of eggs and pollen in a pab4/PAB4 eif4g/eIF4G parent.
The absence of PAB2 or PAB8 expression in pab4/PAB4 eif4g/eIF4G plants altered the genotype of the eggs and pollen participating in successful fertilization events from being predominantly PAB4 eif4g in a pab4/PAB4 eif4g/eIF4G parent to a mix of PAB4 eif4g and pab4 eIF4G eggs and pollen. These results suggest that PAB2 or PAB8 expression contributes to the predominance of PAB4 eif4g eggs and pollen in a pab4/PAB4 eif4g/eIF4G parent that are involved in successful fertilization events, indicating a degree of complexity in the genetic interaction between class II PAB genes and eIF4G during reproduction.
Synthetic lethality was also observed between pab4 and eifiso4g2. pab4/PAB4 eifiso4g2/eIFi-so4G2 plants exhibited non-Mendelian inheritance of the mutant loci although to a lesser extent than that observed in pab4/PAB4 eif4g/eIF4G plants. Reciprocal crosses with WT plants revealed that the eggs and pollen participating in successful fertilization events were predominantly PAB4 eif4iso4g2 or pab4 eIFiso4G2 although pab4 eifiso4g2 eggs and pollen also participated, indicating that expression from PAB4 and eIFiso4G2 is not essential for male or female gametophyte development and function. It is possible that expression from eIFiso4G1 may account for this difference. As in pab4/PAB4 eif4g/eIF4G plants, eggs and pollen from pab4/ PAB4 eifiso4g2/eIFiso4G2 parents that were genotypically PAB4 eIFiso4G2 were under represented. In contrast, however, loss of PAB2 or PAB8 expression in pab4/PAB4 eifiso4g2/eifiso4g2 parents did not alter the genotypic distribution of eggs and pollen participating in successful fertilization events to the same extent as observed in pab4/PAB4 eif4g/eIF4G plants.

Conclusion
The synthetic lethality between pab4 and eif4g and between pab4 and eifiso4g2 suggests that the functional interaction between PAB4 and eIF4G may be similar to that between PAB4 and eIFiso4G2 and these interactions may target similar pathways. Whether these genetic interactions affect global protein synthesis or are of particular importance for expression from specific genes is unknown. The reduced chlorophyll content but relatively normal vegetative growth of pab4/PAB4 eif4g/eIF4G plants, however, suggests that expression from specific genes involved in chloroplast development or function may be dependent on a functional interaction between PAB4 and eIF4G. Similarly, the observed participation of pab4 eif4g ovaries but not pab4 eif4g pollen in crosses resulting in non-Mendelian inheritance suggests that the functional interaction between PAB4 and eIF4G may be important for the expression of genes involved specifically in aspects of pollen development or function. In contrast, the observed reduction in vegetative growth of pab2/8 eifiso4g1/2 quadruple mutants may indicate effects on global protein synthesis. Future work investigating the basis for the synthetic lethality between pab4 and eif4g or between pab4 and eifiso4g2 as well as how these mutant combinations affect male or female gametophyte development in double heterozygous parents will be needed to reveal those pathways that require co-expression of PAB4 and eIF4G (or eIFiso4G2).