Plant materials and breeding:

To analyze the effects of diverse maize zfl alleles relative to mRNA null alleles, we generated 10 maize F₂ populations (H95a, H95b, IL, OH43a, OH43b, M01, M02-1, M02-2, W00, and W03) that segregated for mRNA null alleles of zfl1 and zfl2 and wild-type alleles from five different maize genetic backgrounds (inbred lines H95, IL101, OH43 and W22, and the “Mu-Killer” background) (Martienssen and Baron 1994). Each F₂ population was generated from a cross between a maize plant that carried mRNA null alleles at both zfl1 and zfl2 and individuals from the respective “wild-type” maize line. The mRNA null alleles used were the previously described Mutator transposon insertion alleles zfl1–mum1 and zfl2–mum1, or zfl1–mum2 and zfl2–mum3 (Bomblies et al. 2003). All of the populations segregated for the alleles zfl1–mum1 and zfl2–mum1 except the W03 population, which segregated zfl1–mum2 and zfl2–-mum3.

F₂ populations used to score phenotypes were generated as follows. The breeding schemes used to generate the M01 and W00 populations were previously described (Bomblies et al. 2003). The M02-1 and M02-2 F₂ populations (157 and 158 plants, respectively, “Mu-killer” background) were derived as described for M01 (Bomblies et al. 2003). The H95a and H95b F₂ populations (234 and 72 plants, respectively, H95 background) were derived from a cross between a plant from the M02-1 population that carried zfl1–-mum1 and zfl2–mum1 and inbred line H95. OH43a and OH43b (193 and 105 plants, respectively, OH43 background) were derived from a cross between a plant from the M02-1 population that carried zfl1–mum1 and zfl2–mum1 and inbred line OH43. The W03 F₂ population (99 plants, W22 background) was derived from a cross between two lines, one that carried zfl1–mum2 and the other zfl2–mum3, each backcrossed four generations to inbred line W22. The IL population (262 plants, IL101 background) was derived from a cross between a plant that carried zfl1–mum1 and zfl2–mum1 (from an F₂ population generated as described for the M01 population) and the IL101 inbred maize line.

Quantitative phenotypic variation was scored in field-grown populations cultivated and analyzed during summer (May–September) in an outdoor field plot at the University of Wisconsin West Madison Agricultural Research Station (Madison, WI). All plants were phenotyped and subsequently genotyped to test for associations (see Genotyping below). The W00 population was grown in summer 2000; M01 in summer 2001; M02-1, M02-2, and IL in summer 2002; and H95a/b, OH43a/b, and W03 in summer 2003.

To test whether zfl2 gene activity is necessary for a QTL for ear traits differentiating maize and teosinte that maps to the same chromosomal region, we performed a QTL complementation test. To obtain maize/teosinte hybrid F₂ families for the QTL complementation tests, we crossed a maize plant that was heterozygous for zfl2–mum1 (backcrossed six times into the W22 background) to teosinte (Z. mays ssp. parviglumis collection Iltis and Cochrane 81). We genotyped F₁ plants at zfl2 (see Genotyping below) and selected the following four F₁ individuals: (1) plant 1, which carried wild type maize zfl2 and a teosinte zfl2 allele we named “T1”; (2) plant 2, which carried zfl2–mum1 and T1; (3) plant 3, which carried wild-type maize zfl2 and a different teosinte zfl2 allele we named “T2”; and (4) plant 4, which carried zfl2–mum1 and T2. The T1 and T2 teosinte alleles were defined by restriction fragment length polymorphism (RFLP; see Genotyping below). F₁ plants were grown in isolation and allowed to self-fertilize in the University of Wisconsin greenhouse in winter 2002–2003 to generate the following four parallel F₂ populations: (1) T1W, 193 plants derived from F₁ plant 1; (2) T1m, 178 plants derived from F₁ plant 2; (3) T2W, 168 plants derived from F₁ plant 3; and (4) T2m, 150 plants derived from F₁ plant 4. F₂ populations were grown and phenotyped at Hawaiian Research in Kaunakakai, Molokai Island, HI, in a field plot in winter 2003–2004.

Independent zfl1 mutations:

Two of the 10 maize F₂ populations (M02-1 and IL) carried independent mutations at the zfl1 locus, and thus were used as “control” populations to ask whether associated effects require zfl1 gene activity, as opposed to being caused by segregation of linked genes. The maize inbred line IL101 carries a 220-bp deletion in exon 2, which leads to a frame shift and a premature stop codon in the mRNA (Ed Buckler, personal communication; sequence available at: http://www.panzea.org/db/searches/polyexp_search). The M02-1 population segregated phenotypically zfl double-mutant plants at a ratio of 1/4 (P > χ² = 0.98), while the related M02-2 population segregated phenotypically mutant plants as expected for the 1/16 segregation of recessive mutations in two redundant genes (P > χ² = 0.65). We sequenced zfl1 and zfl2 from the M02-1 population and found that the zfl1 locus derived from the Mu-killer parent had a Ds transposon insertion 30 bases upstream of intron 1. Sequencing RT-PCR products showed that the resulting message is misspliced from the Ds element to exon 2. This generates a message with a frameshift, which truncates exon 1, encodes eight amino acids not normally found in zfl, followed by a premature stop codon in exon 2. Thus, at the zfl1 locus both the IL and M02-1 populations segregated only zfl1–mum1 and independently derived putative null zfl1 alleles (zfl1–IL101 and zfl1–Ds). The M02-1 line, which was generated from an F₁ parent sibling to the M02-2 F₁ parent, thus provides a genetically similar “control” population for M02-2, which inherited a zfl1 allele with a normal sequence from the Mu-Killer parent.

Genotyping:

Plants in F₂ populations were genotyped using a combination of PCR and RFLP approaches. The IL, M01, M02-1, and M02-2 F₂ populations were genotyped using RFLP markers with previously described hybridization conditions (Doebley and Stec 1991). M01 genotyping was previously described (Bomblies et al. 2003). For the M02-1 and M02-2 populations, we performed southern hybridization on genomic DNA restricted with XbaI, while for the IL population, we used HindIII/XbaI double digests. The probe used for the blots was generated from the zfl1 cDNA (Bomblies et al. 2003) using random primers supplemented with zfl-specific primers (5′ CCAACGACGCCTTCTCGG 3′ and 5′ ACATCGACGACGCAGCTAGA 3′). The remaining populations (H95a/b, OH43a/b, and W03) were genotyped using a PCR approach. First, we assayed for the presence of the zfl1–mum1, zfl1–mum2, and zfl2–mum1 Mutator alleles as described previously (Bomblies et al. 2003). We genotyped for the presence of the zfl2–mum3 allele (W03 population) using the primers: 5′ ATTTGTGGCCGCCCAGCTTAGCGA 3′ and 5′ AGAGAAGCCAACGCCA(A/T)CGCCTC(C/T)ATTTCGTC 3′.

We then tested for the presence of wild-type alleles using a zfl2 specific primer set (5′ AGCCTCGCCGTGTCTTCT 3′ and 5′ CCCGTGGACTTGCGAGAC 3′) and a zfl1 specific primer set (5′ GCATTGGAAAACAGTTAC 3′ and 5′ GTCTGCCGTTTGTATAT 3′) spanning the transposon insertion sites. Double-mutant control DNA was included to verify that wild-type bands did not amplify from DNA homozygous for Mutator alleles. This information was combined to assign genotypes for both zfl1 and zfl2 to each individual.

Maize/teosinte hybrids used in the QTL complementation test were genotyped as follows. To initially identify F₁ maize/teosinte hybrid plants that carried the same teosinte alleles in combination with wild-type and mutant maize zfl2 alleles, genomic DNA from F₁ plants was restricted with HindIII and hybridized by Southern blotting with a 392-bp zfl2-specific 5′ region probe generated by PCR using the primers 5′ AGCCTCGCCGTGTCTTCT 3′ and 5′ AAGGCGTCGTTGGGATCCAT 3′. The probe matches the first 372-bp 5′ of zfl2 and did not cross-hybridize with zfl1 under our conditions. DNA from maize line A682 was used as the PCR template. Hybrid F₂ populations were genotyped by PCR for segregating simple sequence repeat (SSR) markers located on chromosome 2. SSR map locations and primer sequences were obtained from the maize mapping project (http://www.maizegdb.org/ssr.php) and tested for segregation in our populations. For the T1W and T1m populations, we used markers umc1165, mmc0231, and bnlg1175, while for T2W and T2m we used umc1227 instead of umc1165 and umc1026 in place of bnlg1175. We genotyped plants at zfl2 using a size polymorphism that results from a 307-bp MITE transposon insertion 329 bp 5′ of the translation start present in the W22 and zfl2–mum1 alleles, but not in the teosinte T1 and T2 zfl2 alleles.

Phenotyping:

To ask whether the zfl genes might associate with quantitative variation for traits important in maize domestication, we scored our maize F₂ populations for 11 traits for flowering time and plant and inflorescence architecture. We measured flowering time by counting vegetative leaf number (LN; with seed leaf as leaf one), days to pollen (DTP; number of days after planting that anthers exserted on the tassel), days to silk (DTS; number of days after planting that silks are visible on ears), and husk leaf number (HLN; number of husk leaves enclosing the uppermost ear). To assay plant architecture, we scored the number of lateral branches, or ears (LBN; number of ear shoots or lateral branches in leaf axils), the node with the lowermost developed ear (LDE; “developed” was defined as any ear whose husk extended beyond the subtending leaf sheath), the node with the uppermost developed ear (UDE), and the number of blank vegetative nodes (producing no ears) between the uppermost ear and the terminal tassel (BLN). We analyzed inflorescence architecture by scoring the number of long branches on the terminal male inflorescence or tassel (TBN), ear phyllotaxy (KRN; the number of rows of kernels around the circumference of the ear averaged over three counts taken 1 cm from the base, at the middle, and ∼2 cm from the end of the ear), and tassel phyllotaxy or rank (TRNK; average of three measurements of the number of rows, or ranks, of spikelet pairs around the rachis circumference taken along the central spike of the tassel).

For the maize/teosinte hybrid families used in the QTL complementation test, we measured ear phyllotaxy and blank node number (BLN). Ear phyllotaxy was scored as ear rank (RNK), the number of grain-bearing cupules around the circumference, rather than as KRN, because maize/teosinte hybrid populations independently segregate for paired vs. single kernels, making rank a more consistent measure of phyllotaxy in these groups. We report two RNK measurements for these families: average ear rank for the terminal ear of the uppermost lateral branch (uRNK) and the terminal ear of the branch below the uppermost branch (sRNK).

Data analysis:

We tested for associations between plant phenotypes and zfl1 and zfl2 genotypes using two-way analysis of variance (ANOVA) in JMP-IN (SAS Institute) and report least-square mean (LSM) trait values for each genotype category for each zfl gene individually. Two-way ANOVA was performed in JMP using the “fit model” function with the model y = zfl1 + zfl2 + zfl × zfl2. Percentage of phenotypic variance explained is reported as an r² value calculated from the sums of squares (SS) for each gene and the whole model in JMP. We tested for effects associated with zfl1 or zfl2 individually using ANOVA by analyzing one subset for each gene in each of the three largest populations (H95a, OH43a, M01) that segregated for one of the zfl genes, but was homozygous for mutant alleles at the paralogous locus. Dominance/additivity (d/a) ratios were calculated using the LSM trait values where a = wild-type trait value/2 − mutant trait value/2 and d = heterozygote trait value − (mutant trait value/2 + wild-type trait value/2). For maize/teosinte hybrid populations, we performed ANOVA for each marker individually for the traits scored and used the resulting trait means to calculate d/a ratios with maize as “wild type” and teosinte as “mutant” in the d/a equations.

As a first step in QTL mapping analysis, we generated genetic linkage maps for the chromosome 2 SSR markers and zfl2 in Mapmaker/EXP v 3.0 (Lander et al. 1987) using the Kosambi mapping function. Windows QTL Cartographer v 2.0 (Wang et al. 2001–2004) was used for QTL mapping in the maize/teosinte hybrid populations by composite interval mapping (CIM) using model 6 (which selects certain control markers to correct for background effects) and the forward regression method with five control markers and a window size of 10 cM. LOD score significance thresholds were estimated for each trait in each population with 1000 permutations of the data.

Maize F₂ populations—developmental timing:

We estimated the timing of reproductive development by counting leaf number (LN), days to pollen shed (DTP; male maturity), days to silk (DTS; female maturity) and the number of husk leaves (HLN) on the ear shoot. LN consistently showed stronger associations with zfl1 than with zfl2 (Table 1). For zfl1, we observed a statistically significant LN decrease associated with increasing number of wild-type zfl1alleles in all seven populations assayed, (Tables 1 and 2) with plants homozygous wild type at zfl1 averaging 4–13% fewer leaves than the homozygous zfl1 mutant class. zfl2 showed a significant association with LN in five of the seven populations (Table 2). HLN decreased significantly with increasing active zfl1 and zfl2 copies in all four populations scored (Tables 1 and 2), with the homozygous wild-type zfl1 class averaging 17–37% fewer husk leaves than the homozygous zfl1 mutant class. zfl1 explained more of the HLN variance than zfl2 in all four populations.

We observed a modest but statistically significant decrease in DTP and DTS with increasing active zfl1 copies in all six populations examined, while for zfl2 DTP and DTS trends were each statistically significant in five of the six populations (Tables 1 and 2). Plants homozygous for mutant zfl1 alleles averaged 2–4% higher DTP and 3–5% higher DTS than plants carrying two wild-type alleles, while plants that had two mutant zfl2 alleles averaged 2–4% higher DTP and 1–5% higher DTS than plants with two wild-type zfl2 alleles. zfl1 explained more variance for DTP and DTS than zfl2 in most of the populations (Table 2).

Plant architecture:

We observed significant associations between the zfl genes and several plant architecture traits: the number of ears or lateral branches (LBN), the lowermost and uppermost nodes bearing developed ears (LDE and UDE, respectively), and the number of blank nodes (BLN). A decrease in LBN was significantly associated with increasing wild-type zfl2 copy number in six of eight populations tested and with increasing zfl1 copy number in four populations (Tables 1 and 3). zfl2 explained more of the trait variance in all of these populations than zfl1. Plants homozygous for wild-type zfl2 alleles averaged 23–44% lower LBN than plants carrying two mutant zfl2 alleles in the populations with significant trends, while plants homozygous for two wild-type zfl1 alleles averaged 7–29% lower LBN. LDE was significantly lower in plants with higher active zfl1 or zfl2 copy number in four of the five populations scored. zfl1 explained more of the LDE variance in all four of these with plants that carried two mutant alleles averaging 11–17% lower LDE than plants that carried two wild-type zfl1 alleles. Plants that carried two mutant zfl2 alleles averaged 5–12% lower LDE than plants that carried two wild-type zfl2 alleles (Table 3). UDE was significantly lower in plants with higher active copy numbers of either zfl1 or zfl2 in four of the five populations scored, while the reverse trend was observed for zfl2 in one population (W03; Table 3). The UDE variance explained by each gene in these groups was similar and UDE averaged 9–13% higher in plants with two wild-type copies of either zfl gene (Table 3). BLN was significantly higher (24–30%) in plants with higher active zfl2 copy numbers in three of six populations examined and with higher zfl1 in one population (9% increase with two wild-type copies of zfl1). zfl2 explained more of the BLN variation in all three lines (Table 3). In contrast, in the W03 population, increasing wild-type copy number from zero to two of either zfl1 or zfl2 was significantly associated with a decrease in BLN (17 and 12%, respectively).

Inflorescence architecture:

We examined variation for inflorescence structure by measuring the number of long tassel branches (TBN), ear phyllotaxy as the number of kernel rows around the ear (KRN), and phyllotaxy or rank of the central tassel spike (TRNK).

TBN showed statistically significant quantitative associations with zfl2 in all eight populations scored for this trait (Tables 1 and 4). In these populations, increasing active zfl2 copy number from zero to two was associated with an increase in TBN of 12–73%, while increasing active zfl1 copy number was associated with a significant increase in TBN (of 5%) in one population (M01) and a statistically significant but strongly overdominant trend in the W00 population (Table 4). Correspondingly, zfl2 explained more of the TBN variance than zfl1 in all populations.

KRN was significantly associated with zfl2 in six of seven populations assayed, while zfl1 genotype showed a significant association with KRN in only two of these (Tables 1 and 4). In the six populations that showed significant KRN associations with zfl2, plants homozygous for wild-type zfl2 alleles averaged 6–13% higher KRN than plants that had two mutant zfl2 alleles. Plants that were homozygous wild type for zfl1 in the H95 and M01 populations averaged 4 and 18% higher KRN, respectively, than plants homozygous mutant for zfl1. In all populations except M01, zfl2 explained more KRN variance than zfl1. TRNK was scored only in one population (M01), in which a statistically significant association with both zfl1 and zfl2 was observed. TRNK was more strongly associated with zfl1, corresponding to 12% higher TRNK in plants that carried two wild-type zfl1 alleles than in plants with two mutant alleles (Table 4).

zfl1 and zfl2 in isolation:

To more specifically analyze the effects associated with each of the two zfl genes, we analyzed two subsets each (one for each gene) from the three largest populations (H95a, OH43a, M01). These subsets contained only individuals homozygous for mutant alleles at one of the two genes, while segregating for the paralog. Analyzing the effects of each of the two zfl genes in absence of gene product from the paralog allowed us to compare the effects of the two genes and to analyze whether either gene alone is sufficient to rescue specific aspects of the double-mutant phenotype.

For LN, DTP, and DTS we observed significant associations with both zfl1 and zfl2 genotype in all six subpopulations. HLN, which was scored for only two of the subpopulations showed a similar trend (Table 5). In keeping with trends observed in whole populations, zfl1 genotype explained more phenotypic variance for these reproductive timing traits than zfl2 did in most of the subpopulations. LBN strongly associated with both genes in the OH43a and M01 subpopulations, and with only zfl2 in the H95a subpopulation (Table 5). zfl2 genotype consistently explained more LBN variance than zfl1. LDE showed a stronger association with zfl1 than with zfl2 in both sets of subpopulations scored, while UDE associated with both zfl1 and zfl2 (Table 5). BLN showed statistically significant associations with zfl1 and zfl2 only in the H95 subpopulations (Table 5).

zfl2 associated with KRN in all three zfl2 subsets (Table 5). Plants homozygous wild type at zfl2 had 24–30% higher KRN than double-mutant plants. KRN showed significant association with zfl1 only in the M01 zfl1 subset, where plants carrying two wild-type zfl1 copies averaged 35% higher KRN than double-mutant plants. TRNK was significantly associated with zfl2 in the M01 subset, where homozygous wild-type plants had 11% higher TRNK than double mutants. zfl1 was associated with a similar trend that was not statistically significant, although this is likely due to higher variance because fewer plants were in this subset. Plants homozygous wild type for zfl1 had 13% higher TRNK than double-mutant plants, a slightly larger effect than that observed for zfl2.

Lines carrying independent zfl1 mutations:

Results from two populations, IL and M02-1, shed light on the important question whether zfl activity itself, as opposed to segregation of a linked gene or genes, was responsible for the observed trait associations. The IL and M02-1 populations carried independent putative null alleles of zfl1 (see materials and methods). As expected, in both populations, phenotypically double-mutant plants segregated at 1/4 instead of 1/16 as observed in populations with functional zfl1 alleles.

In both the IL and M02-1 populations, significant associations were observed with zfl2 genotype for numerous traits (Table 6). In contrast, only one trait, DTS, showed a statistically significant trend for zfl1. The DTS trend observed, however, was only marginally significant, and was opposite to that observed in wild-type populations (lower DTS was associated with increasing zfl1–mum alleles; Table 6). Overall, these results are consistent with the hypothesis that the trends associated with zfl1 genotype in the segregating populations require zfl1 activity rather than being caused by segregation of a linked gene or genes.

Trait associations:

We tested for associations between ear rank of both the terminal ear of the uppermost lateral branch (uRNK) and the terminal ear of the second branch from the top (the branch below the uppermost branch; sRNK), with markers in the previously described QTL regions on chromosome 2. We observed statistically significant associations between chromosome 2 markers, including zfl2 and flanking markers, in the T1W and T2W populations (Table 7). In contrast, no significant ear rank associations were observed in population T1m. In T2m, two markers on chromosome 2 showed statistically significant ear rank associations, but these were associated with overdominant trends and the zfl2–mum1 homozygous maize class did not significantly differ from the teosinte homozygous class (Table 7). This effect may suggest a more complex interaction. For example, there is perhaps a trans-effect on ear rank due to a factor or factors linked to zfl2–mum1 on the maize chromosome that require zfl2 activity, but can act through the teosinte zfl2 allele when it is present.

BLN also showed statistically significant associations with chromosome 2 marker genotypes in populations T1W and T1m. In T1W, BLN was higher in plants homozygous for maize alleles at chromosome 2 markers, but in T1m, the teosinte homozygous class averaged higher BLN (Table 7). This suggests that the teosinte zfl2 allele may also promote higher BLN, but to a lesser degree than the maize allele. T2m showed a similar BLN association trend as T1m, but no significant BLN association was observed in T2W, suggesting that this trait may be sensitive to differences in genetic background or linked genes on chromosome 2 itself.

QTL analysis:

We used QTL mapping to further analyze the genetic effects associated with chromosome 2 in the maize–teosinte hybrid populations. We detected statistically significant sRNK QTL in populations T1W and T2W (Table 8). In each case maize alleles were associated with higher ear rank (the more maize-like phenotype). We did not detect QTL for ear rank in either of the populations carrying mutant maize zfl2 alleles, consistent with the hypothesis that zfl2 activity is important for the chromosome 2 ear rank effect.

The T2W QTL peak for sRNK is centered over the zfl2 locus as expected, but in T1W, the QTL is centered over bnlg1175 instead (Table 8). However, when marker-genotype associations were examined for each of the markers individually, we observed that in T1W zfl2 was associated with sRNK more strongly (r² = 0.09) than the flanking markers, umc1165 (r²= 0.03) and mmc0231 (r² = 0.05), while bnlg1175 was more strongly associated with sRNK (r² = 0.10) than its neighboring locus mmc0231. This observation suggests that the location of the single QTL peak may be a statistical artifact, and suggests that two or more chromosome 2 factors may affect ear rank in this population. However, since no significant associations or QTL were observed in the parallel population, T1m, the chromosome 2 ear rank effect in this population, even if genetically complex, appears to be influenced by maize zfl2 activity.

A significant QTL for BLN was detected in population T1W (Table 8), but not in the parallel population lacking zfl2 activity (T1m). This result is consistent with association effects that suggest that zfl2 activity is important for the BLN effect on chromosome 2 in population T1W. However, no BLN QTL were detected in populations T2W or T2m, suggesting genetic variability for this trait.

Associations suggest partially conserved pleiotropic zfl functions:

In this study we analyzed associations between quantitative phenotypic variation in maize F₂ populations and varying wild-type copy number of the duplicate maize FLO/LFY orthologs zfl1 and zfl2. We observed associations with multiple traits, suggesting that these genes have pleiotropic functions. We caution, however, that while the lack of associations with zfl1 in populations carrying independent zfl1 mutations strongly suggests that the observed associations are indeed due to activity of the zfl genes themselves, further experiments will be necessary to ascertain that the pleiotropic functions suggested by the zfl-associated trends actually reside within the zfl1 and zfl2 loci.

We observed associations of zfl1 (and zfl2 to a lesser degree) genotype with flowering time. While the effect is relatively weak, this may be explained by the fact that many genes appear to affect flowering time in maize (Chardon et al. 2004). A role in flowering time has also been demonstrated for LFY in Arabidopsis with a dosage series experiment that clearly demonstrated that increasing functional LFY copies results in a quantitative dosage-dependent acceleration of flowering time analogous to what we observed in maize (Blázquez et al. 1997). Recently, a meta-analysis of maize flowering time QTL has implicated the zfl1 region in controlling flowering time variation among maize lines, suggesting zfl1 as a candidate locus for this trait (Chardon et al. 2004).

We also observed quantitative associations that implicate zfl1 and zfl2 in various aspects of branching. Lateral branch or ear shoot number and tassel branch number were primarily associated with zfl2, while branch placement was associated with both genes. The zfl genes are associated with increased lateral branch number in the tassel, but decreased lateral branch number during vegetative phases. This suggests that the effects of the zfl genes on branching are complex and depend on other genes. This is not surprising, as FLO/LFY orthologs from other species are associated with similarly complex effects on branching: the flower to shoot conversion observed in flo/lfy mutants in numerous species suggests that the wild-type role of these proteins is to repress branching by conferring floral identity onto lateral structures (Coen et al. 1990; Weigel et al. 1992; Hofer et al. 1997; Souer et al. 1998; Molinero-Rosales et al. 1999; Ahearn et al. 2001). In rice, RFL has been proposed to be involved in suppression of inflorescence branch formation (Kyozuka et al. 1998), while in Arabidopsis, LFY appears to be required for inflorescence branch initiation in a filamentous flower mutant background (Sawa et al. 1999).

Our data showed an association between the zfl genes, especially zfl2, and increased numbers of grain rows around the circumference of the ear. A role for FLO/LFY-like genes in controlling inflorescence phyllotaxy has not been reported in other species. However, FLO/LFY orthologs are involved in promoting whorled organ phyllotaxy during flower development in diverse species (Coen et al. 1990; Weigel et al. 1992; Hofer et al. 1997; Souer et al. 1998; Molinero-Rosales et al. 1999; Ahearn et al. 2001), and in tobacco the FLO/LFY ortholog NFL1 is also involved in controlling leaf phyllotaxy (Ahearn et al. 2001). Thus, our results suggest that a basic role of FLO/LFY orthologs in phyllotaxy may have been appropriated for a novel role in maize inflorescence phyllotaxy, perhaps through an expression pattern change.

Pleiotropic functions for FLO/LFY orthologs have been previously described in other species. For example, in addition to their roles in flower development and reproductive timing, FLO and LFY orthologs in some species also play roles in leaf compounding (Hofer et al. 1997; DeMason and Schmidt 2001), branching (Sawa et al. 1999; Bomblies et al. 2003), and shoot meristem organization (Ahearn et al. 2001). Pleiotropy of this sort has long been recognized as a common feature of developmental regulatory genes (Caspari 1952; Lande 1980). It has previously been suggested that selection pressure acting on traits controlled by pleiotropic genes may cause neutral or even detrimental traits to be selected due to correlation with beneficial traits (Mitchell-Olds 1996a,b; Conner 2002; Albertson et al. 2003). This situation may place limits on the degree to which a particular gene can be selected for its effect on a beneficial trait (Lande 1980; Maynard Smith et al. 1985). In the case of zfl2, the quantitative pleiotropic associations we observe suggest that lines selected for higher kernel row number via zfl2 would likely also flower slightly earlier, average fewer ears placed lower on the plant, and have more blank vegetative nodes. Thus, some domestication-associated traits, such as an increase in blank node number in maize, may be by-products of selection for resource allocation to fewer and larger ears that may have acted in part through zfl2. If any of the secondary effects associated with zfl2 were undesirable to early agriculturalists, these would limit the degree to which zfl2 might have been selected for its useful effects on ear rank or reduced ear number.

Functional divergence of zfl1 and zfl2:

Duplicate genes play a potentially important role in evolution since one or both redundant paralogs may be released from selective constraint for essential functions and thus provide potential “raw material” for evolution (Ohno 1970). Several models of duplicate gene evolution posit that most duplicates are likely to be lost through deleterious mutations, while some gene pairs may be retained via evolution of novel functions or subdivision of a complex ancestral function (Force et al. 1999; Lynch and Conery 2000).

The maize genome contains many duplicate regions (Ahn and Tanksley 1993; Berhan et al. 1993; Devos and Gale 1997; Gale and Devos 1998) due to a genome duplication thought to have occurred ∼11 million years ago in the lineage leading to maize and its relatives (Gaut and Doebley 1997). The maize zfl genes fall into a class of retained duplicates with largely redundant functions, as severe morphological defects are observed only in zfl1/zfl2 double-mutant plants (Bomblies et al. 2003). Despite their functional redundancy in flower development, we observed several differences in quantitative trends associated with the zfl genes. For example, zfl1 was more strongly associated with aspects of flowering time, while zfl2 was more strongly associated with trends affecting lateral branch number and ear phyllotaxy. These results suggest that zfl1 and zfl2 may be evolving subtle differences in function, perhaps through partitioning of a modular ancestral function as predicted by the sub-functionalization model for duplicate gene evolution (Force et al. 1999).

zfl2 and maize domestication:

One of the major differences between maize and teosinte is the phyllotaxy of the ear; teosinte ears are exclusively two-ranked, while maize ears have four or more ranks (Mangelsdorf and Reeves 1939). Langham (1940) argued that a single gene, two-ranked (tr), controls the difference between the two-ranked condition in teosinte and the higher-ranked condition of maize. A major effect region controlling ear rank was later localized to the short arm of chromosome 2 by introgressing this region from Tripsacum dactyloides (Magiure 1961; Galinat 1973), a species more distantly related to maize than teosinte, but also having two-ranked ears. Plants homozygous for the Tripsacum chromosome 2 introgression had (among other traits) fewer tassel branches, fewer total nodes, fewer rows of ovules (lower ear rank), and more ear shoots than siblings heterozygous or homozygous for the maize chromosome 2 (Magiure 1961). Interestingly, many of these trends are similar to trends we observed in plants carrying mutations in zfl2. More recently, QTL mapping experiments have also strongly implicated the chromosome 2 region within which zfl2 maps in the increase in ear rank associated with maize domestication (Doebley and Stec 1991, 1993). Together, these observations suggest the hypothesis that an increase in zfl2 activity during maize domestication may be responsible for multiple traits associated with chromosome 2, including the ear rank effect. As we have previously suggested (Bomblies et al. 2003), zfl may control phyllotaxy through effects on inflorescence meristem size and pattering.

To test the hypothesis that zfl2 activity might underlie this domestication QTL, we performed a QTL complementation test (Doebley et al. 1995; Mackay 2001). Ear rank showed associations with chromosome 2 only in hybrid populations that segregated for wild-type maize zfl2 alleles, and not in populations that segregated only teosinte and mutant maize zfl2 alleles, suggesting that the teosinte zfl2 alleles in these populations may be equivalent to loss-of-function maize zfl2 alleles in terms of their effect on ear rank. The association and QTL mapping results support the hypothesis that a functional maize zfl2 allele is important for the chromosome 2 ear rank QTL effect. We caution that these results do not necessarily indicate that the zfl2 gene itself was under selection during maize domestication, as they do not preclude the possibility that a linked gene acting through zfl2 was selected. An important implication, however, is that activity of zfl2- or a zfl2-dependent developmental pathway appears to be important for a large-effect maize domestication QTL.