As reported by Yamamoto and Ohsako 3 , mfr corresponds to CG5747, whose structure was predicted by the Drosophila Genome Project 19 . The lack of mfr ESTs recovered from various cDNA screens 20 21 and failure to detect mfr transcripts on Northern blots containing poly A+ RNA from 1,000 testes, 500 ovaries, and male and female carcasses (data not shown), indicates that mfr transcripts may be rare. Consequently, we used RT-PCR to search for mfr transcripts and detected transcripts in the testes and ovaries, but not in the carcasses of adult flies that lacked gonads (Figure 1A).

To determine the mfr transcript structure, we performed 5' and 3' RACE using poly A+ RNA isolated from testes and ovaries. Sequence analysis of 10 testicular and 15 ovarian 5' RACE products revealed multiple transcriptional start sites (Figure 1B). All six of the identified testis transcription start sites mapped within a 35 bp region. This pattern of multiple nearby start sites is characteristic of a slippery promoter, which has previously been described by Yasuhara et al., 22 for other Drosophila genes. In contrast, the three identified ovarian transcription start sites mapped over a wider span of 516 bp, with each site distinct from those identified in the testis.

The mfr promoters lack the motifs identified by Ohler et al. 23 as common to Drosophila promoters. However, we note two regions of interest as putative regulatory sequences that are highly conserved among Drosophila species (Figure 1B). A 22 bp sequence, located just downstream of the 5' most testes transcription start site is also found in the homologous mfr region in D. simulans, D. yakuba, D. ananassae, and shows 86% identity in D. pseudoobscura. A 17 bp sequence which is located 87 bp upstream of the most 5' ovarian start site shows complete conservation in D. simulans, D. yakuba, D. ananassae, D. pseudoobscura, and 76% identity in the more distantly related D. mojavensis, and D. virilis. The location of this sequence is also within the 5' UTR of Tsp66E, which is transcribed from the opposite strand of mfr orthologs in all of these species. Therefore, it may be an important regulatory sequence for either gene.

Sequence of 12 testicular and 10 ovarian 3' RACE products indicated that mfr has a single site for polyadenylation, which is located 103 nt downstream of a predicted stop codon. Together, the 5' and 3' RACE studies defined a minimal gene size of 6.761 kb, extending from the most upstream ovarian start site to the polyadenylation site. An estimate of maximum gene size was provided by testing a transgene containing a 10.7 kb genomic fragment for its ability to rescue mfr mutant phenotypes. Two independent insertions of the transgene fully rescued the mfr male sterility in a single copy (Table 1) and also rescued a defect in egg patterning, which is described in more detail below, in two copies (Table 2). Compared to the 6.761 kb region, this transgene includes an additional 131 bp of upstream sequence and ~3.85 kb downstream of the polyadenylation site.

We used the RACE results to design primer sets to recover near full-length mfr cDNAs from polyA+ selected testis and ovary RNA. Analysis of six testicular and nine ovarian cDNAs revised the CG5747 predicted gene structure 19 and identified differently spliced mRNAs with the potential to yield seven different Mfr protein isoforms (Figures 2 and 3).

The longest cDNA (T1), which was recovered from testis RNA, contains 14 exons and an open reading frame (ORF) with three in-frame Met codons located in exon 1 (Figures 1B and 2A). ATG#1 may be the initiating methionine codon for the mRNA corresponding to cDNA T1 because the sequences between ATG#1 and ATG#3 in D. melanogaster and its sister species are largely conserved (D. simulans with 93% identity, D. sechellia 74% identity, D. yakuba 74% identity, and D. erecta 76% identity), suggesting a selective pressure to maintain the region as coding. The longest ORF predicts a protein with 1,684 amino acids, a TM domain near the C-terminus, and five C2 domains (Figure 2B). The C2 domains are denoted C2B-C2F based on their homology to C2 domains in mammalian ferlin proteins 2 . Three additional testis cDNAs (cDNA T2–T4) reflect alternative splicing events that increase the length of the 5' UTR and shorten the length of the ORF (Figure 2A). Corresponding protein isoforms are predicted to have two or three C2 domains, with the shortest isoform (T4) lacking the TM domain (Figure 2B).

Analysis of nine ovarian cDNAs revealed three different splicing variants, and all were different from those seen in testis cDNAs (Figure 3A). The ovarian transcripts vary in coding capacity and the length of the 5' UTR. Corresponding protein isoforms are predicted to have between one and three C2 domains, and only isoform O2 is predicted to have a TM domain (Figure 3B).

For each of the 11 mfr mutations, we assayed the fertility of males that were homozygous or hemizygous in combination with deficiencies Df(3L)h-i22 or Df(3L)ED4415. We also tested the fertility of males with several mfr heteroallelic combinations (see methods). The result in all cases was strong recessive male sterility, with no progeny produced from any of the crosses between mfr males and wild-type females.

To understand how mfr mutation might affect ferlin function, we identified the molecular lesions in each of the 11 mutations. The mutations are distributed throughout the gene and provide information about the functional significance of predicted isoforms and specific residues (Table 3 and Figure 4). In particular, they implicate isoform T1 as essential for male fertility as it is the only isoform expected to be disrupted by all of the mutations. The missense mutation mfr ^Z1250, which changes a methionine to an isoleucine, provides additional evidence that ATG#1 (Figure 1B) encodes the initiating methionine. Also, mfr ^Z6248, which, like mfr ^Z1250, is expected to affect only isoform T1, changes a valine that is conserved in D. simulans, D. sechellia, D. yakuba, and D. erecta to aspartic acid in the region of the protein that is predicted to form the C2C domain. Because valine is a hydrophobic nonpolar amino acid and aspartic acid is a negatively charged amino acid, the mfr ^Z6248 mutation likely disrupts C2C structure. In addition, mfr ^Z1021 replaces a glycine, which is located between C2E and C2F, to a serine. This glycine is conserved in all known and predicted ferlin proteins in other species, indicating a key functional role in ferlins.

Mapping of the mutations also revealed a disproportionately high number of nonsense mutations. Eight mutations are predicted to truncate isoform T1, either by mutating a splice junction (mfr ^Z2942) or in seven cases, by changing an amino acid to a stop codon (Table 3 and Figure 4). Three of the nonsense mutations, mfr ^Z1386, mfr ^Z4070, and mfr ^Z0695 are the best candidates for mfr null mutations since they map to the C2E-C2F region that is shared by all of the predicted testis isoforms.

To characterize the cause of mfr male sterility, we introduced a transgene that expresses the Don-Juan GFP (DJ-GFP) sperm tail marker into five different mfr mutant backgrounds, then monitored efficiency of sperm entry into the egg. Studies with a control strain show that the DJ-GFP assay detects 87% of the sperm entry events in eggs for up to 90 min after egg deposition (see methods). During normal fertilization, sperm entry is followed by events of sperm activation, including sperm plasma membrane breakdown (PMBD), nuclear decondensation, and formation of male pronucleus, in rapid succession. We found that in 85–100% eggs laid by wild type females mated to mfr males, sperm entered the egg (Table 4). However, nearly all retained a highly condensed nucleus (Figure 5). Three mfr alleles, including two nonsense alleles, occasionally allowed sperm to undergo nuclear decondensation and reach the pronuclear apposition stage, but not beyond (Table 4).

Previous studies indicated that the mfr ¹ allele 16 , like snky mutations 18 24 , affects PMBD. To monitor PMBD in eggs fertilized by sperm produced by mfr ^Z0695 males, we used the method of Wilson et al. 24 . Specifically, the membrane protein CD2 was introduced into the sperm plasma membrane through expression of a CD2 transgene during spermatogenesis and its presence was monitored by immunolocalization before and after sperm entry into the egg. We found that the CD2 epitope was retained around the head and tail of mfr ^Z0695 mutant sperm for at least 20 min after egg deposition in 83% of the eggs we monitored (n = 12, Figure 6), supporting an effect of mfr on PMBD.

Because snky is also important for sperm PMBD 18 24 , Mfr could affect this event by controlling the localization of Snky. To test this idea, we introduced Snky-GFP into sperm produced by mfr ^Z0695 males and assayed GFP expression. As shown in Figure 7, this mfr nonsense mutation does not alter the localization of Snky-GFP to the acrosomal membrane 24 .

Expression of mfr transcripts in ovaries suggested a possible function during oogenesis. This role was confirmed by the discovery a recessive egg patterning phenotype induced by a subset of mfr mutations when expressed in females. The defect is visible in late stage eggs as abnormally short and closely apposed dorsal appendages (Figure 8A and 8B), a feature of the eggshell that signifies a ventralized egg chamber 25 .

Penetrance of the egg patterning defect varied among the alleles (Figure 8C), with females hemizygous for the mfr ^Z0695, mfr ^Z1386, or mfr ^Z4070 allele producing 14–22% abnormal eggs. The fact that these three nonsense alleles induced defects, but four nonsense mutations located upstream of the C2D encoding region do not (Figure 4), suggests that normal egg patterning requires a short Mfr isoform. Interestingly, two additional mutations that map downstream of C2D do not induce egg patterning defects. These mutations are mfr ^Z1021, which changes a glycine to a serine, and mfr ^Z2942, which mutates a splice junction to create a premature stop codon in some transcripts. Based on the locations of the nonsense mutations, the ovarian transcript that is important for egg patterning likely encodes the C2E and C2F domains, but does not use the splice junction affected by the mfr ^Z2942 mutation. This structure is not represented among the three identified ovarian cDNAs.

We note that penetrance of the mutant phenotype in eggs produced by hemizygous females does not exceed 22% in any case, indicating that Mfr likely has a redundant function in ovaries. Surprisingly, we found that the percentage of egg patterning defects significantly increases to as high as 41% when the mfr ^Z0695, mfr ^Z1386, and mfr ^Z4070 are placed in heteroallelic combinations (Table 5, Figure 8C). This observation verifies that Mfr proteins are expressed during oogenesis and suggests that the truncated Mfr proteins induce a dose-sensitive disruptive effect on egg patterning.

Next, we investigated whether the defect might be accounted for by disruption of Gurken, a key ligand in the egg patterning pathway. In normal stage 10 egg chambers, Gurken shows a characteristic distribution in the anterior dorsal corner of the oocyte. This distribution permits proper signaling to overlying follicle cells 26 . Follicle cells that receive the Gurken ligand activate a downstream pathway that is responsible for the formation and proper spacing of the dorsal appendages 27 . Immunolocalization of Gurken revealed that in the majority of stage 10 egg chambers produced by control bw;st or mfr ^Z4070/mfr ^Z1386 females, Gurken is properly localized (Figure 9A). However, in 15% of the bw;st egg chambers (n = 46) and 43% of the mfr ^Z4070/mfr ^Z1386 egg chambers (n = 44), Gurken does not extend as far posteriorly (Figure 9B). A χ² likelihood ratio test reveals that this difference is statistically significant (P = 0.0025, R² = 0.0861).

Mutations in mfr also affect embryonic development. Notably, only 43% of the total eggs laid by mfr ^Z4070/mfr ^Z1386 mutant females mated to wild-type males hatch (n = 376 eggs). This hatch rate is similar for eggs with wild-type eggshell morphology and those with defective dorsal appendages; therefore, Mfr's role in embryogenesis may be separate from its role in egg patterning. Furthermore, the hatch rate of embryos of mfr ^Z4070/mfr ^Z1386 mothers is significantly lower than the 84% hatch rate of embryos produced by the mfr ^Z4070 or mfr ^Z1386/TM6 control strain (P < 0.0001 R² = 0.10 n = 340 eggs) and the 52% hatch rate observed for the bw;st parent strain (P = 0.0232 R² = 0.0049 n = 392 eggs).

To determine if the lower hatch rates are caused by visible defects during early embryonic development, mfr ^Z4070/mfr ^Z1386 and control (mfr ^Z4070 or mfr ^Z1386/TM6 and bw;st) females were mated to wild-type males and a minimum of 150 eggs were collected from each mating at 50 and 90 minutes after egg deposition. The embryos were fixed and stained with DAPI to visualize nuclei, and then classified according to mitotic cycle (Figure 10). This analysis revealed that the development of embryos produced from mfr ^Z4070/mfr ^Z1386 mothers is delayed compared to embryos produced from mothers of the two control strains (mfr ^Z4070 or mfr ^Z1386/TM6 and bw;st). Furthermore, an ordinal logistic regression analysis confirms that the effect of maternal strain on the stage of the embryonic development is statistically significant (P < 0.0001, X² = 58.31, d.f. = 2). This effect remains statistically significant when pairwise comparisons between the three strains are made.

Compared to mfr ^Z4070 or mfr ^Z1386/TM6, embryos produced by bw;st females have low hatch rates. However, if the embryos arrested before a the end of cycle 1 are removed from the data set and a subsequent ordinal logistic analysis is performed, the effect of strain on mitotic stage is no longer significant between the bw;st parent strain and the mfr ^Z4070 or mfr ^Z1386/TM6 control strain (P = 0.86, χ² = 0.03). This results suggests that embryos from bw;st mothers show defects in completing cycle 1, but if the embryos complete cycle 1, then the distribution of mitotic stages is similar to embryos from the mfr ^Z4070 or mfr ^Z1386/TM6 mothers through cycle 14. Conversely, after the cycle 1 data is removed, maternal strain retains a significant effect on mitotic stage when embryos from mfr ^Z4070/mfr ^Z1386 mothers are compared to embryos from both bw;st (P = 0.0016 χ² = 9.99) and mfr ^Z4070 or mfr ^Z1386/TM6 (P < 0.0001 χ² = 15.40) mothers. This result indicates that mitotic defects in embryos from mfr ^Z4070/mfr ^Z1386 mothers are not limited to the completion of cycle 1, but instead extend into later cycles.

Strains are described on FlyBase 37 or by Wilson et al. 24 . The mfr mutations (mfr ^Z2713, mfr ^Z0695, mfr ^Z2942, mfr ^Z1021, mfr ^Z6248, mfr ^Z1386, mfr ^Z4070, mfr ^Z3281, mfr ^Z3323, mfr ^Z1250, and mfr ^Z4901) were recovered in a screen for ethyl methanesulfonate (EMS)-induced male-sterile mutations in a bw;st background 17 from the Zuker collection 38 . The mfr ¹ stock was provided by T. Ohsako and M-T. Yamamoto 16 .

The two deficiency chromosomes, Df(3L)h-i22 and Df(3L)ED4415, were used in this study. Although they behaved identically in failing to complement the male sterility associated with each mfr mutation, Df(3L)h-i22 was variable compared to Df(3L)ED4415 in complementation assays of mfr/Df females. Both chromosomes lack the mfr gene, as verified using a PCR assay on embryos homozygous for the deficiencies (K. Okada, personal communication). Therefore, we attributed the difference to a possible genetic suppressor of the female phenotype in the Df(3L)h-i22 stock and used Df(3L)ED4415 for assaying hemizygous female genotypes.

Transgenic lines containing the mfr genomic region were constructed using a 10.7 kb BssHII/BsiWI fragment isolated from BACN18D24 (Children's Hospital Oakland Research Institute, Oakland, CA). After attachment of NotI linkers (New England BioLabs, Ipswich, MA), the fragment was cloned into the pCasPeR4 transformation vector. Standard germ line transformation techniques and genetic crosses were used to construct two independent strains (A and B) of the genotype w ¹¹¹⁸; P [w ^+mc mfr ^+t10.7]A or B/SM1; st mfr ^Z0695/TM6 and test for rescue of the mfr mutant phenotypes.

RNA was harvested from testes, ovaries, and carcasses of male and female flies that were 1–3 days old using the RNeasy kit (Qiagen, Valencia, CA). PolyA+ RNA was isolated by the MicroPoly(A) Purist Kit (Ambion, Austin, TX). To determine the tissue specificity of mfr, first strand cDNA synthesis was performed with SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA). RT-PCR was carried out on cDNA with mfr gene specific primers (located in exon 8 and 14 in testis cDNAs, and exon 7 and 13 in ovary cDNAs) and yip-6 control primers that span an intron. Sequences of these and other primers used in this study are available upon request. PCR products from the testes and ovaries were cloned in the TOPO-TA vector (Invitrogen, Carlsbad, CA), and generated clones were sequenced using Big Dye Terminator Cycle Sequencing Reaction Kit (Applied Biosystems, Foster City, CA) and an ABI/PRISM 3100 Genetic Analyzer.

For RACE analysis on the testes and ovaries, first-strand cDNA synthesis was performed with PowerScript™ Reverse Transcriptase (SMART RACE cDNA Amplification Kit, Clontech, Mountain View, CA) for 5'RACE and MMLV Reverse Transcriptase (First Choice RLM-RACE Kit, Ambion, Austin, TX) for 3' RACE. For both 5' and 3' RACE, PCR products were generated with two different primer sets and cloned using the TOPO-TA vector. Generated clones were sequenced as described above. For 5' RACE, the PowerScript™ Reverse Transcriptase can add between 3–5 C residues to the first strand of cDNA, an activity that later creates ambiguity for defining the exact start site of the cDNA if the corresponding genomic sequence contains one or more Gs. This ambiguity occurred for a subset of mfr start sites and is noted by the brackets in Figure 1B.

Sequences determined to be at the 5' and 3' end of the mfr transcript by RACE analysis were used to design primer sets to isolate near full-length cDNAs. PCR was performed under conditions optimized for long products (Qiagen, Valencia, CA) with a combination of ProofStart DNA polymerase (Qiagen, Valencia, CA) and Taq DNA polymerase (Promega, Madison, WI), and the products were cloned into the TOPO-TA vector. Generated clones of six testis and nine ovary cDNAs were selected and sequenced as described above. Sequences of cDNAs are available in GenBank as: EF120975 (T1), EF120976 (T2), EF120977 (T3), EF120978 (T4), EF120979 (O1), EF120980 (O2), and EF120981 (O3).

To determine the location of the mfr mutations, genomic DNA was extracted from males of the bw;st parent strain or mfr strains. Overlapping primer sets were used to amplify the entire mfr ORF and splice junctions, except 48 bp of the 5' end of exon 2. PCR products were then sequenced. Each mutation was verified by sequencing two independent products.

To assess fertility, single males were mated to three Canton-S (CS) virgin females and cultures were inspected for the presence of progeny over the course of 15 days. At least 10 males were assayed for each mfr allele in hemizygous combinations with Df(3L)h-i22 or Df (3L)ED4415 and for a subset of mfr alleles in heteroallelic combination (mfr ^Z0695/mfr ^Z1386, mfr ^Z2942/mfr ^Z3281, mfr ^Z2942/mfr ^Z4070, mfr ^Z2942/mfr ^Z1386, mfr ^Z2942/mfr ^Z6248, mfr ^Z2942/mfr ^Z1021, mfr ^Z2942/mfr ^Z1250, mfr ^Z2942/mfr ^Z4901). Control crosses used males from the bw;st parent strain or sibling mfr/Balancer males.

In order to compare efficiencies of fertilization of mfr and normal sperm, a transgene expressing the don juan-GFP (dj-GFP) sperm tail marker 39 was introduced into the mfr mutant background to create w; dj-GFP; st mfr/Df(3L)h-i22 males for five alleles: mfr ^Z0695, mfr ^Z6248, mfr ^Z2942, mfr ^Z3281, and mfr ^Z2713. Males were mated to CS females and laid eggs were collected up to 90 minutes after egg deposition then processed as described below to assay GFP. dj-GFP; mfr ⁺ males were used in a control cross to monitor the efficiency of the assay. In this cross, 98% of the eggs hatch (n = 335 eggs), and we detected the DJ-GFP sperm tail marker in 87% of the eggs (n = 60 eggs).

Eggs laid by females were dechorionated in 50% bleach, fixed in 4% paraformaldehyde with an octane overlay, devitellinized with methanol, stained with 1 μg/ml DAPI, and then mounted on slides in Vectashield (Vector Laboratories Inc., Burlingame, CA). Preparations were viewed using a BioRad Radiance 2000 LSCM confocal microscope in conjunction with a Spectra-Physics Mai Tai Laser, a 488 nm Kr/Ar laser, and a Nikon Eclipse E600-FN fluorescent scope. Z-series stacks were compiled using NIH Image and images were edited using Adobe Photoshop 7.0.

Two additional transgenes, expressing CD2 or Snky-GFP, were introduced into a mfr ^Z0695 mutant background. Assays for protein expression in sperm were performed as described by Wilson et al. 24 except a 1:25 dilution of the anti-rat CD2 monoclonal antibody (Harlan Sera-Labs Ltd, Loughborough, UK) and a 1:200 dilution of the Alexa 488 rabbit anti-mouse and goat anti-rabbit antibodies (Alexa 488 Signal-Amplification kit, Molecular Probes, Eugene, OR) were used.

To assay for mfr effects in females, crosses were set up to generate siblings with control and mfr mutant genotypes. These crosses allowed morphological comparisons of eggs laid by bw/+; st mfr ⁺/Df(3L)ED4415 control females to bw/+; st mfr/Df(3L)ED4415 females for each of the 11 mfr alleles, and to eggs laid by heteroallelic bw/+; st mfr ^Z0695/st mfr ^Z1386, bw/+; st mfr ^Z0695/st mfr ^Z4070 and bw/+; st mfr ^Z1386/st mfr ^Z4070 females. A minimum of 176 eggs was examined for each female genotype. Eggs were classified as having either wild-type or mutant morphology based on length and distance of the two dorsal appendages. Eggs were photographed using a Nikon photomicroscope equipped with a CoolSNAP cool-charged coupled device camera (RS Photometrics, Tucson, AZ).

Additional assays in females focused on mainly on the mfr ^Z4070 and mfr ^Z1386 mutations. To examine localization of Gurken during oogenesis, ovaries from bw;st and bw;st mfr ^Z4070/st mfr ^Z1386 females were dissected and fixed in 4% paraformaldehyde in 1 × PBS and heptane. The ovaries were incubated overnight with a 1:100 dilution of a primary mouse Gurken antibody 40 (Developmental Studies Hybridoma Bank, Iowa City, IA) and then subsequently with a 1:1000 dilution of the Alexa 488 goat anti-mouse secondary antibody (Molecular Probes, Eugene, OR). The ovaries were stained with 0.1 mg/ml DAPI, cleared in 80% glycerol overnight at 4°C, then mounted on coded slides so scoring could be done without knowledge of the genotype. Localization of the Gurken signal was scored as abnormal if the signal did not extend posteriorly beyond the oocyte nucleus in stage 10 egg chambers.

To assess female fertility and maternal effects on embryogenesis, bw;st, bw/+;st mfr ^Z1386/st mfr ^Z4070, and bw/+;st mfr ^Z4070 or st mfr ^Z1386/TM6 females were mated to wild-type CS males. Eggs were collected then held at 25°C for 50 min, 90 min, or 48 hours. After 50 or 90 minutes, embryos were dechorionated in 50% bleach, rinsed in 0.2% NaCl/0.2% Triton-X and water, and treated with 0.5 M EGTA. The embryos were vortexed in a 50:50 mixture of octane and methanol. The octane layer was removed and the embryos were rinsed in methanol and a step-wise increase in 0.1% Triton X in 1 × PBS. The embryos were stained in 0.1 μg/ml DAPI, placed in 90% glycerol at 4°C overnight, and staged according to the number of nuclei in the embryo. Total nuclei were counted for embryos in cycle 1 through cycle 8. After cycle 8, the total number of nuclei was extrapolated from the number of nuclei along the embryo periphery. For the embryos held for 48 hours, the hatch rate was determined by counting the unhatched embryos.

All statistical analyses were performed with the program JMP (SAS Institute, Cary, NC). χ² likelihood ratio tests were used to compare the goodness-of-fit between models. Because multiple tests can increase type 1 error, when three pairwise comparisons were made, the sequential Bonferroni correction was used to determine the critical P-value (lowest P < 0.0167, middle P < 0.025, and highest P < 0.05). An ordinal logistic regression model was used to analyze the distribution of embryonic mitotic stages for Figure 10.

The predicted Mfr protein structure was evaluated by SMART 41 42 . The C2 domain closest to the C terminus (C2F) was not predicted by SMART, so the boundaries of this domain were determined by homology to predicted C2F domains of Mfr orthologs in other Drosophila species. Percent identity between sequences was determined with ClustalW 43 .