Identification of tethering domains for protein kinase A type Ialpha regulatory subunits on sperm fibrous sheath protein FSC1.

The fibrous sheath is a unique cytoskeletal structure in the sperm flagellum believed to modulate sperm motility. FSC1 is the major structural protein of the fibrous sheath. The yeast two-hybrid system was used to identify other proteins that contribute to the structure of the fibrous sheath or participate in sperm motility. When FSC1 was used as the bait to screen a mouse testis cDNA library, two clones were isolated encoding the type Ialpha regulatory subunit (RIalpha) of cAMP-dependent protein kinase. Deletion analysis using the yeast two-hybrid system and in vitro binding assays with glutathione S-transferase-FSC1 fusion proteins identified two RIalpha tethering domains on FSC1. A domain located at residues 219-232 (termed domain A) corresponds to the reported tethering domain for a type II regulatory subunit (RII) of cAMP-dependent protein kinase, indicating that this binding domain has dual specificity to RI and RII. Another RIalpha tethering site (termed domain B) at residues 335-344 shows specific binding of RIalpha and had no significant sequence homology with known RII tethering domains. However, helical wheel projection analysis indicates that domain B is likely to form an amphipathic helix, the secondary structure of RII tethering domains of protein kinase A anchoring proteins. This was supported by the finding that site-directed mutagenesis to disrupt the amphipathic helix eliminated RIalpha binding. This is apparently the first report of an RIalpha-specific protein kinase A anchoring protein tethering domain.

The fibrous sheath is a unique cytoskeletal structure in the sperm flagellum believed to modulate sperm motility. FSC1 is the major structural protein of the fibrous sheath. The yeast two-hybrid system was used to identify other proteins that contribute to the structure of the fibrous sheath or participate in sperm motility. When FSC1 was used as the bait to screen a mouse testis cDNA library, two clones were isolated encoding the type I␣ regulatory subunit (RI␣) of cAMP-dependent protein kinase. Deletion analysis using the yeast two-hybrid system and in vitro binding assays with glutathione S-transferase-FSC1 fusion proteins identified two RI␣ tethering domains on FSC1. A domain located at residues 219 -232 (termed domain A) corresponds to the reported tethering domain for a type II regulatory subunit (RII) of cAMP-dependent protein kinase, indicating that this binding domain has dual specificity to RI and RII. Another RI␣ tethering site (termed domain B) at residues 335-344 shows specific binding of RI␣ and had no significant sequence homology with known RII tethering domains. However, helical wheel projection analysis indicates that domain B is likely to form an amphipathic helix, the secondary structure of RII tethering domains of protein kinase A anchoring proteins. This was supported by the finding that site-directed mutagenesis to disrupt the amphipathic helix eliminated RI␣ binding. This is apparently the first report of an RI␣-specific protein kinase A anchoring protein tethering domain.
The mammalian sperm flagellum consists mainly of cytoskeletal structures, including the axoneme, the outer dense fibers, and the fibrous sheath. The axoneme is formed of microtubles organized in a "nine plus two" arrangement, extends the full length of the flagellum, and functions as the motor of the sperm. It is surrounded by the outer dense fibers throughout most of its length. The outer dense fibers are enclosed by the mitochondrial sheath in the middle piece and by the fibrous sheath in the principal piece of the flagellum (1). The outer dense fibers and the fibrous sheath are believed to stiffen the flagellum and to modulate its bending (2,3). In addition, several spermatogenic cell-specific glycolytic enzymes are associated with the fibrous sheath (4 -6). This structure may be a scaffold for enzymes that produce energy required for hyperactivated motility of sperm at the time of fertilization. Further-more, cAMP-stimulated protein phosphorylation is essential for initiation or maintenance of sperm motility (7)(8)(9) and proteins in the fibrous sheath are subject to phosphorylation.
The coordinated regulation of many cellular mechanisms is mediated through cAMP-dependent protein kinase (PKA). 1 There has been much interest in how cAMP can act as a second messenger to regulate different mechanisms at specific sites distributed throughout the cytoplasm and nucleus. One way this is achieved appears to be through diversity of PKA components. PKA is a tetramer of two catalytic (C) and two regulatory (R) subunits, and genes encoding three C subunits and four R subunits have been identified (10 -13). RI␣ and RII␣ subunits are expressed ubiquitously, while RI␤ and RII␤ subunits are expressed mainly in brain and testis. In addition, protein kinase A anchoring proteins (AKAPs) have been identified that are present in specific subcellular sites. AKAPs have tethering domains for regulatory subunits that place PKA in close proximity to specific organelles or cytoskeletal components to localize phosphorylation events that occur in response to activation signals (14 -16).
Ligand overlay assays have indicated that only RII subunits of PKA are tethered to AKAPs. The observation that RII subunits bind to structural proteins in the sperm flagellum (17)(18)(19) was consistent with RII subunits being found mainly in the particulate fraction and RI subunits in the cytosolic fraction of cell homogenates. However, yeast two-hybrid screens recently isolated D-AKAP1 and D-AKAP2 proteins that interact with both RI␣ and RII␣ (20,21). Although D-AKAP1 has a 25-fold lower affinity for RI␣ than for RII␣, this suggests that RI subunits also may determine the subcellular localization of PKA.
Deletion and point mutagenesis approaches have been used with ligand overlay assays to map RII tethering domains on AKAPs. These domains have limited consensus sequences but are predicted to form secondary structures of amphipathic helices that containing acidic amino acids in the hydrophilic region and amino acids with a long aliphatic side chain in the hydrophobic region (14,22,23). The location of the RI tethering domain was not determined for D-AKAP1 or D-AKAP2, but the binding region for RI partially or completely overlaps that for RII in D-AKAP1 (20,21).
To better understand the function of the fibrous sheath of mouse sperm, we have used immunological, biochemical, and molecular biology approaches to identify some of its components. These include three enzymes, glutathione S-transferase (24), glyceraldehyde-3-phosphate dehydrogenase (4), and type I hexokinase (5), and the major structural protein of the fibrous sheath, referred to as fibrous sheath component 1 (25,26). It is remarkable that three of these are products of genes expressed only in spermatogenic cells (Gstm5, Gapds, and Fsc1), while the other is the product of spermatogenic cell-specific transcripts (Hk1-s). Northern analysis and in situ hybridization demonstrated that Fsc1 transcription first occurs in the postmeiotic phase of spermatogenesis when the fibrous sheath is assembled. Other investigators isolated a cDNA encoding a fibrous sheath protein referred to as AKAP82 that is identical to FSC1, except for lacking 9 amino acids at the N terminus (27). An important finding was that RII subunits bind to a 14-amino acid region of AKAP82 (28). This region was predicted to form an amphipathic helix, and the binding in the ligand overlay assay was inhibited by a synthetic peptide of the corresponding sequence (28).
The studies reported here used the yeast two-hybrid system, deletion mutagenesis, and an in vitro binding assay to demonstrate that two RI␣ and one RII␣ tethering domains are present on FSC1. While RI␣ and RII␣ bound to domain A, only RI␣ bound to domain B. This suggests that the amphipathic helix formed by domain B has a secondary structure distinct from that required for RII binding. Taken together with the immunocytochemical evidence that RI associates with the fibrous sheath and outer dense fibers (29), these results suggest that tethering of RI␣ to FSC1 may provide a mechanism for the subcellular localization of PKA to a region of the flagellum important for sperm motility.
Construction of Fsc1 Expression Plasmids-PCR was used to produce the full-length Fsc1 protein coding sequence (GenBank TM accession number U10341) or various deletion mutants. Forward primers containing NdeI sites and reverse primers containing SalI sites were used to generate PCR products for ligation into the pAS2-1 (CLONTECH) yeast two-hybrid system expression plasmid (Table I). Numbering of the amino acid sequence of FSC1 was that of Fulcher et al. (26). The PCR reaction contained 20 ng of mouse Fsc1 cDNA in gt11 (26), 10 pmol of each primer, a 0.1 mM concentration of each of the four deoxynucleotide triphosphates, and Pfu DNA polymerase and reaction buffer (Stratagene, La Jolla, CA) in a 50-l reaction volume. Amplifi-cation was performed for 30 cycles with a temperature profile of 30 s at 94°C, 1 min at 55°C, and 4.5 min at 72°C. The PCR products were digested with NdeI and SalI restriction enzymes, ligated into plasmid pAS2-1 digested with NdeI and SalI, and then transformed in Escherichia coli DH5␣ competent cells (Life Technologies, Inc.). The resulting constructs allowed FSC1 to be expressed in yeast as a fusion protein with GAL4.
To produce constructs that expressed FSC1 protein in E. coli as a fusion protein with glutathione S-transferase, PCR forward primers were used that contained the BamHI sequence to allow ligation into the corresponding site in pGEX-4T-1 (Amersham Pharmacia Biotech). Other conditions for construction of these plasmids were as described above.
The coding sequence of Fsc1 was subcloned into pBluescript II KS ϩ (Stratagene), and then used as a template to introduce the point mutation by using the QuikChange TM site-directed mutagenesis kit (Stratagene). Sequencing of the targeted site was carried out using T7 Sequenase version 2.0 (Amersham Pharmacia Biotech). PCRs using the appropriate primer set were used to amplify DNA sequences corresponding to mutant FSC1 proteins covering residues 237-361. The products were ligated into pGEX-4T-1 to produce constructs that expressed mutant FSC1 proteins in E. coli as described above.
Yeast Two-hybrid Screening of cDNA Library-Yeast Y190 competent cells were prepared with the Yeastmaker yeast transformation system (CLONTECH), based on the lithium acetate method described by Gietz et al. (30), transformed with plasmid pAS/Fsc/1-849 containing the full-length Fsc1 cDNA, and cultured on SD/Trp Ϫ agar plates for 4 days at 30°C. Selected yeast cells containing pAS/Fsc/1-849 were cultured in SD/Trp Ϫ medium and used as the carrier of the bait vector for sequential transformation. A mouse testis cDNA library constructed in the pGAD10 vector (CLONTECH) was screened using the yeast two-hybrid procedure. The pAS/Fsc/1-849 and pGAD10 co-transformants were cultured on SD/Trp Ϫ Leu Ϫ His Ϫ agar plates for 7 days at 30°C.
Detection of FSC1-GAL4 Fusion Protein in Yeast-Yeast cells were transformed with pAS2-1 deletion mutants of Fsc1 and cultured overnight in 1 ml of SD/Trp Ϫ Leu Ϫ selection medium at 30°C. Nine ml of YPD medium then was added, and growth was allowed to proceed to the midlogarithmic phase. Yeast proteins were extracted by the urea/SDS method as described (31) using 250 l of cracking buffer. Proteins from 5 l of extract were separated by SDS-PAGE using 10% (w/v) acrylamide gel and transferred onto Imobilon TM nylon membranes (Millipore

5Ј-CCGCATATGCAGAGGTCAGTTGCCACTCCTGAG-3Ј
237 Forward Corp., Bedford, MA). The membranes were soaked in TBST (150 mM NaCl, 0.1% (v/v) Tween 20, 50 mM Tris-HCl, pH 7.4) solution containing 2% (w/v) gelatin for 1 h and then incubated for 1 h with monoclonal antibody (0.5 g/ml) to the GAL4 DNA-binding domain (CLONTECH) in TBST. After extensive washing with TBST, membranes were incubated for 30 min with 10% (v/v) goat serum in TBST. After washing briefly, membranes were incubated for 30 min with horseradish peroxidase-conjugated goat antiserum to mouse IgG (1:30,000 dilution in TBST) (Sigma) preabsorbed for 30 min with 0.5 mg/ml yeast proteins to eliminate nonspecific antibodies against common yeast proteins. Yeast proteins were prepared from Y190 strain cells sonicated in PBS for 3 min followed by incubation for 30 min with 1% (v/v) Triton X-100 to solubilize membrane proteins. After extensive washing of the membrane with TBST solution, FSC1-GAL4 fusion protein was detected using ECL TM reagents (Amersham Pharmacia Biotech) according to procedures recommended by the supplier.
In Vitro Binding Assay of PKA Regulatory Subunits-E. coli transformed with pGEX-4T-1 plasmid encoding Fsc1 were grown to the midlogarithmic phase at 37°C in LB medium. They were cultured for an additional 90 min at 30°C in the presence of 0.1 mM isopropyl-␤-Dthiogalactopyranoside to induce synthesis of the fusion protein. Crude extracts were prepared by sonicating bacteria in PBS for 30 s followed by incubation for 30 min with 1% (v/v) Triton X-100. Crude extract (15-150 l) was incubated for 2 h at 4°C with 30 l of glutathione-Sepharose resin (Amersham Pharmacia Biotech) in 0.5 ml of PBS, followed by extensive washing with PBS to remove nonspecifically bound proteins.
Extracts from testes of 3-5-month-old CD-1 mice were prepared by homogenization in extraction buffer (140 mM NaCl, 0.1% (v/v) Triton X-100, 0.1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, 10 mM Tris-HCl, pH 7.5). After centrifugation at 10,000 ϫ g for 10 min, supernatant corresponding to 150 g or 1.5 mg of protein was added to the tube containing GST-FSC1 fusion protein immobilized on resin. The volume was adjusted to 0.5 ml with extraction buffer, and the mixture was incubated for 2 h at 4°C. After washing extensively with PBS, proteins were eluted by boiling in SDS gel-loading buffer, the eluates were divided into three parts, and proteins were separated by SDS-PAGE using 10% (w/v) acrylamide gels prepared according to Laemmli (32). Each lane contained that portion of 50 or 500 g of protein from testis extracts that bound to GST-FSC1, while the positive control lane contained 10 g (Fig. 3, lane T) or 5 g (Fig. 5, lane 1) of total testis extract. Regulatory subunits of PKA were detected by Western blotting using rabbit antisera against mouse RI␣, RII␣ (Santa Cruz Biotechnology, Inc., Santa Cruz, CA; 1:1000 dilution each) or RI␤ (1:500 dilution; a gift of Dr. Daniel W. Carr, Oregon Health Sciences University, Portland, OR). Secondary antibody was horseradish peroxidase-conjugated goat antiserum to rabbit IgG (1:10,000 dilution; Santa Cruz Biotechnology). Western blotting was performed as described above except that second antibody was not preabsorbed. Other gels were stained with Coomassie Brilliant Blue to detect proteins. Prestained standard proteins (Bio-Rad) were used to estimate protein mass.

RESULTS
Yeast Two-hybrid Screening of Testis cDNA Library-The yeast two-hybrid procedure (33) was used to screen a mouse testis cDNA library to identify proteins that interact directly with FSC1. Yeast Y190 cells were transfected with a bait plasmid containing the mouse Fsc1 cDNA protein coding region followed by sequential transformation with library plasmid DNA. Secondary transformants corresponding to 5 ϫ 10 6 clones were screened for interaction with FSC1 based on growth on medium lacking histidine. Positive clones were tested for transactivation of LacZ as another marker gene. Of the eight clones positive in both assays, two (FA7 and FC9) were subjected to further analysis in this study.
To confirm association with FSC1, plasmids for the FA7 and FC9 clones were rescued from yeast cells, cloned and amplified in E. coli, and then used to retransform yeast cells. Transformants were selected on medium lacking tryptophan and/or leucine, according to the type and combination of plasmid vectors used. Only transformants that received pAS/Fsc/1-849 in combination with plasmids containing pFA7 or pFC9 grew on SD/Trp Ϫ Leu Ϫ His Ϫ , while transformants that received pAS/ Fsc/1-849, pFA7, or pFC9 alone did not grow (data not shown). Transformants that received plasmid vector(s) were also cultured on YPD medium and used for colony lift assays to detect ␤-galactosidase expression (data not shown). This assay gave the same results as the growth selection assay, confirming that transactivation was due to association between FSC1 and proteins encoded by pFA7 and pFC9. These results also suggested that the interaction is direct and that no additional factors are required for association.
Sequence analysis demonstrated that both the pFA7 and pFC9 cDNAs encode PKA RI␣. The pFC9 cDNA encoded the full-length protein of 381 amino acids, while the pFA7 cDNA encoded 190 residues in the N-terminal half (data not shown). Differences from the previously reported nucleotide sequence (13) were found at the third position in five codons, but this did not alter the deduced amino acid sequence (data not shown). Only the pFC9 RI␣ cDNA was used for further experiments.
Identification of RI␣ Binding Domains on FSC1-The region of the FSC1 protein responsible for RI␣ binding was identified by deletional mutagenesis in the yeast two-hybrid system. The amino acid numbering was based on the FSC1 sequence (26). A series of deletion mutants were constructed in the pAS2-1 plasmid to allow expression of truncated FSC1 polypeptides in yeast cells. These plasmids were used to retransform primary transformants expressing the full-length RI␣ protein. Transformants containing plasmids both for deletion mutants of Fsc1 and full-length RI␣ were selected on SD/Trp Ϫ Leu Ϫ and examined for ␤-galactosidase activity by the colony lift assay. The first series examined were sequential deletions from the N terminus of FSC1. The ␤-galactosidase activity was lost after the deletions reached residue 343 (343-849), while a longer construct (335-849) gave ␤-galactosidase activity (Fig. 1b). This strongly suggests that the peptide sequence within the 335-342 residue region is required for RI␣ binding. We next examined another series of deletion mutations that lacked domain A (see below) in the N-terminal region of FSC1. The polypeptide that included residues 237-334 did not result in ␤-galactosidase activity, while activity was seen with the polypeptide that included residues 237-344 (Fig. 1b). These results confirm that RI␣ binding is associated with residues 335-344. We term this region domain B (Fig. 1c).
The second series of mutants involved plasmids with deletions producing truncations from the C terminus of FSC1. It was found that ␤-galactosidase activity occurred until the deletion reached residue 218 (125-218) (Fig. 1b). Taken together with other results showing that ␤-galactosidase activity was present with polypeptide 125-247 and absent with polypeptide 237-334, this indicated that an RI␣ binding domain lies between residue 219 and 236 in the FSC1 polypeptide (Fig. 1b). Visconti et al. (28) recently mapped an RII tethering domain of AKAP82 to a 57-amino acid region that includes a 14-amino acid sequence similar to an RII tethering domain in other AKAPs. This region of AKAP82 corresponds to residues 219 -232 of FSC1, and our results strongly suggest that this region is also capable of binding RI␣. We term this region domain A (Fig. 1c).
Transactivation of the His3 gene was also examined for all of the deletion mutants (Fig. 1, a and c). Transformants with plasmids encoding deletion mutants of FSC1 and full-length RI␣ were selected on SD/Trp Ϫ Leu Ϫ medium and then transferred onto SD/Trp Ϫ Leu Ϫ His Ϫ medium to test for growth. Representative results (Fig. 1a) show that mutants with deletions up to domain A or B grew on medium lacking histidine (335-849, 237-344, and 125-247), while further deletions resulted in the loss of growth activity (343-849, 237-334, and 125-218). These results were consistent with the ␤-galactosid-ase activity assays (Fig. 1, b and c).
Yeast homogenates were examined for the presence of fusion proteins to determine if failure to transactivate LacZ and His3 was due to failure of the vectors to express. However, Western blotting with an antibody to the GAL4 DNA binding domain demonstrated that the truncated FSC1 proteins were expressed in yeast in which transactivation did not occur (Fig. 2). Fusion protein was detected in homogenates of transformants in which neither LacZ nor His3 genes were expressed (343-849, 359 -849, 397-849, 125-218, 237-334, and 237-325), although some fusion proteins (125-218, 237-334, and 237-325) migrated with a slightly larger (ϳ1 kDa) apparent mass than expected. Since sequencing analysis identified no mutations that would change the translation of FSC1 polypeptides (data not shown), the discrepancy in the apparent mass of some polypeptides may be due to differences in SDS binding by FIG. 1. Deletion mutagenesis assays identify RI␣ binding domains on FSC1. Yeast cells containing plasmids expressing RI␣ were transformed with plasmids containing deletion mutants of FSC1. Tranformants were cultured on selection medium and tested for transactivation of the His3 gene on medium lacking histidine (a). To test for transactivation of the LacZ gene by the colony lift assay (b), transformants were cultured on SD/Trp Ϫ Leu Ϫ medium for 2 days and then transferred to a membrane to assay for ␤-galactosidase activity. In a, transformants received the plasmid encoding the full-length FSC1 protein, containing amino acids 1-849 (1); truncated protein containing amino acids 335-849 (2), 343-849 (3), 237-344 (4), 237-334 (5), 125-247 (6), or 125-218 (7); or plasmid encoding no FSC1 sequence (8). In c, the results are summarized, and an independent assay is shown for transactivation of the His3 gene. different amino acid sequences. Fig. 1c summarizes the results of the transactivation of His3 and LacZ by different mutants. These results strongly suggest that RI␣ binding domains occur at residues 219 -232 (domain A) and residues 335-344 (domain B) of FSC1 and that each domain independently can bind RI␣.
Binding Specificity of Domains A and B-Although the results of deletional mutagenesis studies in the yeast two-hybrid system provide strong evidence that RI␣ can bind to domains A and B of FSC1, additional studies were performed to confirm this association. These studies also examined more closely the specificity of domains A and B because of the report that the region of AKAP82 corresponding to domain A of FSC1 could bind RII subunits in the ligand overlay assay (28).
An in vitro binding assay was used to determine if RI or RII subunits present in testis extracts associate with FSC1-GST fusion proteins. Different Fsc1 constructs containing sequences encoding domain A (A), domain B (B), both domains A and B (AϩB), or the intervening region (N) were cloned into the pGEX plasmid vector, and FSC1-GST fusion proteins were expressed in E. coli (Fig. 3a). The fusion proteins were immobilized on glutathione-Sepharose resin and used to select PKA regulatory subunits from testis extracts containing 50 g (Fig. 3, b-d) or 500 g (Fig. 3, e-g) of protein. An aliquot of the proteins eluted from the resin was separated by SDS-PAGE and stained with Coomassie Brilliant Blue to verify that equivalent amounts of truncated FSC1-GST fusion protein were present in each sample (Fig. 3, d and g). Two other aliquots were analyzed by Western blotting using antibodies to RI␣ (Fig. 3, b and e) or RII␣ (Fig. 3, c and f). The positive control lane (lane T) contained 10 g of testis extract protein, while the other lanes contained the components in 50 g of the testis extract protein that bound to immobilized FSC1-GST.
Under these conditions, the RI␣ subunit was found to bind selectively to domain B (lane B) but not to the intervening region (lane N) or to domain A (lane A) (Fig. 3b). Since the intensity of the signal observed in lane B was comparable with that in lane T, domain B appeared to bind approximately 20% of the RI␣ present in the aliquot of testis extract. Binding of the RI␤ subunit to either domain was not detected in initial exper-iments (data not shown). However, binding of RI␣ to domain A was seen when 10 times the amount of testis extract (500 g of protein) was used (Fig. 3e). The binding of RI␣ to domain A appeared to be at least 10-fold lower than that for domain B, based on the intensity of corresponding bands in Fig. 3, b and e. It was particularly interesting that RII␣ binding to domain B was not detectable (Fig. 3c), even when 500 g of testis extract was used (Fig. 3f). RII␣ was found to bind to domain A, although the amount was relatively low compared with RI␣ binding to domain B (Fig. 3f). This result was consistent with the observation that RII subunits can bind to the N-terminal region of AKAP82 (28).
There was less binding of RI␣ and RII␣ to the fusion protein containing domains A and B (lane AϩB) than to the fusion proteins containing either domain B (Fig. 3b) or domain A (Fig.  3f). This low binding in the presence of both domains may be due to conformational interference between domains A and B for binding of PKA regulatory subunits.
Putative Amphipathic Helix of Domain B Associates with RI␣-Deletional mutagenesis assays in the yeast two-hybrid system demonstrated that a 10-amino acid stretch of domain B is sufficient for RI␣ binding (Fig. 1). In addition, a 14-amino acid stretch of domain A accommodates binding by either RI␣ or RII␣ (Figs. 1 and 3) (28). The amino acid sequences of these two domains are aligned with those of PKA tethering domains in Fig. 4a. Domain A contains 5 hydrophobic amino acids with long aliphatic side chains (leucines 224 and 227, isoleucine 229, valines 221 and 228), which frequently appear in RII tethering domains of AKAPs. These have been shown to be essential for the high affinity binding of RII␤ by AKAP75 (14). In contrast, domain B has only one of these residues (valine 339). Nevertheless, amino acid residues of domain B exhibit a striking segregation of hydrophobic and hydrophilic side chains, as is seen in the RII tethering domains of AKAPs (Fig. 4a). A helical wheel projection analysis indicates that domain B is likely to form an amphipathic helix (Fig. 4b). This secondary structure is commonly found in RII tethering domains of AKAPs and may be necessary for the association of an RII dimer. Site-directed mutagenesis on domain B was applied to determine if an amphipathic helix is an essential secondary structure for RI␣ binding (Fig. 5). These results suggest that domain B has secondary structural features in common with other PKA subunit tethering domains but also has distinct features that confer its specificity for RI␣ binding.

DISCUSSION
Yeast two-hybrid and in vitro binding assays were used to demonstrate that RI␣ and RII␣ subunits of PKA bind to FSC1, the major structural protein of the sperm fibrous sheath. Two subunit-binding sites, termed domains A and B, were identified on FSC1. The binding of RII subunits to domain A was consistent with previous studies. Ligand overlay assays had indicated that RII subunits bind to AKAPs (34), including AKAP82, the N terminus truncated form of FSC1 (27). Domain A corresponds to the region of AKAP82 reported to bind RII (28), and the results of the present study support and extend that finding. However, the binding of RI␣ to domains A and B was unexpected because it is generally assumed that only RII subunits are tethered to AKAPs. Previous studies with ligand overlay assays did not detect RI binding to AKAPs, and most RI subunits are reported to be in the cytosol and most RII subunits are reported to be in the particulate fraction of cell homogenates (35). However, D-AKAP1 and D-AKAP2 bind RI and RII subunits in yeast two-hybrid and in vitro binding assays (20,21). This suggests that the ligand overlay assay fails to detect RI subunit binding to AKAPs. This may be because PKA subunit binding domains are not renatured sufficiently after SDS-PAGE and transfer to nitrocellulose membranes to allow RI binding.
Another unanticipated finding was that both RI␣ and RII␣ bind to domain A of FSC1. However, this is consistent with competition assays indicating that RI and RII subunit binding domains on D-AKAP1 overlap partially or fully (20). The predicted binding domain of D-AKAP1 and experimentally determined binding domain A of FSC1 also have features that appear in the prototypic PKA subunit binding domains of AKAPs (Fig. 4). These include amino acids with long aliphatic side chains and putative amphipathic helical secondary structures (20,28,36). In contrast to domain A, domain B binds only RI␣. Site-directed mutagenesis strongly suggests that domain B also requires an amphipathic helix for binding of RI␣. In addition, domain B has a lower content of amino acids with long aliphatic side chains than other PKA subunit binding domains (14). Interestingly, valine 339 in domain B is not important for RI␣ binding (Fig. 5), but alanine 336 and 340 are essential for high affinity binding of RI␣. 2 (14), human AKAP79 (23), Ht31 (22), and MAP2 (22). Hydrophobic amino acids are boxed. b, the helical wheel projection of FSC1 domain B sequence is drawn as a ␣-helix with 3.6 amino acid residues/turn. Hydrophobic amino acids are indicated by outlined letters.

FIG. 5. Disruption of RI␣ binding activity in FSC1 domain B.
Point mutations were introduced into the Fsc1 cDNA to disturb the putative amphipathic helix in domain B, and then RI␣ binding activity was determined by in vitro binding assays, as in Fig. 3. The proteins that bound truncated FSC1 were analyzed by Western blotting using antiserum to RI␣ (a). Eluted proteins were visualized by Coomassie staining (b). Lane 1 received testis extract not subjected to the binding assay. tures required for RI␣-specific binding that are distinct from those required for RII␣ binding.
Domain B binds only RI␣, and in vitro binding assays suggest that RI␣ binds more strongly to domain B than to domain A. Conversely, low amounts of RII␣ in testis extract were pulled down by the domain A fusion protein. Although these are semiquantitative assays and results may be affected by differences in titers of antibodies to RI␣ and RII␣ and the conformation of binding domains in vitro, they strongly suggest that RI␣ preferentially binds to domain B and may have a significant role in tethering PKA to the fibrous sheath.
A general implication of these findings is that additional levels of PKA and AKAP binding specificity appear to exist beyond those generally recognized. Domain A of FSC1 can bind two types of PKA regulatory subunits, while domain B can bind only one. Conversely, the RI␣ subunit can bind to two AKAP domains in FSC1, while the RII␣ can bind to only one. Further studies will be needed to determine if such associations occur for other AKAP binding sites and PKA regulatory subunits or if these are specific to FSC1. In either case, it will be of interest to learn what features of the FSC1 subunit binding domains are responsible for dual RI and RII binding and for RI-specific binding characteristics.
Although PKA type I holoenzymes and free RI subunits usually are reported to be in the cytosol, there is also evidence for RI compartmentalization in cells. RI is bound tightly to the plasma membrane of human erythrocytes (37), accumulates in lymphocyte caps (38), associates with the sarcolemma of cardiac myocytes (39), and localizes to neuromuscular junctions (40). While RI was reported to be present throughout mouse sperm (28), the association of PKA type I with the fibrous sheath and the outer dense fibers of mammalian sperm was indicated to be detergent-resistant (29). Along with the evidence that RI subunits bind to D-AKAP1 and D-AKAP-2 (20,21), these studies suggest that the binding of RI␣ to FSC1 is not a unique event.
Gene knockout mice were generated to examine the roles of PKA regulatory subunits in vivo (41)(42)(43). In RII␣ knockout mice, PKA remained anchored to AKAPs, and the PKA-dependent potentiation of L-type Ca 2ϩ channels in muscle was unchanged because of compensation by RI␣. This occurred although RI␣ bound to AKAPs at 500-fold lower affinity than RII␣ (41). In RII␤ knockout mice, a compensatory increase of RI␣ occurred in brown adipose tissue, but the PKA type I holoenzyme was more easily inactivated than in wild-type mice, resulting in a lean phenotype (42). RI␤ mutant mice had reduced long term potentiation in the mossy fiber pathway of the hippocampus, despite a compensatory increase in RI␣ (43). However, no abnormalities were observed in sperm function, and the males were fertile for all three knockouts. RI␣ subunits may have been able to compensate for the loss of other regulatory subunits that associate with the fibrous sheath in two ways: by their ability to replace RII␣ binding to domain A and by their specific binding to domain B. If dual specific AKAP sites for RI␣ and RII␣ and monospecific sites for RI␣ exist in other sites and in other cell types, they are also likely to be involved in the compensation for loss of regulatory subunits in knockout mice.