Talin 2 is a large and complex gene encoding multiple transcripts and protein isoforms

Mouse Tln2 spans 414 kb and contains multiple 5′ noncoding exons

Initial studies on human and mouse talin 1 and talin 2 showed that, although they share the same genomic structure, talin 2 is a much larger gene, owing to the larger size of the talin 2 introns [12,13]. Analysis of mouse expressed sequence tags (ESTs) and cDNAs covering the 5′-end of mouse Tln2 now reveals an additional eight 5′-exons spanning 236 kb (Fig. 1A and Table S2). These exons do not encode any ORF in-frame with the rest of the Tln2 coding sequence, and to reflect the absence of coding potential, we numbered them exon −7 (most 5′) to exon 0 with respect to the first known coding exon (exon 1). The two most 5′ exons are embedded in a 1.45 kb CpG island (Fig. 1A). To confirm the presence of transcripts containing both the first coding exon (exon 1) and the most 5′ exon (exon −7), we used RT-PCR on mRNAs isolated from 13 tissues. Sequencing of the 248 bp amplicon detected in all tissues (Fig. 1B) revealed that it contains exon −7, exon −5, exon −2 and exon 1; this is identical to the combination found in EST BQ964581.1 (Fig. 1A). Other alternatively spliced transcripts were expressed at lower levels in some tissues, e.g. brain (Figs 1B and S1). These results: (a) show that previously uncharacterized alternatively spliced 5′-exons are present in Tln2 transcripts; (b) suggest that they originate from a new ubiquitous promoter lying within a CpG island; and (c) demonstrate that Tln2 is much larger (∼ 414 kb) than previously thought.

In order to establish that Tln2 does not extend further in the 5′ direction, we generated RACE cDNA libraries using total RNA from mouse brain and heart. Two rounds of amplification with nested primers in exon −7 revealed the presence of a diffuse 150 bp fragment compatible with a transcription start site in or near the CpG island (data not shown). Cloning and sequencing of the PCR products confirmed that the 5′-end of the Tln2 transcripts lies within the CpG island, although the transcription start site varied slightly within the same tissue and also between tissues (Fig. 1C). This may reflect the absence of a TATA-box, and suggests that the Tln2 promoter associated with the CpG island is a housekeeping promoter that relies on the positioning of various combinations of transcription factors (TFs) for transcription initiation [28,29]. Analysis of the 5′-end of human TLN2 confirms that, as in mouse, transcription is driven from a CpG island situated ∼ 200 kb away from the first coding exon, and that the 5′-UTR exons, some of which are conserved between mouse and human, show complex patterns of alternative splicing in different cell types (Table 1 and Fig. S2).

In silico identification of a conserved minimal promoter associated with the CpG island

As the CpG island is conserved in other mammals (data not shown), we sought to establish whether a conserved promoter is present in the vicinity. We used two different in silico methods (frameworker and dialign tf; Genomatix) to analyse a 2.2 kb segment upstream of exon −7 in six different mammalian species (Fig. S3 and Experimental procedures). Both approaches revealed the presence of the same highly conserved cluster of TF-binding sites just upstream of the transcriptional start sites identified by 5′-RACE in the mouse CpG island. The arrangement and position of these sites is strictly conserved in six different species, and they are predicted to recruit ubiquitously expressed TFs, such as nuclear respiratory factor 1, elongation factor 2 (E2F), cAMP-responsive element binding protein and E-twenty-six (ETS), which are associated with a large number of ubiquitously expressed housekeeping genes [30]. These results strongly suggest that a core promoter is embedded within the CpG island at the 5′-end of the mammalian gene, and that its use underlies the ubiquitous expression of Tln2 transcripts (Fig. 1B). Furthermore, examination of chromatin modification data supplied with the latest human genome build (ensembl v50; Fig. S4) strongly supports the idea of a promoter associated with the CpG island in humans. Indeed, it indicates that a 298 bp region overlapping our minimal promoter bears chromatin features characteristics of a promoter, i.e. presence of histone H2AZ, H3K9 monomethylation, H3K4 dimethylation and trimethylation, and DNAse I hypersensitivity (Fig. S4). It also suggests that, at least in humans, no other sequences upstream of the first coding exon exhibit the same features, and therefore that a single promoter may control the expression of TLN2.

Transcription from the CpG island accounts for the majority of Tln2 expression in adult tissues

To get a better insight into the functional significance of the CpG island and its associated promoter, we analysed mice carrying a gene trap insertion in the intron downstream of exon −3 (Fig. 1A and Experimental procedures), which should interrupt transcription from the CpG region. Mice homozygous for the gene trap (gt/gt) were viable, and we therefore analysed talin 2 levels in various tissues (brain, heart, kidney and spleen) by western blotting (Fig. 1D). Strikingly, the gene trap insertion led to a dramatic reduction in talin 2 levels in all four tissues (Fig. 1D). This indicates that the region upstream of exon −3 contains one or several major site(s) of transcription initiation. It also strongly suggests that the promoter identified in this study is a major Tln2 promoter. However, the gene trap failed to completely ablate talin 2 expression. This could be due to the intrinsic properties of gene traps, as skipping of gene trap cassettes and alternative splicing around the transgene have previously been described [31–33]. Alternatively, it is possible that other Tln2 promoters exist downstream of the gene trap insertion. Nevertheless, the extent of the reduction in talin 2 level observed strongly suggests that transcription upstream of the gene trap, most likely from the promoter in the CpG island, is responsible for the majority of talin 2 production in adult tissues.

A truncated talin 2 is expressed in elongating spermatids of the mature testis

Northern blots of testis mRNA reveal a 4.8 kb Tln2 transcript that is too small to encode the full-length protein [12,24]. A testis EST containing two new 5′-exons (25b and 25c; Fig. 2 and Table S2) and predicted to encode a talin 2 lacking the talin head and half of the rod domain has also been described [24]. To analyse the expression pattern of exons 25b and 25c, we performed RT-PCR on a panel of tissues with primers located in exons 25b and 26 (Fig. 2A). This revealed high-level expression of exon 25b specifically in testis and very low levels in thymus. Surprisingly, the main RT-PCR amplicon in testis is too short to contain exons 25b, 25c and 26, although a minor larger product was also observed (Fig. 2A, 40 cycles). Both products were sequenced, and this revealed that skipping of exon 25c occurs in a proportion of transcripts.

Exon 25b is predicted to be noncoding, whereas exon 25c harbours 84 bp of noncoding sequence and an ATG at position 85–87 in-frame with the full-length Tln2 ORF [24]. Use of this ATG would add a unique 30 amino acid sequence to the N-terminus of this truncated talin 2 isoform, which otherwise would be identical to residues 1121–2542 of the C-terminal half of full-length talin 2. However, the above RT-PCR data suggest that a second protein product may result from the skipping of exon 25c. In this case, the transcript would contain a unique 5′-UTR, and the next in-frame ATG would be located in exon 26. Consequently, the protein would not contain the N-terminal 30 amino acid extension (Fig. 2C), and would be equivalent to residues 1153–2541 of full-length talin 2. To investigate whether exon 25b encodes the 5′-end of the short testis transcript, we first carried out quantitative RT-PCR. This showed that Tln2 transcripts possessing sequences downstream of exons 25b and 25c are 20 times more abundant in testis than transcripts initiating upstream of exon 25 (Fig. S5A). This strongly suggests that the vast majority of Tln2 transcripts in testis initiate downstream of exon 25. 5′-RACE experiments confirmed that exon 25b represents a genuine Tln2 transcription start site in testis. The position of the start site varies slightly in the three clones that we sequenced, and no TATA-box was detected upstream of exon 25b (Fig. 2B). From the above results, we conclude that these short Tln2 transcripts are expressed at high levels specifically in testis, and that they contain a unique 5′-end composed of either exon 25b or exons 25b and 25c.

To establish whether the ∼ 4.8 kb Tln2 transcripts in testis gives rise to truncated talin 2 protein(s), we used a gene trap mouse line [24] in which the transgene is inserted between exons 28 and 29 (Fig. 2A). Because the gene trap cassette is downstream of exons 25b and 25c, it is predicted to interrupt the transcription of both long (ubiquitous) and short (testis-specific) Tln2 transcripts. If full-length talin 2 (270 kDa) and the shorter isoforms (153 kDa) coexist in testis, they should both be detected in wild-type animals with an antibody raised against the C-terminal region of talin 2, whereas both proteins should be absent in homozygous gene trap mice. Furthermore, the gene trap cassette should lead to the expression of two talin 2–βGeo fusion proteins that can be detected by an antibody against β-galactosidase (Fig. 2D). As predicted, the levels of talin 2 proteins were dramatically reduced in tissues from homozygous (gt/gt) mice, and they were replaced by the predicted talin 2–βGeo fusions (Fig. 2D). These results establish for the first time the existence of a truncated talin 2 in testis, which we have named Ts-talin 2 (testis-talin 2). We have not detected Ts-talin 2 in any other tissues.

As Ts-talin 2 lacks the integrin-binding site found in the N-terminal FERM domain of full-length talin, we sought to establish whether it could localize to cell–ECM junctions. Analysis of NIH-3T3 cells transiently expressing green fluorescent protein (GFP)–Ts-talin 2 confirmed that the protein was efficiently expressed (Fig. S6A and Doc. S1), and that it localized to focal adhesions in a similar manner to GFP–full-length talin 2 (Fig. S6B and Doc. S1), mCherry–full-length talin 2 or endogenous full-length talin 2 [25]. Expression of GFP–Ts-talin 2 in talin 1-deficient dJ26.28 cells [15,34] produced essentially similar results, showing that its localization is not dependent upon talin 1 (data not shown).

In order to further understand the role of Ts-talin 2 in testis physiology, we performed western blot analysis of testis extracts from prepubertal and pubertal mice using N-terminal and C-terminal antibodies to detect the full-length and short testis isoforms. The results show that Ts-talin 2 is not expressed in prepubertal testis (2 weeks) but is upregulated at the onset of puberty (4–6 weeks in the C57Bl/6 mouse strain). Concomitantly, the full-length protein is downregulated (Fig. 3A). In addition, our gene trap mouse line, which reports expression of both full-length talin 2 and Ts-talin 2 (Fig. 2A), allowed us to follow the expression pattern of the two proteins during this developmental switch. In heterozygotes, the β-galactosidase fusion protein was expressed in myoid cells and Leydig cells at the periphery of the seminiferous tubules, and this pattern of expression was identical at 2 and 12 weeks of age. However, strong punctate β-galactosidase staining appeared in the seminiferous epithelium after puberty. This was detected in only a subset of immature tubules, and was replaced by more diffuse staining as the tubule progressed towards the spermiation point (Fig. 3B). Dissection of portions of seminiferous tubules, staged by transillumination [35], revealed that these dot-like structures are only present between the end of stage VIII (end of spermiation), when the transition between round and elongating spermatids is initiated, and the end of stage I, when elongating spermatids start forming bundles [36] (data not shown). Squashes of tubules from stages VIII–XII and XII–I [35] were used to analyse further the expression of the β-galactosidase–talin 2 fusion protein. This showed that the dot-like structures are localized in the cytoplasm of early elongating spermatids (Fig. 3C). A proportion of dots were also observed in enucleated vesicles (data not shown), reminiscent of the residual bodies lost during the process of spermatid elongation [37]. Together with our western blotting data, these results strongly suggest that the expression of full-length talin 2 is restricted to myoid and Leydig cells, whereas Ts-talin 2 is characteristic of the early stages of spermatid elongation.

A second, shorter Tln2 transcript is highly expressed in kidney

Northern blotting detects a highly expressed 3.9 kb Tln2 transcript in kidney [12,24]. Examination of mouse ESTs revealed a kidney EST (AI317453) with a novel 5′-exon (numbered 34b) just downstream of exon 34 (Fig. 4A and Table S2). The position of exon 34b suggests that it may represent the first (or one of the first) exon(s) of the 3.9 kb Tln2 kidney transcript (Fig. 4A). RT-PCR showed that exon 34b-containing transcripts are expressed at very high levels in kidney and at lower levels in brain, heart, testis, lung, muscle, and skin (Fig. 4B). Moreover, quantitative real-time RT-PCR showed that exon 34b transcripts are 11 times more abundant in kidney than those initiating upstream (Fig. S5B). Conversely, in small intestine and brain, exon 34b transcripts are expressed at much lower levels than the full-length transcripts (Fig. S5B,C). Together, these results indicate that exon 34b is part of a different Tln2 transcript that is mainly expressed in kidney, where it is by far the most abundant species.

To confirm that exon 34b is actually the first exon of the 3.9 kb kidney Tln2 transcript, we performed 5′-RACE. The data show that transcripts containing exon 34b do not carry any further 5′-exons (Fig. 4B). The longest 5′-RACE product that we have identified does not contain any ATGs in exon 34b that are in-frame with the rest of talin 2, so it is very likely that exon 34b is a noncoding exon. The first ATG in-frame with the Tln2 ORF is found in exon 35, which is common to both long and short transcripts (Fig. 4A,C). The corresponding truncated talin 2 isoform is predicted to be a 90 kDa protein spanning residues 1608–2543 (Figs 4C and 6). This isoform should not be detected by antibodies raised against the N-terminal region of talin 2, whereas an antibody against the C-terminal region should detect both full-length talin 2 and an additional 90 kDa protein, mainly in kidney. As expected, western blotting with the N-terminal antibody showed only full-length talin 2 (data not shown). In contrast, two different C-terminal antibodies detected an additional protein of ∼ 90 kDa that was strongly expressed in kidney (Fig. 4C and data not shown).

To investigate whether a similar talin 2 isoform is expressed in humans, we first probed a northern blot using a conserved mouse Tln2 3′-probe. As in mouse, we observed a 3.9 kb transcript that was specifically and highly expressed in human kidney (Fig. 4D). In addition, a search for sequences homologous to exon 34b in the EST database revealed the existence of a human kidney EST (BI517694) containing at its 5′-end the conserved counterpart of exon 34b. By screening a human kidney cDNA library with a set of nested primers in the conserved exon 34b and a second primer in the vector (Sp6), we were able to amplify a 450 bp cDNA product, which was cloned and sequenced. Alignment between the new human sequence, the mouse 5′-RACE product and the mouse and human ESTs containing exon 34b revealed that exon 34b is highly conserved between the two species, and does not have any coding potential (Fig. 4D). We conclude that the shorter 3.9 kb transcript and exon 34b are conserved in humans, and, as in mice, are highly expressed in kidney.

Alternative splicing of Tln2 coding exons

The existence of alternatively spliced transcripts has been described in vertebrate talin 2 and in the unique ancestral talin gene, which is closely related to talin 2 [12,13]. In order to clarify the extent of alternative splicing in the mammalian gene, we conducted a systematic review of alternatively spliced ESTs in the mouse. We also sought to determine the expression profile of these variants as well as their relationship to the various short transcripts described above. We identified three genuine Tln2 splice variants in mouse ESTs (Fig. 5A). The first alternative splicing skips exon 43 (kidney AF467081). The second one, which contains an extension of exon 54 (exon 54b; Table S2) and inserts 15 amino acids between residues 2458 and 2459 of the protein, is found in four distinct ESTs (brain AB093231, CO044539 and BC079896; 8 days post coitum whole body AK017597), and is equivalent to a splice variant previously described in humans [13]. The third variant EST from a mammary tumour library combines skipping of exons 43 and 44 with the inclusion of exon 54b (BC033288).

RT-PCR showed that exon 43 skipping is common in most adult tissues (Fig. 5B; primers i + j and i + k), although the ratio between transcripts including or excluding exon 43 varies greatly between tissues. Only heart, lung and skin express the exon 43-minus variant at about the same level as the wild-type transcript, whereas it is found at much lower levels in other tissues, and was totally absent from brain. In contrast, skipping of exons 43 and 44, as found in a mammary tumour EST, is a rare event in the tissues tested (Fig. 5B; primers i + k). Next, we searched for transcripts that included exon 54b. Again, this splice variant was detected in all the tissues tested, but at variable levels (Fig. 5C; primers m + n). Strikingly, transcripts containing exon 54b are the major variant in brain (Fig. 5C; primers l + n).

To search for mRNAs lacking exons 43 and 44 but including exon 54b, we first amplified cDNAs containing exon 54b (primers o + q), and then assessed the splicing of either exon 43 alone or exons 43 and 44 using nested PCR (primers i + k; Fig. 5D). We did not detect any products carrying deletions of both exons 43 and 44 (Fig. 5D), which confirmed the very low levels previously observed in the total cDNA population (see above and Fig. 5B). However, transcripts skipping exon 43 and containing exon 54b are a major component of skin, whereas in kidney and muscle they occur at a much lower level. We also failed to detect exon 43 skipping in either the testis or kidney short transcripts (Fig. S7), which suggests that skipping of exon 43 is a specific feature of long transcripts. In contrast, inclusion of exon 54b was found in both the short and long transcripts expressed in brain and kidney, but not in heart or testis (Fig. S7). We conclude that the short Ts-Tln2 transcripts do not skip exon 43 or include exon 54b, whereas in kidney, exon 54b is mainly added to the short transcript, and is much less abundant in full-length Tln2 transcripts. Together, these results provide new evidence of the complexity of Tln2 transcription and suggest that the gene encodes at least three different transcripts, which are subject to alternative splicing in a tissue-specific and/or isoform-specific manner.

EST and cDNA analysis

We used the National Institute for Allergy Mouse Gene Index (version 5) and the Mouse Genome Assembly through the ensembl portal (build 37; NCBI reference sequence version 36) to carry out our analysis. The National Institute for Allergy Mouse Gene Index gene index provided a graphical view of the Tln2 structure, a list of all the known and predicted Tln2 exons, and individual splicing events along with a list of supporting ESTs and cDNAs. This information was analysed manually and compared with the data present in ensembl. Together, this provided a framework with which to decipher Tln2 structure, the existence of short transcripts, and the presence of alternative splicing events. It was also used to design the various primers relevant to the characterization of the gene structure and described in Table S1.

Promoter prediction

blast was used to identify the sequences equivalent to the most 5′ mouse Tln2 exon (exon −7) in human, chimpanzee, mouse, rat, dog and cow genomic sequences. A 2.2 kb sequence including the 5′-end of the CpG island was extracted from each individual genomic sequence via ensembl, and analysed using the genomatix promoter analysis suite (http://www.genomatix.de). Briefly, we used frameworker and the most stringent settings to identify the most complex model (framework) of TF-binding sites that occurred in the same order and in a similar region and that were common to all six sequences. In addition, dialign tf [54] was used to identify TF-binding site matches within the multiple sequence alignments using default parameters. TF-binding site matches were identified by comparison with a matrix of descriptions with matinspector. To minimize artefacts due to the small size of TF-binding sites, we increased the threshold of similarity to the matrix to 0.99, and identified only TF-binding site matches that were common to all six sequences.

RNA extraction

Tissues from wild-type C57/Bl6 mice were dissected and stored in RNAlater (Invitrogen, Carlsbad, CA, USA), and total RNA was extracted from a 20 mg piece of tissue using the RNAeasy Miniprep kit (Invitrogen), both according to the manufacturer’s instructions. RNA concentrations were quantified at 260/280 nm and treated accordingly with RQ1 DNase I (Promega, Madison, WI, USA; 1 unit·μg⁻¹ RNA). After addition of 600 μL of RLT buffer (Invitrogen; RNAeasy miniprep kit), samples were purified through a second RNA column (Invitrogen; RNAeasy miniprep kit) to eliminate the enzyme. Yields were quantified at 260/280 nm prior to storage at −80 °C.

Human RNA, tissue northern blot and kidney cDNA library

Human foreskin fibroblast and adult heart total RNAs were a kind gift from K. Clark (University of Leicester, UK); heart RNA was from Stratagene (Santa Clara, CA, USA; #540011). One microgram of RNA was reversed transcribed using a standard protocol (see below). A human Multiple Tissue Northern Blot was purchased from BD Biosciences Clontech (Palo Alto, CA, USA) and hybridized as previously described [12] with a 598 bp mouse cDNA probe spanning exons 52–55. In order to map the 5′-end of the short Tln2 transcript present in human kidney, plasmid DNA from a human kidney cDNA library (kindly provided by M. J. A. Tanner, University of Bristol) was screened by PCR. We took advantage of the directional cloning of the cDNAs in the vector, and primer SP6-2 in the plasmid sequence flanking the 5′-end of the cDNA inserts was used in combination with primer hT2-3R, designed in reverse orientation in the TLN2 sequence near the putative 5′-end of the short transcript (Table S1). A 450 bp amplicon was cloned and sequenced.

Reverse transcription

DNAse-treated total RNA (1 μg) was reversed transcribed with Superscript III (Invitrogen) and random primers (Promega) as recommended by Invitrogen, with 250 ng of random primers in a total volume of 20 μL. In addition, non-reversed transcribed controls were generated in the same conditions by replacing the Superscript III enzyme with water. One to two microlitres of these reactions was used in PCRs with the primers described in Table S1. All primers used to detect cDNA were designed to span at least one exon–intron boundary, to avoid detection of residual genomic DNA.

PCR

The amplification of small amplicons (< 2 kb) was performed in standard conditions at an annealing temperature of 55 °C, using 600 nm of the primers described in Table S1 and PCR Reddy Mix (ThermoScientific, Walthman, MA, USA) in a volume of 15 μL. When a second round of nested PCR was required, 1 μL of the first round was used in a 15 μL reaction as described above. Long-range PCR (> 2 kb) was performed using the Expand High Fidelity PCR kit (Roche, Indianapolis, IN, USA), according to the manufacturer’s recommendations, with 300 nm of each primer. One microlitre of these reactions was used subsequently as template in a second nested PCR with ReddyMix as described above.

Real-time quantitative RT-PCR

Expression of Tln2 variant mRNAs was quantified using real-time PCR in a BioRad MiniOpticon Thermocycler with 2x iQ SyberGreen Mastermix (BioRad, Hercules, CA, USA) and the specific primer sets described in Table S1 and in the figure legends. Amplification conditions, annealing temperature and primer pair efficiency (95–105%), were optimized for each primer pair prior to the experiment, to validate the comparison between different primer pairs. Primers were used at a final concentration of 300 nm in a 25 μL reaction volume. One microlitre of random-primed cDNA from relevant tissues was added to each reaction, and each cDNA sample was amplified in triplicate to take into account variations in amplification. Gapdh was used as an internal normalization control. Relative abundances were calculated using the 2^−ΔΔCt method [55].

Gene trap ES cells

The gene trap ES cell line RRI4343 (plasmid insertion) was obtained from BayGenomics (Stamford, CA, USA), and the FHCRC-S1-D6 gene trap (lentiviral insertion) from P. Soriano’s laboratory (Seattle, WA, USA). These ES cell lines were grown in DMEM with 4.5 g·L⁻¹ glucose plus glutamine and 10% fetal bovine serum on 0.1% gelatin (RRI4343) or mitomycin C-treated MEF feeder layers (FHCRC-S1-D6) in the presence of 1000 U·mL⁻¹ leukaemia inhibitory factor, as described previously [19]. The clones were expanded and stained with X-gal to check for β-galactosidase expression, and genomic DNA was extracted and amplified by PCR to verify the presence of the correct gene trap insertion. The insertion point of the gene trap construct in cell line RRI434 was determined by Chen & Lo [24]. We used this information to design our own set of genotyping primers, 5′GT4/GT4Vec (gene trap allele, 450 bp) and 5′GT4/3′GT4 (wild-type allele, 700 bp; for sequences, see Table S1), which were used in a single multiplex PCR. The genomic region upstream of the insertion point of the lentivirus in clone S1-D6 was cloned and sequenced by P. Soriano’s team, and the sequence information provided was used to design primer pairs SGT5′For/SGT5′Rev (gene trap allele, 684 bp) and SGT5′For/SGT3′Rev (wild-type allele, 536 bp; see sequences in Table S1), which were used in separate standard PCR reactions.

Generation of transgenic mice and genotyping

ES cell clones RRI434 and FHRC-S1-D6 were injected into C57BL/6 blastocysts. Germline transmission from male chimeras was assessed through the presence of Agouti pups in the offspring. Animals carrying the gene trap insertions were identified by PCR using genomic DNA extracted from ear snips, using the HotSHOT technique as previously described [56]. Heterozygous gene trap mice (gt/+) were either outbred onto a C57/Bl6 background to establish a colony or intercrossed in order to generate homozygotes (gt/gt). Combinations of primers (see gene trap ES cells and Table S1), detecting both the presence and the absence of the gene trap at the insertion point, allowed discrimination between heterozygous and homozygous mice.

Protein lysates from tissues and western blotting

Tissues were dissected from genotyped animals, flash frozen in liquid nitrogen, and stored at −80 °C. The frozen tissues were lysed and homogenized with a Polytron tissue homogenizer as described in [57], in the presence of a cocktail of protease inhibitors (Calbiochem, San Diego, CA, USA; Set III; 1 : 100 dilution) plus a calpain II inhibitor (E64d; Sigma, St Louis, MO, USA; 1 μg·mL⁻¹). Protein concentrations were determined in 5 μL aliquots in triplicate using the Bradford assay (Coomassie Plus Kit; ThermoScientific) with BSA standards. Aliquots of tissue lysates containing 40 μg of protein were denatured at 100 °C for 2 min in the presence of 5 mm Tris/HCl (pH 6.8), 4% glycerol, 1.6% SDS, 2% bromophenol blue, and 1%β-mercaptoethanol. The samples were subjected to SDS/PAGE (6% gels), and the proteins were electroblotted onto Hybond P membranes using wet transfer in 2.5 mm Tris base and 9.6 mm glycine. Excess protein binding sites were blocked by incubation in 5% dried milk in NaCl/Tris (pH 7.4). The long isoform of talin 2 was detected using a mouse monoclonal antibody (68E7) raised in this laboratory against an N-terminal recombinant mouse talin 2 peptide (amino acids 482–911) and diluted 1 : 200 in blocking buffer. Polypeptides containing the C-terminal region of talin 2 were detected with an in-house mouse monoclonal antibody (53-121A1, epitope 2479-2494; 1 : 50 dilution) or a polyclonal antibody raised in rabbits against residues 2477–2491 of human talin 2 (1 : 3500 dilution). Vinculin was detected using the mouse monoclonal antibody F9 as previously described, with a 1 : 400 dilution [13]. β-Galactosidase was detected with a rabbit polyclonal primary antibody (Abcam, Cambridge, UK; ab4761, 1 : 1000 dilution). Bound primary antibodies were detected by adding anti-mouse IgGs (GE Healthcare UK Ltd, Chalfont St Giles, UK; 1 : 2500 dilution) or anti-rabbit IgGs (Sigma; A6154, 1 : 3500) coupled to horseradish peroxidase. Horseradish peroxidase activity was detected using an ECL Kit (SuperSignal West Pico; ThermoScientific), and autoradiographic films were exposed for varying amounts of time. Semiquantitation of the relative abundance of talin 2 across samples was performed using Adobe photoshop. Films were scanned and the images were inverted. Rectangles of constant surface area were drawn around the talin 2 and vinculin bands, and the mean luminosity of each rectangle was determined in Adobe photoshop. Levels of talin 2 were obtained by normalizing the talin 2 to the vinculin signals and by comparing the values between samples.

Cryosections and X-gal staining

Testes were dissected and frozen in optimal cutting temperature (OCT) medium (R. A. Lamb, London, UK) and 20 μm frozen sections were fixed in 2% paraformaldehyde and 0.1% glutaraldehyde for 5 min, and stained for β-galactosidase activity overnight at 37 °C in 1 × NaCl/P_i with 2.5 mm X-gal, 5 mm potassium ferrocyanide, 5 mm potassium ferricyanide, 2 mm MgCl₂, 2% sodium deoxycholate and 0.4% NP-40. The slides were washed three times in 1 × NaCl/P_i, and nuclei were counterstained with haematoxylin. After being rinsed in cold water, the slides were dehydrated in 70, 90 and 100% ethanol, postfixed twice in xylene, and mounted in DPX synthetic resin.

Seminiferous tubule dissection and X-gal staining

The testis was decapsulated, and the tubules were staged and dissected as previously described [35]. The relevant portions of tubules were squashed on a slide using a coverslip, and flash frozen in liquid nitrogen before being fixed in 2% paraformaldehyde for 5 min. X-gal staining was performed as described for cryosections, and nuclei were counterstained with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI; Invitrogen) for 5 min at room temperature. Slides were subsequently washed three times in 1 × NaCl/P_i before mounting. Differential interference contrast and DAPI-stained images were acquired on a Leica DM5000B microscope fitted with a 40× objective and a Leica DFC420C imaging system.