Harp (harmonin-interacting, ankyrin repeat-containing protein), a novel protein that interacts with harmonin in epithelial tissues


  • Communicated by: Shoichiro Tsukita

  • Present addresses: INSERM U145, IFR 50, Faculté de Médecine, 06107 Nice Cedex 2, France; bTakasago Research Institute, Kaneka Corporation, 1–8, Miyamae-Machi, Takasago-Cho, Takasago-Chi, Hyogo 676-8688, Japan


Mutations in the triple PDZ domain-containing protein harmonin have been identified as the cause of Usher deafness syndrome type 1C. Independently, we identified harmonin in a screen for genes expressed in pancreatic β cells. Using a yeast two-hybrid assay, we show that the first PDZ domain of harmonin interacts with a novel protein, designated harp for harmonin-interacting, ankyrin repeat-containing protein. This interaction was confirmed in an over-expression system and in mammalian cells, and shown to be mediated by the three C-terminal amino acids of harp. Harp is expressed in many of the same epithelia as harmonin and co-localization of native harp and harmonin was demonstrated by confocal microscopy in pancreatic duct epithelium and in a pancreatic β-cell line. Harp, predicted molecular mass 48 kDa, has a domain structure which includes three ankyrin repeats and a sterile alpha motif. Human harp maps to chromosome 16, and its mouse homologue to chromosome 7. Sequences with similarity to harp include the sans gene, mutations of which are responsible for deafness in the Jackson shaker 2 (js) mutant mouse and in human Usher syndrome type 1G. The functional domain structures of harp and harmonin, their interaction under native conditions and their co-localization suggest they constitute a scaffolding complex to facilitate signal transduction in epithelia.


The organization of signalling complexes within specific intracellular compartments is necessary for the regulation of cellular responses to a range of stimuli. Scaffolding, anchoring and adaptor proteins play essential roles to integrate signalling pathways. These proteins contain modular structural domains, such as ankyrin repeats, Src-homology 2 (SH2) and Src-homology 3 (SH3) motifs and PDZ domains, responsible for mediating protein–protein interactions [reviewed in (Pawson & Nash 2000)]. PDZ domains, originally identified as sequence repeats in PSD-95, Dlg and ZO-1 [reviewed in (Sheng & Sala 2001)], are protein modules of approximately 90 amino acids that mediate associations with consensus sequences usually located at the C-termini of their partner proteins.

Defects in the PDZ domain-containing protein harmonin are responsible for Usher syndrome type 1C (USH1C), an autosomal recessive disorder characterized by congenital sensorineural deafness, vestibular dysfunction and blindness due to progressive retinitis pigmentosa (Verpy et al. 2000). In USH1C-affected families with additional features, including severe hyperinsulinism, enteropathy and renal tubular dysfunction, a 122-kb homozygous deletion was identified in the chromosomal region containing harmonin (Bitner-Glindzicz et al. 2000). Harmonin exists as several alternatively spliced isoforms generated from 28 exons (Verpy et al. 2000). Two of these, designated type a, contain three PDZ domains and one coiled-coil domain located between PDZ domains two and three. Two type b isoforms contain an additional coiled-coil domain as well as a proline, glutamic acid, serine and threonine (PEST) motif, while type c isoforms contain only one coiled-coil domain and are truncated after the second PDZ domain. AIE-75, a type a isoform, was identified as an autoantigen in X-linked autoimmune enteropathy (Kobayashi et al. 1999). In addition, multiple isoforms of harmonin were identified as targets of autoantibodies present in some patients with colon and renal cancer (Scanlan et al. 1998). The harmonin b isoforms are expressed specifically in the inner ear, whereas other isoforms have broader tissue distribution patterns (Scanlan et al. 1999; Verpy et al. 2000).

We identified harmonin in pancreatic β cells by PCR-based representational difference analysis of the mouse pancreatic insulin-producing βTC3 and glucagon-producing αTC1 cell lines (Niwa et al. 1997). Harmonin is expressed at the apical surface of epithelial cells specialized in fluid secretion (Kobayashi et al. 1999; Scanlan et al. 1999). The presence of PDZ domains in harmonin implies that it may coordinate signalling protein complexes at the plasma membrane. To gain further insight into the physiological role of harmonin in epithelia, we used the yeast two-hybrid system to identify proteins that interact with its individual PDZ domains. We describe the isolation and characterization of a novel ankyrin repeat-containing protein we term harp (harmonin-interacting, ankyrin repeat-containing protein), a member of a family of structurally related proteins conserved across species, whose three C-terminal amino acids interact specifically with the first PDZ domain of harmonin and which is co-localized with harmonin in epithelia.


Yeast two-hybrid screening for harmonin-interacting proteins

After identifying harmonin in mouse pancreatic β cells (Niwa et al. 1997), we sought to identify proteins that interacted with harmonin. A mouse kidney cDNA library was screened using the PDZ domains of harmonin as bait. Preliminary studies indicated that constructs encoding fusion proteins of the GAL4-activation domain and PDZ1, PDZ2 or PDZ1-2 domains were appropriate, whereas fusion proteins that included domain PDZ3 of harmonin auto-activated the Lac-Z gene. For this reason, only constructs containing PDZ1 and/or PDZ2 were used as baits in the yeast two-hybrid screens. Only clones encoding proteins that interacted with a fusion protein containing the first two PDZ domains of harmonin, as well as one containing either of the single PDZ domains (Table 1), were further investigated. One of these clones encoded d-dopachrome tautomerase (DDT), a protein of unknown function with a putative immuno-inflammatory role inferred from its homology with macrophage migration inhibitory factor (Esumi et al. 1998). The C-terminal three amino acids (TFL) of DDT constitute a putative class I PDZ domain-interacting motif, characterized by the concensus sequence S/T-X-Φ where X is any amino acid and Φ is hydrophobic (Sheng & Sala 2001). However, we were unable to demonstrate an interaction of DDT with harmonin in vitro (data not shown) and did not investigate it further. Another clone encoded β-catenin, which bound the harmonin PDZ1/PDZ2 fusion protein. Further investigation revealed that β-catenin did not interact with either PDZ1 or PDZ2 individually in the yeast two-hybrid screen. While the binding of target molecules to PDZ domains may depend on conformation (Grootjans et al. 2000), the interaction that we detected between β-catenin and harmonin in the yeast-two-hybrid screen was weak and we could not replicate it using full-length over-expressed proteins either in vitro or in mammalian cells.

Table 1. Harmonin PDZ domain–mediated interactions


Harmonin-interacting proteins detected by yeast-two hybrid analysis
Clone number*IdentityInteractionC-terminal sequence
HI-1 d-dopachrome tautomerase++–MT F L
HI-2β-catenin+–DT D L
HI-3harp++–DT S L


Previously identified harmonin–PDZ interactions
ProteinPDZ1PDZ2PDZ3C-terminal sequenceReference
  • *

    Clones isolated from yeast two-hybrid screens of an adult mouse kidney cDNA library fused to the GAL4 activation domain, using PDZ domain 1 and PDZ domain 2 of harmonin as bait. Putative interacting clones were further assessed for binding specificity by determining their ability to bind a construct encoding the amino terminal half of harmonin, including both PDZ1 and PDZ2 domains.

  • The C-terminal sequence of myosinVIIa does not correspond to a classical PDZ domain binding motif consensus.

sans+–DT E L(Weil et al. 2003)
MCC2+–DT F L(Ishikawa et al. 2001)
Harmonin+–LT F F(Siemens et al. 2002)
CDH23++–I T E L(Boeda et al. 2002; Siemens et al. 2002)
Myosin VIIa+–R S G K (human)–R S G R (mouse)(Boeda et al. 2002)

Analysis of the Blast database using sequence from clone HI-3, which interacted with PDZ1 and PDZ1/2 of harmonin, revealed it to be a novel cDNA, with similarities to some ESTs as well as to a region of human chromosome 16. Clone HI-3 contained a single open-reading frame followed by a 3′-untranslated region of 143 nucleotides encoding a putative protein of 423 amino acids (Fig. 1A). Analysis of the amino acid sequence using the Smart program (http://smart.embl-heidelberg.de/) revealed that the protein comprised three N-terminal ankyrin repeats and a C-terminal sterile alpha motif (SAM) (Fig. 1B). Therefore, we designated this protein as harp (harmonin-interacting ankyrin repeat-containing protein) and deposited its nucleotide sequence in the GenBank database under Accession number AF524030. Using the deduced protein sequence to perform a Blast search, significant homology was revealed to a mouse protein (Accession numbers XP111242 and XP137901) and a human protein (Accession numbers AK091243 and BAC03619), which we termed harp-related, as well as to a Drosophila melanoganster protein (CG13320, Accession number NP610829) and an Anopheles gambiae protein (Accession number EAA04767). Alignment of these proteins (Fig. 2) revealed that members of this family have a conserved protein domain structure (three N-terminal ankyrin repeats and a C-terminal SAM domain) with the highest degree of similarity in their common motifs, and all conserve the class I PDZ-binding motif at their C-terminal end. During preparation of this manuscript, the mouse and human harp-related proteins were identified as sans (scaffold protein containing ankyrin repeats and SAM domain), a protein mutated in the Jackson shaker (js) mouse model of deafness (Kikkawa et al. 2003) and in humans with Usher syndrome type 1G (Weil et al. 2003). Overall, there is 40% amino acid identity between harp and sans proteins, the highest conservation being in the ankyrin repeat region (70%) and the SAM domain (55%) and the lowest in the central intervening region (24%).

Figure 1.

Sequence and domain structure of harp. (A) Nucleotide and deduced amino acid sequence of mouse harp (mmHARP). Numbering of amino acids and nucleotides is indicated on the left and right margins, respectively. The three ankyrin repeats (amino acids 31–60, 64–93, 97–126) are underlined. The sterile alpha motif is shaded. Amino acid differences between human (hsHARP) and mmHARP are shown. (B) Schematic of the putative domain structure of harp.

Figure 2.

Harp-related proteins. Comparison of harp family members. The amino acid sequences of mouse and human harp, mouse and human sans, Drosophila and Anopheles harp proteins were aligned by the ClustalV method using a PAM250 matrix. Residue identities and similarities are highlighted in blue and yellow, respectively. The ankyrin repeats are indicated with a solid underline and the SAM domain with a dashed underline.

Chromosomal localization of harp family genes

Radiation hybrid panel mapping identified the best-fit location for harp to mouse chromosome 7 between D7Mit133 (proximal) and D7Mit240 (distal). The human homologue physically mapped to BAC clone CIT987SK-334D11 (Accession number AF001550) and was located on the short arm of chromosome 16 between D16S3045 and D16S3201 (http://genome.ucsc.edu/). This mapping of apparently homologous genes was confirmed by the syntenic homology between mouse chromosome 7 and human chromosome 16 as a result of the ordering of harp in relation to two other known genes (ZP2 and CRYM) on the human BAC clone and within the current linkage map of mouse chromosome 7 (http://www.informatics.jax.org/). The mapping coordinates for the mouse gene have been submitted to the database at The Jackson Laboratory website (http://www.jax.org/resources/documents/cmdata/rhmap/). Human and mouse harp genes are encoded upon two exons and show 81% identity at the amino acid level. Comparison of the mouse and human sans cDNAs to the UCSC Genome Bioinformatics databases revealed that, analogous to harp, mouse and human sans genes are also encoded by two exons with similar splicing sites to those of harp genes. The mouse sans gene has been mapped to chromosome 11 near the js locus, which is responsible for deafness in the Jackson shaker 2 (js) mouse mutant strain (Kikkawa et al. 2003). The human sans gene was mapped to chromosome 17q25 within the critical region of USH1G (Weil et al. 2003).

Tissue expression of harp, harmonin and sans RNAs

Northern blot analysis of mRNAs from various mouse tissues (Fig. 3A) revealed a 2.5-kb harp transcript in liver and kidney. A 3.8-kb transcript was detected in liver and kidney and a less abundant 4.0-kb transcript also in liver, kidney and weakly in testis, indicating possible alternatively spliced forms of harp in different tissues. To investigate the expression of harp in relation to harmonin and sans, RNA samples from several mouse tissues were subjected to RT-PCR using gene-specific oligonucleotide primers (Fig. 3B). Under the PCR conditions used, harp transcripts were detected in liver, kidney, small intestine, colon, pancreas (weak signal) and purified pancreatic islets. Harmonin transcripts were detected in brain, heart, kidney, small intestine, colon, testis (weak signal), pancreas and islets. Sans transcripts were only detected in brain and testis. Both harmonin and harp were expressed in the βTC3 cell line.

Figure 3.

Tissue expression of harp, harmonin and sans mRNAs. (A) Northern blot analysis. A mouse multiple tissue Northern blot (Clontech) was hybridized with a harp-specific cDNA probe as described in Experimental procedures. (B) RT-PCR. Total RNA extracted from different tissues was (+) or was not (–) reverse transcribed and subjected to PCR using gene specific oligonucleotides for harp, harmonin and sans, β-actin was amplified as an internal control. Amplified products were run on 1.5% agarose gels. NT, not tested.

The interaction between harp and harmonin

The interaction between harp and the PDZ1 domain of harmonin was confirmed in vitro using a pull-down assay with lysates of transfected COS7 cells expressing either FLAG-tagged harp (Fl-harp) or a FLAG-tagged truncated form of harp lacking the three C-terminal amino acids (Fl-harpΔ3) (Fig. 4). In this assay, GST-PDZ1, -PD22 or -PD23 fusion proteins acted as baits for Fl-harp or Fl-harpΔ3. As shown (Fig. 4), harp interacts specifically with the PDZ1 domain of harmonin, and this interaction is mediated by the three C-terminal amino acids (TSL) of harp because the truncated form (Fl-harpΔ3) fails to bind GST-PDZ1 of harmonin.

Figure 4.

The interaction between the C-terminus of harp and PDZ1 of harmonin COS7 cells were transfected with constructs encoding FLAG-tagged harp or FLAG-tagged harp minus the three C-terminal amino acids (Flag-harpΔ3). Lysates prepared as described in Experimental procedures were incubated with GST-fusion proteins, including GST alone, GST-PDZ1, GST-PDZ2 or GST-PDZ3, immobilized on glutathione-Sepharose. Following stringent washing, bound proteins were analysed by SDS–PAGE and Western blotting using the anti-FLAG M2 monoclonal antibody (upper panel, right section). Total cellular lysates from transfected cells were included in the blot to determine levels of expression of harp fusion proteins (upper panel, left section). Approximately equivalent amounts of GST-fusion proteins were included in the assay (bottom panel).

To investigate the interaction of harp and harmonin in mammalian cells, we used cells either stably (RIN-GFP-harp) or transiently (βTC3-GFP-harp) transfected with a GFP-harp fusion construct. A 75-kDa protein corresponding to GFP-harp was immunoblotted in both sets of transfectants with the anti-harp 1 serum (Fig. 5A). Complexes containing endogenous harmonin were precipitated from transfectant cell lysates with anti-harmonin serum and co-precipitated GFP-harp was immunoblotted with either anti-GFP serum (Fig. 5B, top panel, lanes 4 and 8) or anti-harp 1 serum (Fig. 5C, lane 4). Membranes were also probed with anti-harmonin serum. A 64-kDa protein corresponding to endogenous harmonin was detected at similar levels in all total cell lysates (Fig. 5B, bottom panel, lanes 1, 3, 5 and 7) and in anti-harmonin immunoprecipitates (Fig. 5B, bottom panel, lanes 2, 4, 6 and 8; Fig. 6C, lane 6).

Figure 5.

The interaction between harp and harmonin in mammalian cells. (A) Total cell lysates (50 µg protein/lane) of RIN cells stably transfected and selected for GFP-harp expression (RIN-harp) and βTC3 cells transiently transfected with GFP-harp (βTC3-harp) were analysed by SDS–PAGE and Western blotting. GFP-harp protein was detected with anti-harp 1 serum. Lysates from non-transfected RIN and βTC3 cells were included as controls. (B) Triton X-100 cell lysates (lanes 1, 3, 5 and 7) and anti-harmonin immunoprecipitates (lanes 2, 4, 6 and 8) from non-transfected and GFP-harp transfected RIN and βTC3 cell lines were analysed by SDS–PAGE and Western blotting. GFP-harp was detected with an anti-GFP serum. (C) Lysates of stable GFP-harp RIN cell transfectants were incubated with pre-immune rabbit serum at 2 µg/mL (lanes 1 and 3) or with anti-harmonin serum at 2 µg/mL (lanes 2 and 4) and proteins bound to protein G-Sepharose were separated by SDS–PAGE followed by Western blotting with a pre-immune rabbit serum (left panel) or an anti-harp 1 rabbit serum (centre panel). The membrane blotted with pre-immune rabbit serum was stripped and re-probed with anti-harmonin serum (right panel). (D) Triton X-100 cell lysates αTC1 and βTC3 cells were precipitated with pre-immune rabbit IgG (Pre-immune), anti-harmonin IgG (Harmonin), anti-harp 1 IgG (Harp1), anti-harp 2 IgG (Harp2) and anti-sans IgG (Sans) chemically cross-linked to protein G-Sepharose, separated by SDS–PAGE and transferred on to a nitrocellulose membrane. The membrane was then probed with anti-harmonin serum (left panel). The βTC3 section of the blot was stripped and re-probed with anti-harp 1 serum (bottom panel). As a control, βTC3 cell lysates (right panel) were incubated with GST, GST-PDZ1 or GST-C-terminal harp fusion proteins, immobilized on glutathione-Sepharose, separated by SDS–PAGE and Western blotted. The blot was then probed with anti-harmonin serum.

Figure 6.

Tissue expression of harp and harmonin proteins. (A) Schematic of the domain structures of the three major subclasses of harmonin isoforms. (B) Western blotting of harmonin and harp in mouse tissues. Lysates from various mouse tissues and from the βTC3 cell line (50 µg of protein/lane) were resolved by SDS–PAGE, transferred to a nitrocellulose membrane and blotted with antiserum raised against PDZ domains 2 and 3 of harmonin. Major bands corresponding to harmonin types a, b and c are indicated with arrows (top panel). Selected mouse tissues were similarly analysed by Western blotting with anti-harp 1 serum (bottom panel).

We next determined that harp and harmonin interact under native conditions. Lysates of βTC3 and αTC1 cells were incubated with anti-harmonin, anti-harp 1 and 2, anti-sans or control pre-immune IgG chemically crosslinked to protein G-Sepharose, resolved by sodium dodecyl sulphate–polyacrylamide gel electrophoresis (SDS–PAGE) and Western blotted with antisera to harmonin or harp. Harmonin was absent in αTC1 cells but a 64-kDa band corresponding to harmonin was precipitated with anti-harmonin IgG and co-precipitated with anti-harp 2 IgG from βTC3 cells (Fig. 5D, left panel, arrow), confirming the native interaction between the two proteins. Stripping and probing the βTC3 section of the membrane with anti-harp 1 serum revealed a 48-kDa band corresponding to harp that precipitated with anti-harp 2 IgG and co-precipitated with anti-harmonin IgG; however, very low levels of harp were precipitated with anti-harp 1 IgG (Fig. 5D, bottom panel). Taken together with the Western blot results in Fig. 5A and 5C, these findings indicate that anti-harp 1 IgG precipitates native harp inefficiently, and fails to co-precipitate harmonin, because it recognizes the C-terminus of harp where harmonin is bound. This would imply that, under these conditions, harmonin is in excess. As a final control, harmonin in the same βTC3 lysates was shown to bind a GST-harp C-terminus fusion protein, but failed to interact with GST alone or GST-PDZ1 (Fig. 5D, right panel, black arrow).

Tissue expression of harp and harmonin proteins

Harmonin is known to have multiple isoforms, classified as types a, b and c (Verpy et al. 2000) (Fig. 6A). Using an antiserum that recognizes both PDZ2 and PDZ3 domains of harmonin, which should detect all harmonin isoforms so far described, the expression of harmonin was analysed in mouse tissues by Western blot (Fig. 6B, top panel). Harmonin isoforms were predominantly detected in the small intestine, brain, kidney, pancreas and isolated pancreatic islets. An isoform of approximately 100 kDa corresponding to harmonin type b was detected in islets, small intestine, pancreas, kidney and brain. Type c isoform bands of 45, 54 and 59 kDa were observed in small intestine, kidney and less prominently in brain. Two isoforms of approximately 75 and 64 kDa corresponding to harmonin type a were detected in islets, small intestine and kidney. The 64-kDa form is consistent with the predicted size of the type a isoform, and corresponds to the cDNA we found differentially expressed in the pancreatic β cell line, βTC3 (Niwa et al. 1997).

The expression of harp in selected mouse tissues was analysed by Western blot using the harp 1 antiserum (Fig. 6B, bottom panel). Harp was detected as a 48-kDa species in epithelial tissues: kidney, small intestine, pancreas and liver, and was absent in heart, spleen and brain.

Immunohistochemistry of adult mouse pancreas and e17.5 embryonic mouse tissues was performed with anti-harp and anti-harmonin sera. In the adult pancreas, the strongest staining for both harp (Fig. 7A) and harmonin (Fig. 7J) was observed in the pancreatic duct epithelium, with much weaker staining of the islets of Langerhans. Both harp and harmonin showed comparable expression patterns in epithelia of the developing lung (Fig. 7B,K), kidney (Fig. 7C,L), salivary glands (Fig. 7D,M) and the cochlea (Fig. 7E,F,N,O); staining was specifically blocked with harp C-terminal peptide (Fig. 7G,H,I) or GST-PDZ2/3 fusion protein (Fig. 7P,Q,R), respectively. The staining patterns were similar but not identical, both proteins being expressed most prominently at the cell membrane. Harmonin had a wider intracellular distribution, probably because all of its isoforms are detected by the antiserum. Harmonin isoforms can have distinct intracellular localization; for example, harmonin b, the longest harmonin isoform, directly binds actin filaments, inducing the formation of large actin bundles in the stereocilia of inner hair cells, whereas harmonin a and harmonin c do not (Boeda et al. 2002). Co-localization of harp and harmonin in pancreatic ducts and in the βTC3 cell line was demonstrated by confocal microscopy (Fig. 8). Harp and harmonin are co-localized in small intralobular (Fig. 8, top row) and larger interlobular (Fig. 8, second row) ducts. At higher magnification, the two proteins clearly show the same spatial distribution in ductal epithelial cells (Fig. 8, third row) and in βTC3 cells (Fig. 8, bottom row).

Figure 7.

Immunohistology of harp and harmonin in selected mouse epithelial tissues. Immunohistochemistry with anti-harp 1 (panels A–I) and anti-harmonin (panels J–R) sera was performed on Histochoice-fixed sections of mouse tissues, as described in Experimental procedures. Shown are sections of adult mouse pancreas (A and J) and of e17.5 embryonic epithelia including lung (B and K), kidney (C and L), salivary gland (D and M), cochlea (E and N) and cochlear duct (F and O). To demonstrate specificity, sections of salivary gland (G and P), cochlea (H and Q) and cochlear duct (I and R) were stained with anti-harp 1 serum pre-incubated with harp C-terminal peptide (G, H and I), or with anti-harmonin serum pre-incubated with GST-PDZ2/3 fusion protein (P, Q and R).

Figure 8.

Co-localization of harp and harmonin. Adult mouse pancreas (top three rows) or βTC3 cells (bottom row) were double stained with anti-harmonin (green) and anti-harp (red) rabbit IgG as described in Experimental procedures. Merged images (yellow) are indicated.


We identified harmonin initially as a subtractive product in a representational difference analysis screen for genes expressed in the βTC3 insulin-secreting β-cell line, compared with the αTC1 glucagon-secreting α-cell line (Niwa et al. 1997). A role for harmonin in auditory signal transduction has been inferred by the finding that mutations in the harmonin gene are associated with congenital deafness in Usher syndrome type 1C (Bitner-Glindzicz et al. 2000; Verpy et al. 2000). In the sensory cells of the inner ear, harmonin was reported to interact with myosin VIIa (Boeda et al. 2002), cadherin-23 (Boeda et al. 2002; Siemens et al. 2002), the harp-related protein sans (Kikkawa et al. 2003; Weil et al. 2003) and with itself (Siemens et al. 2002), creating a scaffolding complex required for steocilia hair bundle function. It is worth noting that most of the harmonin PDZ domain-mediated protein interactions identified to date occur with PDZ1. This suggests that, when expressed within the same intracellular compartments, strict mechanisms must control the interactions of these proteins. While the domain-structure of harmonin indicates that it acts as a scaffolding protein, the identity of other harmonin-interacting proteins and their exact mechanism of action in auditory as well as other epithelia remain unclear.

Using the yeast two-hybrid assay, we identified three candidate harmonin-interacting proteins: DDT, β-catenin and a novel protein. Each of these proteins harboured a PDZ-consensus binding motif at its C-terminus. The interaction between DDT and harmonin could not be replicated in vitro and was not pursued. β-Catenin has previously been shown to interact with PDZ domain-containing proteins through its C-terminal amino acids, TDL (Perego et al. 2000). Its potential role as a harmonin-binding protein is supported by its expression in the sensory epithelium of the inner ear (Warchol 2002) and in other epithelia such as pancreatic islet cells (Collares-Buzato et al. 2001) where harmonin is also expressed. Furthermore, β-catenin co-precipitates with vezatin and E-cadherin (Kussel-Andermann et al. 2000), which are implicated in the maintenance of stereociliary stability (see below). Although we could not reproduce the interaction in vitro, β-catenin remains a putative harmonin-interacting protein.

The novel protein, designated harp, also identified in the yeast two-hybrid system, was demonstrated to interact specifically with the first PDZ domain of harmonin. Structurally, harp has three N-terminal ankyrin repeats and a C-terminal sterile alpha motif (SAM) domain, and its three C-terminal amino acids (TSL) conform to a class I PDZ-binding motif. Its domain structure is identical to that of the harp-related protein sans (Kikkawa et al. 2003; Weil et al. 2003). Mutations in sans are responsible for Usher Syndrome 1G (Weil et al. 2003). Sans was shown by co-transfection to interact with PDZ1 of harmonin. We suggest that harp, together with sans and the two protein variants found in insects, belongs to a larger protein family that appears to have the common property of interacting with harmonin.

The structural domains in harp and related proteins mediate protein–protein interactions. Ankyrin repeats are 33-residue units observed in over 400 proteins including transmembrane proteins, cytoskeletal proteins, trancription factors and cyclin inhibitors. Their function as protein-binding modules has been well documented in several systems (Sedgwick & Smerdon 1999; Bennett & Chen 2001), particularly in linking integral membrane proteins, including ion channels and cell adhesion molecules, with the spectrin-based cytoskeleton. SAM domains are 60- to 70-amino acid sequences originally identified in yeast Ste4 and Byr2 (Ponting 1995) and subsequently found in a diverse range of proteins (Schultz et al. 1997). They have been shown to promote protein–protein interactions by forming homo- or hetero-oligomers with other SAM domains (Thanos et al. 1999). SAM domains may also complex signalling molecules, as in the case of the ELK receptor, in which ligand binding-induced phosphorylation of a conserved tyrosine within the SAM domain enables association with the SH2 domain of Grb10 (Stein et al. 1996). A number of proteins have a domain composition similar to that of harp. These include the SHANK family of somatostatin receptor-interacting proteins, which contain six N-terminal ankyrin repeats and a C-terminal SAM (Lim et al. 1999; Zitzer et al. 1999) and are believed to multimerize as homo- or heteromers to cross-link proteins at postsynaptic sites. The GASZ protein contains four ankyrin repeats and a C-terminal SAM domain and is believed to be an important cytoplasmic signal transducer mediating protein–protein interactions during germ cell maturation and preimplantation embryogenesis (Yan et al. 2002).

Harmonin is expressed in embryonic and adult epithelia (Kobayashi et al. 1999; Scanlan et al. 1999) (Góñez, L. J., Naselli, G., Braakhuis, A and Harrison, L. C., manuscript in preparation). We found that harp was co-expressed with harmonin in adult pancreatic ductal epithelium and in the embryonic epithelia of the lung, kidney, salivary glands and cochlea. However, RT-PCR and Northern blotting revealed that harp transcripts are expressed in the liver where harmonin transcripts or protein were not detectable. The domain structure of harp suggests that it is involved in binding different proteins, but these could vary depending on the tissue and/or cellular context. This may apply to harmonin as well, considering the number of isoforms expressed in different tissues. Indeed, while the present study was in progress, MCC2, a putative tumour suppressor molecule, was identified by yeast two-hybrid screening as a binding partner for the first PDZ domain of harmonin/AIE75 (Ishikawa et al. 2001). However, it remains to be confirmed that MCC2 and harmonin interact physiologically in mammalian cells, and the cellular localization of MCC2 has not been reported. Our RT-PCR results suggest that harp and sans have mutually exclusive tissue distributions and therefore may be tissue-specific partners of harmonin. Moreover, neither harp nor sans are expressed in the heart where MCC2 shows the highest expression (Ishikawa et al. 2001), providing further evidence of tissue specificity of harmonin-binding partners.

Of particular interest, in view of harmonin's role in auditory signal transduction, is the co-expression of both harp and harmonin in epithelium of the embryonic ear. Using the harp sequence that we had deposited in GenBank, Weil and co-workers (Weil et al. 2003) could not detect harp in the inner ear vestibule by RT-PCR. However, immunohistochemical staining presented here demonstrates that harp is expressed in the developing mouse cochlea. Several proteins involved in auditory signalling have been discovered by identifying mutations that lead to hearing impairment or vestibular dysfunction, and the role of some has been at least partly defined by identifying their binding partners. Thus, mutations in myosinVIIa, Cdh23, Pcdh15 and harmonin are responsible for different subtypes of Usher syndrome, and subsequent studies have shown that myosin VIIa, Cdh23 and Pcdh15 are indispensable for stereociliary stability [reviewed in (Gillespie & Walker 2001; Muller & Littlewood-Evans 2001)]. In mouse models, mutations of myosin VIIa, Cdh23 and sans have similar phenotypes with splayed stereocilia in hair cells of the inner ear, and each of these three proteins has been shown to bind harmonin through PDZ domain interactions. Notably, mutations in Pcdh15, which also has a C-terminal PDZ-binding motif (TSL), result in a very similar phenotype. It would therefore be of interest to determine if Pcdh15 is involved in the same protein complex. Cytoskeletal components such as actin, espin, fimbrin and other myosin isoforms appear to be important for cross-linking of stereocilia, while vezatin and protein kinase A have been identified by virtue of their myosin VIIa-binding properties. The harmonin b isoform also acts as an F-actin bundling protein (Boeda et al. 2002), linking the cadherins and the cytoskeleton. Although several intermolecular interactions have been identified, the picture remains incomplete and the protein network involving harmonin is likely to be more intricate. This is suggested by the identification of both harp and sans as harmonin-interacting proteins, which themselves are likely to bring new partners to the complex.

The functional domain structures of harp and harmonin, their demonstrated interaction under native conditions and their tissue co-localization support the hypothesis that they constitute a scaffolding complex to facilitate signal transduction in epithelia. Determining the nature of the proteins that interact with the PDZ domains of harmonin and with the protein binding modules of harp should help to elucidate signal transduction in epithelia.

Experimental procedures

Yeast two-hybrid screening

To identify mammalian proteins that interact with the PDZ domains of harmonin, an adult mouse kidney MATCHMAKER library (Clontech Laboratories Inc., Palo Alto, CA) was transformed into the Saccharomyces cerevisiae Y153 reporter strain bearing a fusion protein between the GAL4 DNA binding domain and either the first or second PDZ domain of harmonin. Positive cDNA clones that resulted in a His3+/LacZ+ phenotype were then screened for interactions against the other PDZ domain and also against a fusion protein including both PDZ domains 1 and 2. The bait constructs encoding the PDZ domains were made in a modified pGBT9 vector (Schneider et al. 1997) containing the PheS gene which allows positive selection of the prey vector upon transformation into bacteria by plating onto agar plates containing 10 mm DL-4-chlorophenylalanine (Sigma-Aldrich Pty. Ltd, Castle Hill, NSW, Australia). Otherwise, standard yeast transformation, two-hybrid-screening and His3 and LacZ reporter analyses were performed as previously described (Hoffman & Winston 1987; Fields & Song 1989).

cDNA cloning and sequencing

Following isolation of plasmids encoding putative harmonin-interacting proteins, sequencing was performed by the dideoxy-terminator method using a 320A sequenator (Applied Biosystems, Foster City, CA). The Blast algorithm was used to scan databases for protein and DNA homologies.

Chromosomal localization

The mouse chromosomal location of harp was determined using specific primers on the T3 Mouse Radiation Hybrid Panel (Research Genetics Inc., Huntsville, AL) (McCarthy et al. 1997). The primers chosen for mapping correspond to sequences within the 3′ untranslated region and the first intron of the harp gene (5′-TGAGGGTTTAGATGCACAGG-3′/5′-GATCAATGTGCTCAGGCAGA-3′ for intron 1; 5′-CTGCTGCTTTGCTCTGATGA-3′/5′-TGGTTCCAGGAGAACTCAGC-3′ for 3′ UTR). The 3′ untranslated region and intron sequences were determined by a Blast search using the harp cDNA sequence through the use of the Celera Discovery system and Celera's associated databases (Celera Genomics, Rockville, MD). These primers generate specific products under the PCR conditions used: 95 °C, 58 °C, 72 °C for 20, 20 and 30 s, respectively, for 40 cycles in the presence of 500 µm dNTPs, 0.2 units of Taq polymerase (Boehringer Mannheim, Castle Hill, NSW, Australia). The best map order was established for the gene by submitting PCR results to The Jackson Laboratory Mouse Radiation Hybrid Database (http://www.jax.org/resources/documents/cmdata/rhmap/). The human chromosome location of the homologous gene for harp was determined by a Blast search and identification of a human BAC clone. The BAC clone location was determined by accessing the database at http://genome.ucsc.edu/.

Expression constructs

Expression constructs were generated by subcloning full-length cDNAs isolated from the yeast two-hybrid screen into the pEGFP-C2 (Clontech) and the pCMV-Tag vectors (Stratagene, La Jolla, CA), resulting in green fluorescent protein (GFP)-tagged or FLAG-tagged proteins, respectively. Truncated constructs lacking three C-terminal amino acids were amplified by PCR using Pfu polymerase (Promega) with oligonucleotide primers that included restriction sites to allow cloning into the pEGFP-C2 vector in the appropriate frame (5′-CGGGATCCAAATGTCTAGGCGCTATCAC-3′/5′-CCGCTCGAGTTAGTCAACCAACTGCCCAGGCTG-3′ for harp). Glutathione-S-transferase (GST)-fusion proteins containing PDZ1, PDZ2 and PDZ3 of harmonin were constructed by PCR amplification using Pfu polymerase (with sense/anti-sense primers for PDZ1 5′-AAGGAATTCCGCTTGGACCGTC-3′/5′-CTCCTCTCGAGACTCTT-3′; PDZ2 5′-GAGGTGAATTCGGCTCAC-3′/5′-CAGCTCGAGGCCGGCTCC-3′; PDZ3 5′-CCAGAGCGGATCCCAGGGAAG-3′/5′-CAGCAACCTCGAGGTCATT-3′), purified products were digested with appropriate enzymes and ligated into pGEX-4T1 (Amersham-Pharmacia Biotech, Uppsala, Sweden).

Reverse transcription polymerase chain reaction (RT-PCR)

Mouse tissues were dissected from 8-week-old CBA mice, washed in ice-cold phosphate-buffered saline (PBS) and RNA extracted using RNAzol (Tel.Test, Inc., Friendswood, TX). Dnase I-treated RNA was reverse transcribed with 200 U of MMLV RT (Invitrogen, Life Technologies, Carlsbad, CA) in the presence of 0.5 µm random hexanucleotides (Bresatec, GeneWorks Pty. Ltd, Thebarton, SA, Australia) and 200 µm dNTPs. One-tenth volumes of the first strand synthesis reactions were amplified by PCR in PCR buffer (Perkin Elmer Inc., Shelton, CT) containing 200 µm dNTPs, 1 U Taq polymerase and 1 µm each of gene specific sense/anti-sense oligonucleotide primers (5′-GGCAACCTCGAGGCCCTAGAA-3′ and 5′-GAAGCCGTTCGCA CTCTTTCA-3′ for harp, 5′-CCAGAGCGATCGCAGGGAAG-3′ and 5′-AGCAACCTCGAGGTCTAT-3′ for harmonin, 5′-GGTCACCTGCACTGCCTGTCCTTCCTC-3′ and 5′-GCCCTGGCGCACAAACATCACATCA-3′ for sans and 5′-GTGGGCCGCCCTAGGCACCA-3′ and 5′-CTCTTTGATGTCACGCACGATTTC-3′ for β-actin). PCR reactions were performed for 35 cycles (95 °C/30 s; 58 °C/1 min; 72 °C/1 min) and amplified products were analysed on 1.5% agarose gels.


FLAG-tagged proteins were detected using an anti-FLAG M2 monoclonal antibody (Sigma-Aldrich). GFP fusion proteins were detected using an anti-GFP rabbit serum obtained from Dr Emmanuela Handman (The Walter and Eliza Hall Institute). Anti- harmonin rabbit serum was generated to purified protein encompassing the PDZ domains 2 and 3 of harmonin, obtained by cleavage of the GST-PDZ2/3 fusion protein with thrombin (Amersham-Pharmacia Biotech). Anti-harp 1 and anti-harp 2 rabbit sera were generated, respectively, against a peptide corresponding to the C-terminus of harp (415-PGQLVDTSL-423) and a peptide within the central intervening region between the third ankyrin repeat and the SAM domain (197-SKGIKDTFKIKSKKNK-212). Anti-sans rabbit serum was also generated against a peptide within the central intervening region between the third ankyrin repeat and the SAM domain (221-ERRKQGGEGTFK-232). Peptides were synthesized with an extra cysteine residue at the N-terminus and then coupled to diphtheria toxoid (DT) via a maleimidocapoyl-N-hydroxysuccinimide (MCS) linker. Free and coupled peptides were purchased from Mimotopes Pty. Ltd, Clayton, Vic., Australia. Harmonin PDZ2/3 protein or DT-coupled peptides (0.5 mg) were emulsified with Freund's complete adjuvant and injected at multiple subcutaneous sites into two rabbits each. Following two booster immunizations 6 and 8 weeks later, rabbits were bled and sera stored at −20 °C. Immunoglobulins (IgGs) were purified from antisera by protein G-Sepharose (Amersham-Pharmacia Biotech) affinity chromatography.

Cell culture and transfection

All culture media were from Life Technologies–Gibco (Rockville, MD). The RIN insulinoma cell line was maintained in mouse tonicity-RPMI, 10% foetal calf serum (FCS) and antibiotics. RIN cell transfections were performed by electroporation. Stable transfectants expressing GFP-harp were obtained by selection in culture medium containing 0.5 mg/mL geneticin (Life Technologies–Gibco). COS7 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) with 10% FCS and antibiotics. COS7 cells plated at a density of 1 × 105 per well were transfected within 24 h of plating with 1 µg of DNA using FuGENE (Roche Diagnostics Australia Pty. Ltd, Castle Hill, NSW, Australia), according to the manufacturer's instructions. βTC3 cells were maintained in high glucose (4.5 g/L) DMEM with 10% FCS and antibiotics. For transfection, βTC3 cells were seeded at 80% confluency in 6-well plates and transfected with 5 µg of DNA/well using Lipofectamine 2000 (Invitrogen).

Preparation of pancreatic islets and mouse tissue homogenates

Islets were isolated from 6- to 8-week-old C57BL/6 or CBA mice by intraductal collagenase digestion followed by bovine serum albumin gradient centrifugation as previously described (Lake et al. 1987; Thomas et al. 1998). Mouse tissues were dissected from 6- to 8-week-old CBA mice and washed once in ice-cold PBS. Purified islets and tissues were homogenized in PBS containing 1 mm phenylmethylsulphonyl fluoride and 1 mg/mL aprotinin by 20–30 strokes in a chilled Dounce homogenizer. Triton X-100 (1% v/v) was then added, incubated 30 min on ice and the lysates clarified by centrifugation at 12 000 g for 10 min at 4 °C. Protein concentration was determined using the Bio-Rad Protein Assay kit (Bio-Rad Laboratories, Hercules, CA).

GST fusion protein purification, pull-down assays and Western blot analysis

For GST fusion protein purification, bacterial expression constructs encoding GST, GST-PDZ1, GST-PDZ2 and GST-PDZ3 of harmonin were used to transform competent E. coli. GST fusion proteins were produced by induction of 500 mL log phase cultures with 1 mm isopropyl-β-d-thiogalactopyranoside, followed by sonication to lyse the bacteria. Following centrifugation, lysates were incubated with glutathione-Sepharose beads (Amersham-Pharmacia Biotech) according to the manufacturer's instructions. GST fusion proteins or GST alone (5 µg each) immobilized on glutathione-Sepharose beads were incubated with 500 µL of TritonX-100 lysates from COS7 transfectants expressing either FLAG fusion proteins encoding the indicated putative harmonin-interacting proteins or FLAG fusion proteins of mutants missing the three C-terminal amino acids which encompass the putative PDZ binding motif. After binding overnight at 4 °C, the beads were washed three times in lysis buffer, twice in lysis buffer containing 0.5 m NaCl and once in PBS. Bound proteins were separated by SDS–PAGE, transferred to nitrocellulose membranes (Protran BA 83, Schleicher and Schuell, Dassel, Germany) and immunoblotted with antibodies as described in Results and figure legends. Horseradish peroxidase (HRP)-conjugated sheep anti-rabbit and HRP-sheep anti-mouse antibodies (Silenus Laboratories, Hawthorn, Vic., Australia) were used at 1 : 5000 dilution for detection by chemiluminescence with the Lumi-Light Western blotting substrate (Roche).

Immunoprecipitation studies

For immunoprecipitation of FLAG fusion proteins, cells were lysed in Triton X-100 lysis buffer (50 mm HEPES, pH 7.5, 10% glycerol, 1% TritonX-100, 150 mm NaCl, 10 mm MgCl2, 1 mm EGTA) 48 h after transfection. Lysates from transiently transfected or non-transfected cells were pre-cleared by centrifugation at 12 000 g for 10 min at 4 °C and incubated with 10 µL of anti-FLAG M2 agarose affinity gel (Sigma-Aldrich) for 2 h at 4 °C. Proteins immobilized on agarose were then washed and separated by SDS–PAGE and detected by Western blotting.

For co-immunoprecipitations of native harp and harmonin, βTC3 and αTC1 cells were lysed in Triton X-100 lysis buffer supplemented with 1 mm sodium vanadate, 1 mm phenylmethylsulphonyl fluoride, 1 mg/mL aprotinin and 5 mm dithiothreitol. Lysates were pre-cleared and incubated with purified anti-harp, anti-sans or anti-harmonin IgG chemically crosslinked on to protein G-Sepharose. Immune complexes were washed three times with high salt buffer (Triton X-100 lysis buffer supplemented with 500 mm NaCl), followed by two washes in low salt buffer and one wash in PBS, before resolution by SDS–PAGE in 10–20% Tris-glycine Novex gels (Invitrogen) and Western blotting.

Northern blot analysis

A mouse multiple tissue Northern blot (Clontech) was probed with a 32P-labelled cDNA probe corresponding to full-length harp. Hybridization was performed for 2 h at 68 °C in ExpressHyb Hybridization Solution (Clontech). Filters were washed in 2× saline-sodium citrate (SSC) solution, 0.05% SDS for 30 min at room temperature followed by 0.1× SSC, 0.1% SDS for 30 min at 50 °C, and exposed to Hyperfilm MP (Amersham Pharmacia Biotech) for 24 h at −70 °C.


Mouse tissues were fixed for 2–4 h in Histochoice tissue fixative (Sigma-Aldrich), dehydrated and embedded in paraffin. Tissue sections (4 µm) were placed on uncoated glass slides, deparaffinized in xylene and re-hydrated in graded alcohols. Endogenous peroxidase was blocked by immersion in 0.03% hydrogen peroxide for 15 min. Purified rabbit IgG was diluted in PBS to 100 µg/mL, added to the slides and incubated for 60 min at room temperature in a humidified chamber, followed by incubation with HRP-sheep anti-rabbit secondary antibody (Silenus) diluted 1 : 50 in PBS for 30 min at room temperature. The sections were then visualized using 3,3′-diaminobenzidine as chromogen and counterstained with Mayer's haematoxylin. To confirm the specificity of staining, anti-harp 1 IgG was pre-incubated for 16 h at 4 °C with 20 µg/mL of C-terminal peptide and anti-harmonin IgG was pre-incubated for 16 h at 4 °C with 100 µg/mL GST-PDZ2/3 fusion protein before addition to the slides.

Confocal microscopy

Frozen adult mouse pancreata were embedded in optimal cutting temperature (OCT) reagent (Sakura Finetechnical Co. Ltd, Tokyo, Japan) sectioned at 5 µm, and fixed in 4% paraformaldehyde. βTC3 cells were grown to 50% confluency on sterile 18 × 18 mm microscope glass cover slips (Chance Propper Ltd, Smethwick, UK), the cover slips were washed once in PBS and fixed in 4% paraformaldehyde. The sections and cover slips were incubated with anti-harp 1 IgG followed by goat anti-rabbit IgG conjugated to Alexa Fluor 568 (Molecular Probes, Eugene, OR) and anti-harmonin IgG conjugated to Alexa Fluor 488 using a Zenon labelling kit (Molecular Probes). Confocal images were taken using a Leica TCS4 SP2 spectral confocal scanner and a Leica DMIRE2 microscope (Leica Microsystems) equipped with a 100× oil immersion objective.


This work was supported by the Juvenile Diabetes Research Foundation (JDRF)-National Health and Medical Research Council (NHMRC) of Australia Special Program Grant #219172. We thank Andrea Braakhuis for technical assistance and Dr Andrew Holland for critical review of the manuscript.