Proneural and proneuroendocrine transcription factor expression in cutaneous mechanoreceptor (Merkel) cells and Merkel cell carcinoma

Abstract

Merkel cells form part of the peripheral neuroendocrine system of the skin and act as mechanoreceptors in touch response. Merkel cell carcinoma (MCC) is a rare, aggressive disease with similarities to small cell lung cancer (SCLC), which is also of neuroendocrine origin. We previously identified a novel DNA binding protein complex specific for MCC suspension cell lines, termed Merkel nuclear factor (MNF) by its binding to the POU-IV family DNA binding consensus sequence. Here we report that MNF contains the POU-IV family member Brn-3c and that Brn-3c is expressed in normal Merkel cells. Additionally, Brn-3c protein reactivity is restricted to a subset of MCC biopsies and is not seen in biopsies revealing adherent, variant cell lines lacking neuroendocrine markers. Recently, proper development of murine Merkel cells was shown to require the proneural basic helix-loop-helix transcription factor, atonal family member, MATH1. We demonstrate a correlation between Brn-3c and HATH1 reactivity in MCC biopsies and cell lines with retention of neuroendocrine phenotype. In SCLC, the related basic helix-loop-helix transcription factor HASH1 is responsible for neuroendocrine phenotype, but HASH1 transcripts were not detected in MCC cell lines. We propose that HATH1 and Brn-3c may form a transcriptional hierarchy responsible for determining neuroendocrine phenotype in Merkel cells and that lack of Brn-3c and/or HATH1 in MCC may indicate a more aggressive disease requiring closer patient follow-up. © 2002 Wiley-Liss, Inc.

Merkel cells, found in the basal layer of the epidermis, are slow acting mechanoreceptors providing sensory information for touch and hair movement.1, 2 In the developing embryo their role seems to involve the formation of the subepidermal nerve plexus,3, 4, 5 hair follicles and sweat glands.6, 7, 8 The cells are neuroendocrine (NE) in origin, expressing neuron-specific enolase (NSE) and chromogranin (CHM), a basic peptide of neurosecretory granules9, 10, 11; they also express cytokeratin 20.12 Neuroendocrine cells are thought to be postmitotic; in fact, mitotic Merkel cells have never been observed in the skin but can be detected in skin grafts some 6 weeks after transplant, suggesting the presence of as yet unidentified stem cells; there is some debate in the literature as to whether these cells are of neural crest or epithelial origin.12, 13

Neuroendocrine tumours may be caused by aberrant expression of genes normally only active in precursor cells or perhaps when cells migrate from the neural crest to the skin. Merkel cell carcinoma (MCC; also known as neuroendocrine or small cell carcinoma of the skin) is a particularly aggressive form of skin cancer. Survival rates more closely equate to small cell lung cancer (SCLC), another NE-derived tumour, than other skin cancers.14 It is locally aggressive (50–80% of cases having regional lymph node involvement at presentation), with a high incidence of distant metastases (25–40%), and a significant mortality (with 25–30% of patients dying of their disease within 3 years).15

Cell lines derived from these tumours have been classified into 2 groups, classic and variant,16, 17 following the classification of Carney and Gazdar for SCLC lines.18, 19 Classic cell lines retain NE marker expression (such as chromogranin, CK20 and low molecular weight keratins) and contain neurosecretory granules, as evidenced by electron microscopy. In contrast, variant lines of MCC and SCLC do not contain neurosecretory granules and lose expression of NE markers with no reactivity for chromogranin, somatostatin, synaptophasin CAM.5.2 or neurofilament protein,17, 19 although MCC1, MCC14/1 and MCC26 retained reactivity for CK20,16 and the tumours from which they were derived expressed the epithelial marker CAM 5.2 in a dot-positive fashion at varying levels.17 No adherent MCC lines demonstrated reactivity for S100, reactivity of which is indicative of melanoma, or of leukocyte common antigen, which is indicative of lymphoma or high molecular weight keratin, which is indicative of squamous cell carinoma.17

Within these 2 groups, there is a further subdivision into 4 morphologic types, with type IV lines being adherent monolayers, whereas types I, II and III grow as suspension cultures that differ in their morphology from being tightly packed 3-dimensional clusters (type I) to loose 2-dimensional clusters (type III).16, 17, 18, 19

We have previously shown that MCC suspension cell lines express Merkel nuclear factor (MNF) DNA binding activity20 detected using the OA25 DNA fragment probe,21 which contains the Brn-3 family consensus sequence.22, 23 The Brn-3 or POU-IV family of transcription factors consists of 3 members in mammals, Brn-3a, -3b and -3c.24 Of these, Brn-3c has been shown to be critical for proper development of inner ear hair cells,25, 26 another peripheral nervous system mechanoreceptor. It has been postulated that the Merkel cell may have a similar mechanosensory function to inner ear hair cells due to the presence of cytoplasmic projections between neighbouring keratinocytes.27 Further support of functional similarity between inner ear hair cells and Merkel cells comes from work in Drosophila, in which both sound and touch responses are eliminated in atonal mutant flies. atonal is a member of the basic helix-loop-helix (bHLH) family of transcription factors with high evolutionary conservation.28, 29 Additionally in mice, the murine atonal ortholog 1, MATH1, is critical for both inner ear hair cell and Merkel cell development.29, 30

We now show that MNF contains Brn-3c and that Brn-3c is restricted to a subset of MCCs from which suspension cell lines arise on culture. Furthermore, normal Merkel cells in neonatal foreskin express both Brn-3c and the human atonal ortholog HATH1, but only classic type I MCC cell lines and not type II or type III cell lines express HATH1, even though these lines are reactive for Brn-3c. Neither Brn-3c nor HATH1 is expressed in adherent type IV Variant MCC cell lines.

bHLH, basic helix-loop-helix; CHM, chromogranin; EMSA, electrophoretic mobility shift assay; MCC, Merkel cell carcinoma; NE, neuroendocrine; NSE, neuron-specific enolase; RGC, retinal ganglion cell; SCLC, small cell lung cancer.

MATERIAL AND METHODS

Cell Lines and Biopsies

The MCC cell lines (MCC1, MCC5, MCC6, MCC13, MCC14/1, MCC14/2, MCC19 and MCC26) were all established in the Queensland Radium Institute Laboratory (QIMR) and have been described elsewhere.16, 17 The T95-45D MCC cell line was established at the Center for Medical Genetics, Ghent, Belgium. Other MCC cell lines were a gift from Dr. S.G. Ronan and Dr. T.K. Das Gupta (USIO; Department of Surgical Oncology, University of Illinois, Chicago, Illinois) and Dr. S.T. Rosen (MKL-1, MKL-2; Department of Medicine, Surgery and Pathology, Northwestern University, Chicago, Illinois). SCLC cell lines were CorL88 (a gift from Dr. P. Twentyman, Cambridge, UK), NCI-H69, NCI-H146 (obtained from the ATTC, Rockland, MD) and GLC-4 was a gift from Dr. M. Maliepaard (Department of Experimental Therapy, The Netherlands Cancer Institute, Amsterdam, The Netherlands). Ewing's sarcoma cell lines (Schultz, SKES and RDES) were established at QIMR. The neuroblastoma cell lines (NGP and SK-N-AS) were described earlier.31, 32 The large cell carcinoma cell line L111 was described by Center et al.33 and was a gift from Dr. R. Bowman (QIMR). All lines were grown in RPMI-1640/10% FCS at 37°C and maintained in a water-jacketed incubator at 37°C and 5% CO₂ at high humidity. Routine tests for mycoplasma using the Hoechst 33258 stain34 were negative. Biopsy material was collected at the time of establishment of the cell lines and stored at −70°C.16, 17

Extraction of DNA Binding Proteins

Whole cell extracts were prepared from exponentially growing cultures following the methods described by Schreiber et al.35 DNA restriction fragment probes were prepared by end-labelling of HindIII-EcoRI digests of OA2521 with polynucleotide kinase and γ-³²P-dATP. Nuclear protein extract (1–2 μg as determined by the Bradford assay) was used in an electrophoretic mobility shift assay (EMSA) with ³²P-labelled OA25 probes. The SV40 mutant Octa1 probe dpm8 with 3 μg of poly (dI-dC) as a nonspecific binding probe was used as a control. Samples were preincubated with the binding mix containing the poly (dI-dC) for 15 min at room temperature prior to incubation for another 15 min with the specific DNA probe. When antibody was used, this was added to the nuclear extract in the 15 min preincubation step. The profiles were resolved on 6% polyacrylamide/Tris-glycine gels, which were subsequently dried onto blotting paper and exposed to XAR-5 film (Kodak, Rochester, NY) overnight in sealed cassettes at −70° C. Antibodies used were anti-Brn-3a, anti-Brn-3b (gifts from Dr. E.E. Turner,36, 37 Department of Psychiatry and Program in Neuroscience, University of California, San Diego, CA) and anti-Brn-3c (BAbCo, Richmond, CA and described in ref. 24).

RT-PCR

Total RNA was extracted using Total RNA Isolation Reagent (Applied Biotechnologies, Surrey, UK) from cells in exponential growth, and 2 μg of RNA was reverse-transcribed with SuperScript II (Life Technologies, Bethesda, MD) and oligo-dT primers (Boehringer Mannheim, Indianapolis, IN) for cell lines or random hexamers (Promega, Madison, WI) for tumour samples. RNA isolated from cell lines was treated with DNase I prior to the reverse transcription reaction. The resulting cDNA was amplified with primers specific for each gene. Primers for brn-3c, HASH1 and HES1 were as described.24, 38, 39 For HATH1, primers and PCR conditions were as per human STS UT6525 (GenBank accession number L30585), which was used to map HATH1.40 The product resulting from HATH1 RT-PCR was cloned into pGEM-T (Promega), sequenced using dye terminator chemistry and found to be identical to HATH1 after a BLAST search. Biopsy cDNA was amplified in the presence of α³²P-dCTP in the PCR mix to allow for autoradiography. For all genes, RT-PCR products were electrophoresed through agarose gels in 1 × TAE buffer and visualised by ethidium bromide staining (for cell lines) or autoradiography (for biopsy material).

Immunohistochemistry

Cell lines were prepared as previously.41 Frozen sections of MCC biopsy material were prepared, and immunohistochemistry was performed essentially as described previously,42 except that sections were blocked with 1% goat serum/0.1% Triton X-100 in TBS (pH 7.4) for 15 min at room temperature. Antibodies were incubated overnight at room temperature in a humidified chamber at either 1:400 dilution for Brn-3c (BAbCo) and MATH1/HATH1 (a gift from Dr. J.E. Johnson,43 UT, Southwestern Medical Center, Dallas, TX) or 1:200 dilution for NSE, in 1% goat serum/0.1% Triton X-100 in TBS, pH 7, followed by secondary biotinylated goat-rabbit IgG (Zymed, San Francisco, CA; prediluted) and streptavidin-horseradish peroxidase (Zymed; prediluted). The chromophore was 3,3′-diaminobenzidine (DAB) with H₂O₂.

Fluorescent Microscopy

Human neonatal foreskin was obtained through routine circumcision, washed in PBS and cut into small sections before being incubated in Dispase II (50 μg/ml in PBS) for 3 hr at 37°C. Epidermal sheets were teased from the dermis, fixed in acetone at −20°C and stored at −20°C until required. After rehydration in PBS, epidermal sheets were, blocked for 90 min at room temperature in 3% FCS in PBS and incubated with primary antibody to chromogranin (Biogenex, San Ramon, CA), Brn-3c (BAbCo) or HATH1, or a combination of both, overnight, at 4°C (1:250 dilution in blocking solution). Washes were 5 × 5 min in PBS/0.05% Tween 20 followed by 1 wash in PBS only, all with gentle agitation at room temperature. Secondary antibody either alone or in combination [Alexa 488-conjugated anti-mouse for chromogranin and cytokeratin-20 (Dako, Carpinteria, CA), and Alexa 594-conjugated anti-rabbit for Brn-3c and HATH1; Molecular Probes, Eugene, OR] diluted to 1:500 in blocking solution for 1 hr at 37°C in the dark. Washes, in the dark, were repeated before mounting onto slides in Antifade (Vectashield, Vector, Burlingame, CA) with a coverslip for viewing by fluorescent microscopy using an Axioscop microscope (Carl Zeiss) fitted with epifluorescence optics. Photographs were taken on Kodak Tri-X 400 film, and processed slide film images were then scanned to convert to a digital image.

Western Analysis

Total protein lysates were prepared by harvesting cells, washing once in ice-cold PBS and lysing in 1% SDS in cell lysis buffer [10 mM Tris, pH 7.4, 20% glycerol, 2 mM phenylmethylsulfonyl fluoride (PMSF) and a Protease Inhibitor Cocktail Tablet at 10 ml (Roche, Mannheim, Germany)] followed by sonication. Lysates containing equal total protein as determined by BCA (Pierce, Rockford, IL) were denatured by addition of dithiothreitol (DTT; to 10 mM) and heating to 60°C. Samples were loaded onto 5% SDS-PAGE stacking gels, and proteins were resolved by electrophoresis at 200 V through 10% SDS-PAGE resolving gels. Protein was transferred overnight at 4°C in buffer (14.4 g/L glycine, 3.05 g/L Tris and 20% methanol) to nitrocellulose membranes. Membranes were blocked at room temperature for 1 hr in blocking solution [5% skim milk powder in 0.1% Tween 20/PBS]. Primary antibody reactive to the 28 kDa isoform of HES1 (BD Biosciences, San Jose, CA) was incubated at a 1:1,000 dilution in 5% skim milk powder in 0.1% Tween 20/PBS for 90 min at room temperature, followed by 2 × 2 min and 3 × 5 min washes in 0.1% Tween 20 /PBS. Membranes were incubated for 90 min at room temperature in appropriate horseradish peroxidase-conjugated secondary antibodies, diluted to 1:2,000 in blocking solution followed by 2 × 2 min and 3 × 5 min washes in 0.1% Tween 20 /PBS; then ECL detection was performed according to the manufacturer's protocol (NEN, Boston, MA).

RESULTS

MNF Contains Brn-3c

We have previously reported on the novel DNA binding activity of MNF, found in MCC suspension cell lines but not present in MCC adherent lines or other neuroendocrine cell lines using the nonconsensus octamer recognition sequence OA25.20 Since the OA25 DNA fragment contains the Brn-3 family consensus DNA binding sequence,22, 23 attempts were made to inhibit binding with antibodies specific for Brn-3a,44 Brn-3b,36 and Brn-3c.24 Introduction of Brn-3c antibody specifically inhibited MNF migration (Fig. 1, closed arrow) and produced a supershift of the MNF complex (Fig. 1, open arrow), without altering the binding activity of other proteins, including the brn-2 POU domain complexes N-Oct-3 and N-Oct-5, which have strong binding activity for the OA25 probe45(Fig. 1 and summarised in Table I).No effect on MNF activity was seen with antibodies against Brn-3a and Brn-3b (data not shown). These results clearly show that MNF contains Brn-3c.

Figure 1. EMSA analysis of MCC cell lines. EMSA of 2 MCC suspension cell lines (MCC6, MCC19); 2 MCC adherent cell lines (MCC13, MCC26); and a Ewing's sarcoma cell line (Schultz). Nuclear extracts were incubated with OA25 probe in the presence (+) and absence (−) of anti-Brn-3c antibody. Closed arrow indicates the position of MNF binding complex and open arrow the position of the inhibited (supershifted) MNF complex, identifying MNF as containing Brn-3c. Oct-1 is a ubiquitously expressed POU domain complex, N-Oct-2 a POU domain protein ubiquitously expressed in neural tissue and N-Oct-3 and N-Oct-5 are the complexes of the brn-2 POU domain gene. FP free probe.

Table I. Immunohistochemical Analysis of Brn-3c and Hath1 in Merkel Cell Carcinoma Biopsies and Cell Lines1

Tumour/cell line	Growth habit/classification	NSE	CHM	Brn-3c	HATH1
1 NSE, neurone-specific enolase; CHM, chromogranin; TC, tissue culture cell line; S, grows in suspension; A, grows as an adherent monolayer. Morphological types as defined for small cell carcinoma cell lines: type I, tight 3-D cluster; type II, loose 3-D cluster; type III loose 2-D cluster; type IV, adherent monolayer; c-MCC, classic MCC cell line; v-MCC, variant MCC cell line. Neg, no reaction; +/−, trace amount of staining; +, 0–25% of cells staining; ++, 25–50% of cells staining; +++, 50–75% of cells staining; ++++, 75–100% of cells staining. ND, not done. 2 MNF detected by EMSA. 3 No MNF present by EMSA.
MCC1		+	ND	+++	+++
MCC1 TC	S, III, v-MCC	++++	Neg	++++2	Neg
MCC2		+/−	ND	++	+
MCC2 TC	S, II, v-MCC	++	Neg	ND	ND
MCC4		++	ND	+	++
MCC5		++	ND	+	Neg
MCC5 TC	S, I, c-MCC	++++	++++	++++2	++
MCC6		++++	ND	+++	++++
MCC6 TC	S, I, c-MCC	++++	++++	++++2	++++
MCC8		Neg	ND	+++	++
MCC10		+/−	ND	+	Neg
MCC11		Neg	ND	+/−	Neg
MCC13		Neg	ND	Neg	Neg
MCC13 TC	A, IV, v-MCC	Neg	Neg	Neg3	Neg
MCC14		+/−	ND	+/−	Neg
MCC14/1 TC	A, IV, v-MCC	+	Neg	Neg3	Neg
MCC14/2 TC	A, IV, v-MCC	Neg	Neg	Neg3	Neg
MCC16		Neg	ND	Neg	Neg
MCC17		+	ND	+	++
MCC19 TC	S, II, v-MCC	++++	ND	+++2	Neg
MCC22		Neg	ND	Neg	+/−
MCC26 TC	A, IV, v-MCC	+	Neg	Neg3	Neg

Normal Merkel Cells Express Brn-3c

To determine whether normal Merkel cells express Brn-3c, we examined human neonatal foreskin epidermis by fluorescent microscopy using a monoclonal antibody to CHM, a basic peptide found in neurosecretory granules and specific for Merkel cells within the epidermis.11 Colocalisation of CHM and Brn-3c fluorescence was seen in cells in the basal layer of the epidermal sheet (Fig. 2^a–c).46 Additionally, confocal microscopy of human scalp hair follicles after dual labelling with cytokeratin 20 and Brn-3c showed colocalisation of label at the base of the follicle where Merkel cells are positioned (data not shown).

Figure 2. Expression of Brn-3c and HATH1 in normal neonatal foreskin. Acetone-fixed epidermal sheets, teased from neonatal foreskin after treatment with dispase, were reacted with antibodies to CHM and Brn-3c (a–c) or CHM and HATH1 (d–f) followed by secondary antibodies conjugated to Alexa dyes (Molecular Probes), and activity was visualised and photographed using an Axioscop microscope (Carl Zeiss) fitted with epifluorescence optics and Kodak Tri-X 400 film. Digital images of fluorescence were then converted to greyscale. Row 1 shows (a) brightfield, (b) CHM and (c) Brn-3c for CHM and Brn-3c dual label. Row 2 shows (d) brightfeld, (e) CHM and (f) HATH1 for CHM and HATH1 dual label. Scale bar = 5 μm.

Merkel Cell Carcinomas Express Brn-3c

RT-PCR amplification of cDNA using primers specific for brn-3c produced a product of expected size from suspension cell lines but failed to produce a product from adherent cell lines (Fig. 3a, panel 1). Moreover, RNA extracts of MCC biopsy material corresponding to the cell lines were available for some samples, and brn-3c expression was confined to those tumours that resulted in MCC suspension cell lines and therefore was not a tissue culture artifact (Fig. 3b). Two of 4 SCLC suspension cell lines (NCI-H69, CorL88) also had brn-3c product, but none of the Ewing's sarcoma or neuroblastoma cell lines or the large cell lung cancer cell line L111 expressed brn-3c.

Figure 3. Expression of proneuronal and proneuroendocrine transcription factors in MCCs. (a) Total RNA from cell lines was reverse transcribed, and the resulting cDNA was amplified using specific primers, electrophoresed through an agarose gel and visualised by ethidium bromide staining. brn-3c (row 1), HATH1 (row 2), HASH1 (row 3) and HES1 (row 4). (b) Total RNA extracts of MCC biopsy material were reverse transcribed, and PCR was performed with specific primers for brn-3c with incorporation of α-³²P-dCTP in the nucleotide mix. Resulting PCR products were electrophoresed through an agarose gel and visualised by autoradiography.

To confirm that only MCC suspension cell lines expressed Brn-3c, immunohistochemical analysis of MCC cell lines was undertaken. MCC suspension cell lines demonstrated strong nuclear reactivity, whereas adherent lines exhibited nonspecific cytoplasmic staining (Fig. 4). Further confirmation was given in a larger series of frozen sections of biopsy material. Of 13 MCC biopsies, 8 exhibited strong nuclear reactivity, 3 were negative except for nonspecific staining (MCC13, MCC16, MCC22) and 2 (MCC14, MCC11) exhibited trace amounts of reactivity (Fig. 4; summarised in Table I). Of the biopsies in which cell lines were established, only those biopsies from which MCC classic lines were derived exhibited strong nuclear reactivity.

Figure 4. Examples of immunoreactivity of MCC cell lines and tumour biopsies for Brn-3c, HATH1 and NSE. MCC6 exhibits strong nuclear staining for both Brn-3c and HATH1, as well as strong cytoplasmic staining for NSE in the suspension cell line and in the tumour biopsy from which it was derived. MCC14/2, an adherent cell line, exhibits only nonspecific reactivity for Brn-3c and is negative for HATH1 and NSE. The tumour biopsy from which it was derived has weak reactivity for NSE and Brn-3c but is negative for HATH1. Scale bar = 0.5 μm.

Merkel Cell Carcinomas Contain HATH1

Gene deletion studies in the mouse have recently demonstrated a requirement for MATH1 in the inner ear, prior to Brn-3c proteins being detectable.29 We therefore extended our study of MCC cell lines to examine expression of the human ortholog HATH1. Only the MCC suspension cell lines gave a positive PCR product with primers for HATH1, and no transcript was seen for any SCLC, Ewing's sarcoma or neuroblastoma cell line (Fig. 3a, panel 2). Additionally, examination of methanol-fixed cells and frozen biopsy material with an antibody originally raised against MATH1,43 but reactive for HATH1, revealed reactivity in some MCC cell lines and biopsies. There was strong nuclear reactivity in 2 of 4 MCC suspension cell lines examined (MCC5, MCC6). However, 2 suspension lines, MCC1 and MCC19, and all the adherent MCC cell lines had no reactivity for HATH1 (Fig. 4). The MCC1 tumour was strongly reactive for HATH1, but no biopsy material was available for MCC19. Conversely, the MCC5 tumour biopsy was unreactive to HATH1 antibody, but the cell line was positive although not as strongly so as the MCC6 cell line (summarised in Table I).

HATH1 Expression in Normal Merkel Cells

Fluorescent microscopy was again utilised to determine whether normal Merkel cells within epidermal sheets of neonatal foreskin expressed HATH1. Expression of HATH1 protein colocalized with that of chromogranin (Fig. 2d–f), demonstrating that Merkel cells within the neonatal epidermis express HATH1.

Control of NE Phenotype in Merkel Cells

In SCLC, which MCC closely resembles, the related bHLH transcription factor HASH1 (for human achaete-scute homolog) has been shown to be responsible for the NE phenotype;47, 48 in addition, HASH1 is negatively regulated by the HES1 transcription factor, whose expression abolishes HASH1 in these cells.49 We therefore examined the expression of these genes in MCC cell lines to determine whether HASH1 regulated the NE phenotype in MCC and/or whether HES1 prevented it in adherent lines. RT-PCR analysis failed to detect HASH1 transcripts in any MCC or Ewing's sarcoma cell line examined, although product was seen in SCLC and neuroblastoma cell lines, as reported elsewhere47, 50 (Fig. 3a, panel 3), indicating that the NE phenotype in MCC is not due to HASH1. HES1 transcripts were detected in all cell lines examined (Fig. 3a, panel 4), but as RNA levels may not reflect protein levels, HES1 protein levels were determined by Western analysis. Examination of protein lysates indicated that the 28 kDa isoform of HES1 was present in all cell lines examined (MCC, SCLC and Ewing's sarcoma) in approximately equal quantities (Fig. 5) and thus is unlikely to be responsible for the lack of the NE phenotype in MCC adherent cultures.

Figure 5. Western analysis of neuroendocrine cell lines for HES1 protein. Protein lysates of MCC (3 suspension; MCC1, MCC5 and MCC6, and 3 adherent, MCC13, MCC14/1 and MCC14/2), SCLC (NCI-H69, CorL88) and Ewing's sarcoma (Schultz) cell lines as well as the cervical carcinoma cell line, HeLa as control, were electrophoresed at equal protein loading through a 10% SDS-PAGE gel, and the protein was transferred to nitrocellulose for Western analysis with antibody to HES1. All cell lines examined exhibited strong reactivity to HES1 antibody in a band of 28 kDa size.

We have shown that HATH1 and Brn-3c proteins are present in normal Merkel cells and in the classic MCC suspension cell lines (MCC5, MCC6), which have strong expression of CHM and NSE. The variant MCC suspension lines MCC1 and MCC19, although still expressing NSE, had no reactivity for HATH1 but were reactive for Brn-3c. All MCC adherent cell lines that were unreactive for CHM and had either no or only trace amounts of NSE failed to express HATH1 or Brn-3c protein, and no transcripts of either gene were detected.

DISCUSSION

Merkel cell carcinoma is an aggressive skin cancer with a poor prognosis, and the present treatment regimes are not effective.15, 51 Until we have a better understanding of the biology, we are unlikely to improve patient outcomes or have more selective treatments. Two families of proteins involved in maintaining cell fate with a restricted expression pattern to cells of neural origin are the POU-IV (Brn-3) class of POU domain transcription factors and the bHLH family member atonal.

Drosophilaatonal and the murine ortholog MATH1 have been extensively studied, but less is known of the human ortholog HATH1. Recently, Ben-Arie and co-workers,29 using a mutant mouse containing lacZ in the coding region of the MATH1 locus, were able to demonstrate that β-galactosidase-positive staining cells in sections of hairy skin, footpad and vibrissae were Merkel cells. Examination of markers for Merkel cells in MATH1 nullizygous mice revealed that Merkel cells were still present, indicating a requirement for MATH1 in Merkel cell maturation but not genesis.29

Both MATH1 and Brn-3c are required for correct development of the murine inner ear hair cells, and, not surprisingly, the phenotypes of brn-3c^−/− mice and MATH1^−/− mice are similar.26, 52 Furthermore, a 8 bp deletion in the homeodomain coding region of the brn-3c gene results in low-affinity DNA binding and has been implicated in hereditary hearing loss (DFNA15) in humans.53 Interestingly, the inner ear of the HES1^−/− mouse contains supernumerary inner ear hair cells, and HES1 has been postulated to regulate MATH1 expression negatively.54 Regulation of these genes appears to be through modular elements both 5′ and 3′ of the coding region, and, in the mouse, includes autoregulation by MATH1.30

Recently, gene deletion studies in mice, chickens and Xenopus have indicated that expression of brn-3a and brn-3b genes may be dependent on MATH5.55, 56, 57 MATH5 has been implicated in retinal ganglia cell (RGC) development, as knockout mice lack 80% of RGCs, and virtually no Brn-3b-positive RGCs were seen, suggesting that MATH5 acts upstream of Brn-3b.57 Although such data have not been reported, these mice are unlikely to express Brn-3a either, as brn-3b^−/− mice have no detectable Brn-3a.26 No human ortholog of MATH5 has been described; however, a sequence with similarity to MATH5 has been identified within a clone (GenBank accession number NT_029388) that maps to 10q21, a region where we have recently demonstrated high rates of loss of heterozygosity (LOH) in MCC,58 and previous comparative genome hybridisation (CGH) studies of SCLC and MCC showed frequent losses.59 Given the transcriptional hierarchy between MATH5 and brn-3b, the phenotypes of the knockout mice and the colocalisation of MATH1 and Brn-3c protein in Merkel cells of neonatal epidermal sheets, it is conceivable that a similar hierarchy exists between MATH1/HATH1 and Brn-3c in normal peripheral nervous system mechanoreceptors in mice and humans.

We have shown that normal Merkel cells in humans express both HATH1 and Brn-3c and that a subset of MCCs contains these proteins, with classic MCC cell lines retaining expression of both proteins, whereas variant lines may express Brn-3c but do not express HATH1. Although the number of cell lines examined is small, it appears that HATH1 expression may correlate with CHM expression, as all lines that were reactive for HATH1 were also reactive for CHM, a peptide of neurosecretory granules. This finding would suggest that HATH1 may be required for maintenance of the NE phenotype in MCCs. Additionally, as no MCC cell lines examined express HASH1 transcripts, the NE phenotype in MCC cell lines is unlikely to be under HASH1 control, as reported for SCLC,47, 48 and other transcriptional regulators must maintain the NE phenotype in these cells. Our study, however, does not exclude the possibility that HASH1 may be involved in determination of normal Merkel cell fate, and perhaps examination of mice lacking MASH148 may elucidate whether this gene is required for normal Merkel cell development.

We propose that Brn-3c and HATH1 are possible candidates for regulation of cell fate and maintenance of the NE phenotype in Merkel cells, and our results suggest that HES1 may not be responsible for the decrease in NE expression in variant cell lines. Given that in SCLC, in which the numbers of cases are large enough to show statistical significance, patients from whom variant cell lines are established have a poorer prognosis.18, 19 Expression of HATH1 and/or Brn-3c may prove to be a useful prognostic marker, with lack of expression indicating a subset of MCCs in which aggressive treatment and closer follow-up of patients may be warranted.

Acknowledgements

J.H.L. was the recipient of an NH&MRC scholarship, A.L.C. was the recipient of a University of Queensland postgraduate scholarship. We thank Drs. M.M. Hughes, V.M. Hinkley, R.W. Allison, W. Cockburn, T.J. Harris and O. Williams for their support in collecting the specimens and Mr. G. Dietrich for photography. We are grateful for the gifts of Brn-3a and Brn-3b (from Dr. E.E. Turner) and MATH1 (from Dr. J.E. Johnson) antibodies.