Báo cáo khoa học: The phosphorylation pattern of human as1-casein is markedly different from the ruminant species

doi:10.1046/j.1432-1033.2003.03755.x

Eur. J. Biochem. 270, 3651–3655 (2003) (cid:1) FEBS 2003

The phosphorylation pattern of human as1-casein is markedly different from the ruminant species

Esben S. Sørensen, Lise Møller, Maria Vinther, Torben E. Petersen and Lone K. Rasmussen*

Protein Chemistry Laboratory, Department of Molecular Biology, University of Aarhus, Denmark

lated at Ser18 and Ser26. Both phosphorylation sites are located in the amino acid recognition sequence of the mammary gland casein kinase. Notably, no phosphoryla- tions were observed in the conserved region covering resi- dues Ser70–Glu78, which is extensively phosphorylated in the ruminant as1-caseins.

Keywords: as1-casein; human milk; mammary gland casein kinase; phosphorylation.

Caseins are highly phosphorylated milk proteins assembled in large colloidal structures termed micelles. In the milk of ruminants, as1-casein has been shown to be extensively phosphorylated. In this report we have determined the phosphorylation pattern of human as1-casein by a combi- nation of matrix-assisted laser desorption mass spectrometry and amino acid sequence analysis. Three phosphorylation variants were identified. A nonphosphorylated form, a variant phosphorylated at Ser18 and a variant phosphory-

counterpart by a lower degree of phosphorylation [10]. For many years it was generally accepted that as1-casein was absent or present in only very small amounts in human milk [2]. In the mid-1990s, two groups isolated and sequenced a minor 27-kDa casein component that was identified as being the human counterpart of as1-casein [11,12]. In addition, it was shown that this as1-casein component forms disulfide-bonded heteromultimers with j-casein in human milk [12]. The molecular cloning and sequencing of mRNA transcripts revealed the presence of three forms of as1-casein in human milk [13,14]. In the present study, we report the phosphorylation pattern of human as1-casein.

Materials and methods

Materials

Caseins are the predominant milk proteins of most mammalian species [1]. In ruminants, about 75% of the milk protein content is constituted of caseins. The corresponding figure for human milk is only about 40% [2]. In the milk of ruminants, caseins interact with calcium phosphate forming large stable colloidal particles termed micelles. These micellar complexes make it possible to maintain a supersaturated calcium phosphate concentra- tion in milk, providing the newborn with sufficient calcium phosphate for the mineralization of the rapidly growing calcified tissues. In this context, the phosphory- lation of the individual caseins plays a significant role in the interaction with calcium phosphate and thereby the organization of the micelles. The ruminant caseins, which are the most intensely studied, comprise as1-, as2-, b- and j-casein. Their phosphorylation pattern has been the basis of many studies and the general feature is that they are highly phosphorylated proteins, phosphorylated by the mammary gland casein kinase [3–7]. The primary requirement for phosphorylation by this kinase is a glutamate, a phospho- serine or an aspartate two residues to the C-terminal side of the phosphoacceptor site (S-x-E/Sp/D) [8,9].

Trypsin (EC 3.4.21.4) was obtained from Worthington Biochemical Corporation (Freehold, NJ, USA). Vydac C4 and C18 reverse-phase resins were from The Separations Group (Hesperia, CA, USA) and the RP C2/C18 column was from Amersham Biosciences AB (Uppsala, Sweden). Reagents used for sequencing were from Applied Biosys- tems (Foster City, CA, USA). All other reagents were of analytical reagent grade.

Compared with ruminants, human milk contains a very low concentration of calcium phosphate and the function of casein in delivering calcium to the neonate is therefore muted in this species. In human milk, the predominant caseins are j- and b-casein, which differ from its ruminant

Purification of human as1-casein

Human as1-casein was purified from human milk as described [12]. During this procedure, the protein was reduced and alkylated to dissociate the disulfide-linked complex consisting of as1- and j-casein. To remove small residual amounts of j-casein, the protein was subjected to reverse-phase chromatography on a Vydac C4 reverse-phase column. The purity of the resulting as1-casein was verified by SDS/PAGE and N-terminal amino acid sequence analysis.

Correspondence to E. S. Sørensen, Protein Chemistry Laboratory, Department of Molecular Biology, University of Aarhus, Science Park, Gustav Wieds Vej 10, DK-8000 Aarhus, Denmark. Fax: + 45 8 6136597, Tel.: + 45 8 9425092, E-mail: ess@imsb.au.dk Enzyme: Trypsin (EC 3.4.21.4) *Present address: Symphogen A/S, DK-2800 Denmark. (Received 5 May 2003, revised 14 July 2003, accepted 16 July 2003)

3652 E. S. Sørensen et al. (Eur. J. Biochem. 270)

(cid:1) FEBS 2003

Generation, separation and characterization of peptides

Table 1. Characterization of peptides from the tryptic digest of human as1-casein. Peak numbers designations correspond to those of Fig. 1. The amino acid sequence was identified by sequence analysis and/or MALDI-TOF MS. Calculated MH+, calculated protonated mono- isotopic masses; observed MH+, molecular monoisotopic protonated mass determined by MALDI-TOF MS.

Approximately 300 lg of reduced and alkylated human as1-casein was digested with trypsin using a ratio of enzyme to substrate of 1 : 100 (w/w) in 0.1 M ammonium bicar- bonate, pH 8.1, at 37 (cid:2)C for 6 h. Separation of the peptides was carried out by reverse-phase HPLC on a Vydac C18 column and detected in the effluent by measuring the absorbency at 226 nm (as described in the legend to Fig. 1). Fraction 35 (Fig. 1) was rechromatographed by reverse- phase HPLC on a SMART-system equipped with a 2.1 · 100 mm C2/C18 RPC column using a gradient of acetonitrile in 0.05% heptafluorobutyric acid at 25 (cid:2)C.

Peak number Sequence Calculated MH+ Observed MH+

Peptides were characterized by mass spectrometric- and amino acid sequence analysis. Mass spectrometric analyses of the peptides were performed using a MALDI-TOF mass spectrometer (Voyager DE PRO, Applied Biosystems Inc.). Theoretical peptide masses were calculated using the GPMAW program (Lighthouse Data, Odense, Denmark). Amino acid sequence analysis was performed on an automated amino acid sequencer (ABI 477A/120A; Applied Biosystems Inc.). To locate phosphoserines in the sequence, phosphopeptides were treated with ethanethiol to convert phosphoserine into S-ethylcysteine [15] which can be identified by amino acid sequence analysis as PTH- S-ethylcysteine after its release in the corresponding cycle. PTH-S-ethylcysteine eluted just before the diphenylthiourea peak in the system used [16].

45–50 8–11 43–50 45–53 1–7 84–90 54–67 39–42 28–36 4–7

Results and discussion

a Peaks from rechromatography of fraction 35 from Fig. 1.

Human as1-casein was purified in a reduced and carboxy- methylated state as described [12]. The protein was digested with trypsin and the resulting peptides were separated by reverse-phase chromatography (Fig. 1). Fractions were collected and the peptides were characterized by mass spectrometric- and sequence analysis. The combined results are shown in Table 1. The amino acid sequence of human

as1-casein is shown in Fig. 2. Peptides identified by mass spectrometric analysis and/or sequence analysis are under- lined. As seen in the Fig. 2, peptides covering the entire

11 16 17 19 23 25 25 26 26 28 35–1a 35–2a 36 37 38 41 43 44 49 53 54 54 62 65 733.37 564.28 990.51 1105.55 879.59 972.36 1605.71 515.33 1143.46 498.34 12–27+2P 1942.82 1718.75 68–83 1043.6 4–11 12–27+1P 1862.86 12–27 1782.89 12–36+2P 3067.34 12–36+1P 2987.34 2267.16 91–109 904.44 164–171 2580.21 142–163 2591.24 111–132 1170.57 133–141 3731.75 133–163 6303.97 111–163 733.29 564.25 990.55 1105.5 879.6 972.29 1605.77 515.27 1143.38 498.24 1942.78 1718.75 1043.62 1862.78 1782.86 3067.26 2987.26 2267.15 904.38 2580.44 2591.18 1170.5 3731.12 6303.61

Fig. 1. Reversed-phase separation of a trypsin digest of human as1-casein. Human as1-casein was digested with trypsin as described in Materials and methods. Peptides were eluted with a gradient of 80% acetonitrile in 0.1% trifluoroacetic acid (dotted line) on a Vydac C18 (10 lm) column (4 · 250 mm). The col- umn was operated at 40 EC and the flow rate was 0.85 mLÆmin)1. Peptides were detected in the effluent by recording the absorbance at 226 nm (solid line), collected manually and characterized as described in the text.

(cid:1) FEBS 2003

Phosphorylation of human as1-casein (Eur. J. Biochem. 270) 3653

Our laboratory has much experience of employing MALDI-TOF mass spectrometric analysis for identification and localization of phosphorylation sites in proteins [16–18]. MALDI-TOF mass spectrometric analysis of a phospho- peptide results in a spectrum with an easily identifiable fragmentation pattern which is characteristic for phospho- rylated serines. These spectra contain a series of peaks separated by approximately 98 Da, which represents the fragmentation of a phosphoserine to dehydroalanine.

sequence of human as1-casein have been identified and characterized in this study.

Glycosylation

In this work, we have analyzed all fractions from the reverse-phase HPLC separation of the tryptic digest of human as1-casein (Fig. 1) by MALDI-TOF mass spectro- metric and N-terminal sequence analysis. We identified four fractions with the characteristic fragmentation pattern of peaks at 35, 37, 41 and 43. A representative MALDI-TOF spectrum of peak 41 showing the characteristic fragmenta- tion of a phosphopeptide is shown in Fig. 3. Peak 35 was found to contain two peptides which potentially could be phosphorylated, thus the fraction was rechromatographed by reversed-phase HPLC on a SMART HPLC system to separate the two components, 35–1 and 35–2. Peak 35–2 gave a mass spectrum with only one mass at 1718.75 Da which is identical with the calculated protonated mass for the tryptic peptide covering residues 68–83, thereby showing that this peptide is not phosphorylated in human as1-casein. This observation was confirmed by Edman sequencing of the peptide, which showed normal yields of PTH-serine in all relevant cycles. Mass analysis of peak 35–1 showed a mass of 1942.78 Da, as well as two populations of ions at approximately )98 Da and )196 Da. This triplet of ions, each separated by approximately 98 Da, indicates that peak 35–1 contains a phosphopeptide with two phosphoserines. Furthermore, the observed mass at 1942.78 Da correlates with the calculated protonated mass (1942.82 Da) for the tryptic peptide covering residues 12–27 and containing two phosphorylations (159.93 Da). Mass analysis of peak 37 showed a mass of 1862.78 Da, and a single fragmentation ion at )98 Da was observed, indicating the presence of a single phosphorylation in the peptide. Furthermore, the

Three asparagine residues in human as1-casein (Asn14, Asn54, Asn154) are located in the putative glycosylation sequence Asn-X-Ser/Thr. In the case of Asn14 and Asn154, a neighbouring proline residue in position X corrupts the glycosylation sequence and renders it unfit for glycosylation. Regarding Asn54, this study did not show any evidence for glycosylation of this residue in human as1-casein. Mass spectrometric analysis of peak 25 (Fig. 1) containing the peptide Asn54–Lys67 showed a mass of 1605.77 Da which corresponds to the calculated protonated monoisotopic mass (1605.71 Da) of the unmodified peptide sequence, thereby showing that Asn54 was not glycosylated in human as1-casein.

Likewise in this study, we observed no O-glycosylations

in human as1-casein.

Phosphorylation

Fig. 2. Localization of phosphorylations in human as1-casein. The amino acid sequence was deduced from the cDNA sequence [11]. Solid lines indicate isolated and characterized peptides (Table 1). P denotes identified phosphorylation. Peptides are numbered according to the reversed-phase elution profile in Fig. 1.

Human as1-casein contains 16 serines and four threonines, where nine of the serines and one threonine are located in the recognition sequence of the mammary gland casein kinase [8,9]. The recognition sequence (Ser/Thr-X-Glu/ Ser(P)/Asp), comprises an acidic residue, glutamic acid, aspartic acid or a phosphorylated residue, as the second amino acid to the C-terminal side of the serine or threonine to be targeted. Especially interesting, human as1-casein contains a serine rich region, SSISSSSEE(70–78), where five of the serines are located in the recognition sequence of the mammary gland casein kinase. This region is highly conserved among all species with known as1-casein sequences (for alignment see [13]), and in all analyzed species a high degree of phosphorylation has been observed in this region [3–5].

Fig. 3. MALDI-TOF MS of peak 41 from Fig. 1. The protonated mass at m/z 3067.26 corresponds to the peptide 12–36 including to phosphorylations. The characteristic fragmentation pattern confirms the presence of two phosphorylations in the peptide.

3654 E. S. Sørensen et al. (Eur. J. Biochem. 270)

(cid:1) FEBS 2003

mass 1862.78 Da correlates with the calculated protonated mass of residues 12–27 containing a single phosphorylation (1862.86 Da). Finally, mass analysis of peak 38 showed a mass which correlates with the mass of the peptide covering residues 12–27 without any modifications. In conclusion, we have observed the peptide 12–27 in three different forms, with zero, and one and two phosphate groups attached. Peaks 41 and 43 represent peptide 12–36 with two and one phosphorylated groups, respectively. These peptides, result- ing from incomplete cleavage at Arg27, do not contain any additional serines or threonines compared with peaks 35 and 37, and thus they were not characterized further. The peptide, LQNPSESSEPIPLESR(12–27)

(peak 35–1), contains three serines located in the recognition sequence of the mammary gland casein kinase (Ser16, Ser18 and Ser26). To determine which of the serines are in fact phosphorylated, we subjected the two peptides to an ethanethiol treatment followed by Edman sequencing as outlined in Materials and methods. The ethanethiol treat- ment converts the labile phosphoserine residues into S-ethylcysteine, which is more stable and able to withstand the relatively harsh Edman chemistry during automated sequencing [15]. Furthermore, PTH-S-ethylcysteine elutes in an open window just before the diphenylthiourea peak in the on-line HPLC system used in these studies. Sequence analysis of peptide 35–1 succeeding the ethanethiol treat- ment revealed PTH-S-ethylcysteine in cycles 7 and 15, corresponding to Ser18 and Ser26 in human as1-casein. Likewise, sequence analysis of peptide 37, after the ethanethiol treatment gave PTH-S-ethylcysteine in sequence cycle 7, corresponding to Ser18 in human as1-casein. The yields of PTH-serine in cycles corresponding to Ser16 and Ser19 were as expected, indicating that these residues were not phosphorylated.

In conclusion, these studies show that human as1-casein exists in three phosphorylation variants. A nonphosphory- lated form, a variant containing a single phosphorylation at Ser18 and a variant phosphorylated at Ser18 and Ser26. It is difficult to determine the quantitative relation between the three phosphorylation variants, but judged by the reversed-phase HPLC trace in Fig. 1, the variant containing a single phosphorylation at Ser18 is the major variant ((cid:1) 50%), followed by the nonphosphorylated form ((cid:1) 30%) and the doubly phosphorylated variant ((cid:1) 20%). The degree of identity between human and other known as1-casein sequences is overall low (alignment of sequences is shown in [6,13]). The phosphorylations have been charac- terized in ovine (Ovis aries), caprine (Capra hircus), bovine (Bos taurus), water buffalo (Bubalus bubalis) and camel (Camelus dromedarius) as1-casein. Generally, all of these species have been reported to have a significant higher number of phosphorylations than shown to be the case for human as1-casein in this study. Bovine as1-casein is phos- phorylated at up to nine positions depending on the genetic variants [3], the ovine as1-casein is phosphorylated at up to 11 positions [5], the water buffalo as1-casein is phospho- rylated at 6–8 positions [6], the caprine counterpart is phosphorylated at 9–10 positions [4], and camel as1-casein is phosphorylated at up to six serines [7]. The most striking difference in the phosphorylation pattern between human as1-casein and its ruminant counterparts is the shortage of phosphorylation in the serine-rich region consisting of residues Ser70–Glu78 in the human sequence. This region has been shown to be highly phosphorylated in all the above mentioned species. In this study, we have isolated the tryptic peptide covering residues Met68–Lys83, which contains the serine-rich region, in a nonphosphorylated form, and no traces of a phosphorylated form of this peptide were observed.

As a control experiment (data not shown) human and bovine b-casein were purified, tryptic digests were generated and these were separated by reversed-phase HPLC using the system and column described in Materials and methods. In MALDI-TOF MS analyses of the human b-casein digest, two peptides with protonated masses of 2407.85 and 2327.80 were identified. These masses correspond to the expected masses of the peptide 1–18 of human b-casein with four phosphorylations (2408.00) and three phosphoryla- tions (2328.00), respectively. Likewise in the digest of bovine b-casein, a peptide with a protonated mass of 3122.27 was observed, corresponding to the peptide 1–25 of bovine b-casein with four phosphorylations (3122.40 Da). These results indicate that the protocol used for identification of phosphorylation sites is capable of handling highly phos- phorylated peptides. Furthermore, the methods used in the present study have previously been used for identification of phosphorylation sites in several milk proteins in our laboratory, most prominently the 28 phosphorylation sites in bovine milk osteopontin [16]. Therefore it is not likely that the lack of identification of a highly phosphorylated peptide in as1-casein is due to limitations of the techniques used.

The lack of phosphorylation at other positions reported to be modified in ovine, caprine and bovine as1-casein (serines 41, 46, 48, 75 and 115 in all species, and Ser12 in ovine as1-casein, all numbers referring to the ruminant sequences), can simply be explained by sequence substitu- tions at these positions, leaving no hydroxyamino acids to be phosphorylated at these positions in human as1-casein. The region containing the phosphorylated residues, Ser18

Finally, MALDI-TOF MS analysis of native human as1-casein showed ions corresponding to a mass of approxi- mately 20 232 Da, which correlates well with the calculated mass of human as1-casein including two phosphate groups (20246 Da) (Fig. 4).

Fig. 4. MALDI-TOF MS of intact human as1-casein. The peak at m/z 20233 corresponds well to the calculated mass of human as1-casein including two phosphate groups (20246 Da). The peak at m/z 10121 represents the doubly protonated species (M2H+).

(cid:1) FEBS 2003

Phosphorylation of human as1-casein (Eur. J. Biochem. 270) 3655

in human as1-casein is not especially well and Ser26, conserved among the other analyzed species. The sequence containing Ser18 in the human as1-casein is part of exon 3 in the human as1-casein gene, which is not present in the ruminant species [13]. Ser26, situated in exon 5 of the human as1-casein gene, is not conserved in any other species except the wallaby, in which the phosphorylation pattern has not been determined [19].

produces multiple forms of mature caprine as1-casein. Eur. J. Biochem. 249, 1–7.

5. Ferranti, P., Malorni, A., Nitti, G., Laezza, P., Pizzano, R., Chi- anese, L. & Addeo, F. (1995) Primary structure of ovine as1- caseins: localization of phosphorylation sites and characterization of genetic variants A, C and D*. J. Dairy Res. 62, 281–296. 6. Ferranti, P., Scaloni, A., Caira, S., Chianese, L., Malorni, A. & Addeo, F. (1998) The primary structure of water buffalo as1- and b-casein: characterization of a novel b-variant. J. Protein Chem. 17, 835–844.

7. Kappeler, S., Farah, Z. & Puhan, Z. (1998) Sequence analysis of Camelus dromedarius milk caseins. J. Dairy Res. 65, 209–222. 8. Mercier, J.C. (1981) Phosphorylation of caseins, present evidence for an amino acid triplet code posttranslationally recognized by specific kinases. Biochimie (Paris) 63, 1–17.

9. Lasa-Benito, M., Marin, O., Meggio, F. & Pinna, L.A. (1996) Golgi apparatus mammary gland casein kinase: monitoring by a specific peptide substrate and definition of specificity determi- nants. FEBS Lett. 382, 149–152.

10. Greenberg, R., Groves, M.L. & Dower, H.J. (1984) Human beta- casein: amino acid sequence and identification of phosphorylation sites. J. Biol. Chem. 259, 5132–5138.

During the review of these results, we were puzzled by the lack of phosphorylation in the conserved region Ser70– Glu78, which is so extensively phosphorylated in the ruminant species. To test whether our results were repre- sentative, milk from three different women was analyzed. as1-Casein was purified and the reversed-phase traces of tryptic digests of the protein were compared and found to be identical in all cases, thereby showing similar phosphoryla- tion of the protein in different individuals. The phosphory- lation pattern of human as1-casein described here, and especially the lack of phosphorylations in the region Ser70– Glu78, is therefore unlikely to be a result of intra-species post-translational polymorphism in the protein. However, it should be emphasized that it is more difficult to show the absence of a modification convincingly than its presence; hence the existence of minor species partially phosphory- lated at the region discussed can not be entirely excluded.

11. Cavaletto, M., Cantisani, A., Gluffrida, G., Napolitano, L. & Conti, A. (1994) Human as1-casein like protein: purification and N-terminal sequence determination. Biol. Chem. Hoppe-Seyler 375, 149–151.

12. Rasmussen, L.K., Due, H.A. & Petersen, T.E. (1995) Human as1- casein: purification and characterization. Comp. Biochem. Physiol. 111B, 75–81.

13. Johnsen, L.B., Rasmussen, L.K., Petersen, T.E. & Berglund, L. (1995) Characterization of three types of human as1-casein mRNA transcripts. Biochem. J. 309, 237–242.

14. Martin, P., Brignon, G., Furet, J.P. & Leroux, C. (1996) The gene encoding as1-casein is expressed in human mammary epithelial cells during lactation. Lait 76, 523–535.

15. Meyer, H.E., Hoffmann-Posorske, E., Korte, H. & Heilmyer, M.G. Jr (1986) Sequence analysis of phosphoserine-containing peptides: modification for picomolar sensitivity. FEBS Lett. 204, 61–66.

The deletion of 11 amino acids at positions 59–69 and of 37 amino acids at positions 59–95 in caprine as1-casein leads to the variants D and F. In both cases these deletions, which start at the same position of the polypeptide chain, include the major phosphorylation site of the protein [20]. In ruminant milk, as1-casein, as well as the other three caseins as2-, b- and j-casein is present in micellar structures responsible for the calcium transport to the neonates. Compared with ruminant milk, the milk of primates holds a much lower concentration of calcium and a function of as1- casein in calcium transport in human milk is not likely. Recent studies of caprine as1-casein suggest that the protein interacts with the other caseins in the rough endoplasmic reticulum and that the formation of this complex is required for their efficient export to the Golgi apparatus [21]. Whether a similar scenario exists in the human system remains to be elucidated.

16. Sørensen, E.S., Højrup, P. & Petersen, T.E. (1995) Posttransla- identification of tional modifications of bovine osteopontin: twenty-eight phosphorylation and three O-glycosylation sites. Protein Sci. 4, 2040–2049.

17. Sørensen, E.S. & Petersen, T.E. (1994) Identification of two phosphorylation motifs in bovine osteopontin. Biochem. Biophys. Res. Commun. 198, 200–205.

Acknowledgments

18. Rasmussen, L.K., Sørensen, E.S., Petersen, T.E., Nielsen, N.C. & Thomsen, J.K. (1997) Characterization of phosphate sites in native ovine, caprine and bovine casein micelles and their case- inomacropeptides: a solid-state 31P NMR and sequence and mass spectrometric study. J. Dairy Sci. 80, 607–614. Special thanks to H. Breinholt and K.-E. Højbjerg, Department of Obstetrics and Gynaecology, University Hospital of Aarhus, for providing the individual milk samples.

References

19. Ginger, M.R., Piotte, C.P., Otter, D.E. & Grigor, M.R. (1999) Identification, characterisation and cDNA cloning of two caseins from the common brushtail possum (Trichosurus vulpecula) 1. Biochim. Biophys. Acta 1427, 92–104. 1. Jennes, R. & Holt, C. (1987) Casein and lactose in milk of 31 species are negatively correlated. Experentia 43, 1015–1018.

2. Kunz, C. & Lo¨ nnerdal, B. (1990) Casein and casein subunits in preterm milk, colostrum, and mature human milk. J. Pediatr. Gastroenterol. Nutr. 10, 454–461. 20. Brignon, G., Mahe, M.F., Ribadeau-Dumas, B., Mercier. J.C. & Grosclaude, F. (1990) Two of the three genetic variants of goat alpha s1-casein which are synthesized at a reduced level have an internal deletion possibly due to altered RNA splicing. Eur. J. Biochem. 193, 237–241.

3. Mercier, J.C., Grosclaude, F. & Ribadeau-Dumas, B. (1971) Structure primaire de la case´ ine as1-bovine. Eur. J. Biochem. 23, 41–51. 21. Chanat, E., Martin, P. & Ollivier-Bousquet, M. (1999) as1-casein is required for the efficient transport of b- and j-casein from the endoplasmic reticulum to the Golgi apparatus of the mammary epithelial cells. J. Cell Sci. 112, 3399–3412. 4. Ferranti, P., Addeo, F., Malorni, A., Chianese, L., Leroux, C. & Martin, P. (1997) Differential splicing of pre-messenger RNA