Tmp21 and p24A, two type I proteins enriched in pancreatic microsomal membranes, are members of a protein family involved in vesicular trafficking.

We report here on the isolation, cloning, and expression of two Mr 21,000 proteins from rat pancreatic acinar cells, the rat-Tmp21 (ransembrane rotein, Mr 21,000) and the rat-p24A. Both proteins are transmembrane proteins with type I topology and share weak but significant homology to one another (23% identity). We further show the cloning and characterization of the human homologs, hum-Tmp21, which is expressed in two variants (Tmp21-I and Tmp21-II), and hum-p24A. Tmp21 proteins and p24A have highly conserved COOH-terminal tails, which contain motifs related to the endoplasmic reticulum retention and retrieval consensus sequence KKXX. The rat-p24 sequence is identical to the hamster CHOp24, a recently characterized component of coatomer-coated transport vesicles, which defines a family of proteins (called the p24 family) proposed to be involved in vesicular transport processes (Stamnes, M. A., Craighead, M. W., Hoe, M. H., Lampen, N., Geromanos, S., Tempst, P., and Rothman, J. E. (1995) Proc. Natl. Acad. Sci. U. S. A. 92, 8011-8015). Sequence alignment and structural features identify the Tmp21 protein as a new member of this p24 family. Northern analysis of various tissues indicates that the Tmp21 proteins and the p24A protein are ubiquitously expressed. The integral membrane components Tmp21 and p24A are localized in microsomal membranes, zymogen granule membranes, and the plasma membrane and are absent from the cytosol. Both p24A and Tmp21 show weak homology to the yeast protein Emp24p, which recently has been shown to be involved in secretory protein transport from the endoplasmic reticulum to the Golgi apparatus. This leads us to conclude that the receptor-like Tmp21 and p24A are involved in vesicular targeting and protein transport.

The current knowledge of vesicular transport mechanisms concerns mainly the budding, docking, and fusion of transport vesicles (reviewed by Rothman, 1994). Less is known about the sorting and accumulation of specific proteins in transport vesicles and their delivery from the ER 1 to the Golgi apparatus. Characterization of transmembrane proteins that act like re-ceptors could give new insights into the mechanisms underlying transduction of luminal information to the cytosol.
Several transmembrane proteins of the ER and the Golgi apparatus carry a short cytoplasmically exposed COOH-terminal peptide sequence with the dilysine (KKXX) motif. This motif serves as a retention and retrieval signal that brings proteins back from a sorting compartment such as the Golgi complex to the ER (Jackson et al., 1990). The mechanisms responsible for ER retrieval and retention are not well understood, but it is known that yeast and mammalian dilysinetagged ER-resident transmembrane proteins interact with the coatomer in cell lysates Lowe and Kreis, 1995). Mutations that affect the ER retention capacity of the motifs abolish binding of the coatomer . A defect in retrieval was also observed in mutants with a defect in the genes coding for the coatomer proteins ␣-COP (RET1), ␤Ј-COP (SEC27), and ␥-COP (SEC21) . From these results, it has been suggested that coatomer plays an essential role in retrograde Golgi-to-ER transport and retrieval of dilysine-tagged proteins back to the ER.
Recently, an integral membrane component of coatomer (COPI)-coated vesicles, termed p24, was characterized (Stamnes et al., 1995). This protein defines a family of proteins from plants, yeast, and mammals with integral membrane character. One of these p24 proteins, the yeast Emp24p protein (endomembrane protein precursor of M r 24,000) was shown to be involved in the sorting and/or concentration of a subset of secretory proteins in ER-derived COPII transport vesicles  that mediate ER-to-Golgi trafficking in yeast (Salama et al., 1993).
Here we report on the identification of two M r 21,000 proteins from rat pancreatic microsomal membranes and the molecular cloning of their human homologs. These M r 21,000 proteins with receptor-like structure belong to two subfamilies of the evolutionarily conserved p24 protein family. One of these proteins, p24A, is identical to the CHOp24 that defines the p24 family. The second protein, Tmp21, shows 23% identity to p24A and represents a new subfamily, to which belong two variants (Tmp21-I and Tmp21-II). The Tmp21 protein carries a KKLIE sequence at the COOH terminus, which might be recognized by the coatomer.

Preparation of Pancreatic Acinar Cells and Subcellular Fractions-
Acinar cells were isolated from rat pancreas by collagenase digestion (Streb and Schulz, 1983). Zymogen granules were prepared from isolated acinar cells on a Percoll gradient as described (Fuller et al., 1989). * This study was supported by Grants SFB 246, B14 from the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ To whom correspondence should be addressed. Plasma membranes were purified using the MgCl 2 precipitation method (Bayerdörffer et al., 1985). Microsomal membranes and the cytosol were prepared by differential centrifugation, as described previously .
Protein Purification and Sequencing-Preparative isoelectric focusing was performed using a Rotofor cell (Bio-Rad). Protein samples were supplemented with 1% Triton X-100 and 1% ampholytes pH 5-7. Separated M r 21,000/22,000 proteins of the microsomal membrane fraction were focused in the pH range 5.8 -6.1, subjected to SDS-PAGE (Laemmli, 1970) and stained with Coomassie Blue; then the bands of interest were excised. Protein sequencing was performed using Edman degradation of the whole proteins and tryptic fragments.
Antibody Production and Western Blot Analysis-Antibodies against rat-Tmp21-I and rat-p24A were produced by immunizing rabbits with synthetic peptides coupled to hemocyanin containing the amino acid sequence of the NH 2 terminus for rat-Tmp21-I (ISFHLPVNSRKC) and for rat-p24A (YFVSIDAHAEEC). For Western blot analysis, proteins were electrophoretically separated on SDS-PAGE and transferred to nitrocellulose membranes. Nitrocellulose blots were blocked with 3% milk powder in Tris-buffered saline (TBS: 10 mM Tris-HCl, pH 8.0, 150 mM NaCl) for 60 min, followed by a 90-min incubation with various antibodies diluted in TBS plus 0.2% Tween 20 as follows: 1:3000, anti-Tmp21 and anti-SSR␣; 1:2000, anti-p24A and anti-calnexin; and 1:500, anti-Rab3A/B and anti-Ras. Bound antibodies were visualized with horseradish peroxidase-conjugated goat anti-rabbit IgG using the ECL™ detection kit (Amersham Corp.) or with alkaline phosphatase coupled to goat anti-mouse or goat anti-rabbit IgG, according to the instructions of the manufacturer (Bio-Rad).
Amplification and Cloning of cDNA Probes-Degenerated oligonucleotides encoding Tmp21 and p24A peptide sequences were synthesized and used for the reverse transcription-polymerase chain reaction. Reverse transcription-polymerase chain reaction was carried out using 25 pmol of degenerated primers, 66 mM Tris-HCl, pH 8.8, 2 mM MgCl 2 , 16.6 mM (NH4) 2 SO 4 , 200 M deoxynucleotide triphosphates, 170 g/ml bovine serum albumin, 10% dimethyl sulfoxide, and 2.5 units of Taq polymerase (Life Technologies, Inc.) at different annealing temperatures (50 -55°C). As a template, we used cDNA obtained from standard reverse-transcriptase reactions (Sambrook et al., 1989) with RNA, which had been isolated from different human and rat tissues by the method of Chomczynski and Sacchi (1987). The sequences of the primers for the first amplification step of hum-Tmp21-II were: forward-1, 5Ј-ATGTCIGGTTTGTCTGGCCCICC; reverse-1, 5Ј-CAT(C/T)TC(T/ C)TC(C/T)TCTCT(T/C)TT(T/C)TTCAT; for the second amplification: forward-2, 5Ј-CCACCAGCCCGGCGCGGCCCITT(T/C)CC; reverse-2, 5Ј-CATGTAGGCAAAGTCGTTAAC. The resulting amplification products were ligated in pUC18 using the SureClone™ kit (Pharmacia), and the plasmids were recovered according to the method of Del Sal et al. (1988). DNA inserts were sequenced using the T7 Sequencing™ kit (Pharmacia), and inserts encoding Tmp21 and p24A peptide sequences were separated from pUC18 by digestion with KpnI and BamHI, followed by electrophoretic DNA separation. The resulting cDNA inserts were isolated from the agarose gels and used for cDNA screening.
cDNA Screening and DNA Sequencing-We have prepared a random and an oligo(dT)-primed cDNA library from rat pancreatic tissue in ZAPII and ZAP-Express vectors (Stratagene). In addition, we used premade ZAPII-cDNA libraries (Stratagene) from human and rat brain. About 10 6 plaques from different cDNA libraries were screened by hybridization with 32 P random-labeled partial Tmp21-cDNA and p24A probes (see "Northern Analysis") under high stringency hybridization conditions (Sambrook et al., 1989). DNA sequencing was performed with a T7 Sequencing™ kit (Pharmacia).
Data Base Searches and Sequence Analysis-Protein and nucleic acid searches were performed using the TBLASTN and the FASTA algorithm of the European Molecular Biology network service (GENIUS Network Service; Deutsches Krebsforschungszentrum, Heidelberg, Germany). DNA fragment assembly, protein hydropathy analysis, identification of heptad repeats, alignments, and all other sequence-dependent analyses were performed using the Wisconsin Sequence Analysis Package from the Genetics Computer Group.
Northern Analysis-Total RNA was prepared by the method of Chomczynski and Sacchi (1987) from different rat tissues, from the rat pancreatic acinar tumor cell line AR4 -2J, and from the human pancreatic duct cell line CAPAN. The RNA was separated by formaldehydeagarose gel electrophoresis and transferred on Hybond-N ϩ membrane (Amersham Corp.). Tmp21-and p24A-specific transcripts were visualized by hybridization with 32 P-labeled cDNA fragments (p24A transcripts with a 265-bp probe (bases 324 -589 in accession no. X92097) and Tmp21 transcripts with a 485-bp probe (bases 30 -515 in accession no. X97442)). In addition, a 900-bp probe (XhoI fragment of rat-Tmp21-I) was used containing 200 bp of the coding region and about 700 bp from the complete 3Ј-untranslated region.
Expression of rat-Tmp21-I and hum-p24A in Escherichia coli-Using Pwo-polymerase (Boehringer), the rat-Tmp21-I and the hum-p24A were amplified from NotI-linearized cDNA screening clones with primers that span the whole mature proteins (Tmp21-I-primer: forward, 5Ј-ATCTCCTTCCATCTACCC, and reverse, 5Ј-TTACTCTATCAACTTCT-TGG; hum-p24A-primer: forward, 5Ј-TATTTCGTTAGCATCGACG, and reverse, 5Ј-TTAAACAACTCTCCGGAC). The amplification products were blunt-inserted into the XmaI restriction site of the expression vector pMal-C2 (New England Biolabs) and expressed in the E. coli strain TB1. The resulting maltose-binding fusion proteins were purified using the maltose-binding affinity for maltose. The proteins of interest were cleaved from the maltose-binding protein by factor Xa and characterized by Western blot analysis. The expressed proteins were used to confirm the identity of the cloned Tmp21 and p24A cDNAs to the microsomal proteins and for affinity purification of the antibodies described above.
Triton X-114 Extraction of Microsomal Membranes-Isolated pancreatic microsomes (see above) were extracted three times with Triton X-114 as described by Bordier (1981). The resulting aqueous and Triton X-114 protein phases were analyzed by immunoblotting with the Tmp21-and p24A-specific antibodies. To determine whether Triton X-114 extraction was complete, we monitored the distribution of the secretory protein amylase, which was found preferentially in the aqueous phase.
Trypsin Digestion Assay of Microsomes-Intact microsomes were resuspended in 280 mM mannitol, 5 mM Hepes, 10 mM KCl, and 1 mM MgCl 2 , pH 8.0. The microsomes were divided into three aliquots: the "undigested input" aliquot A, the "trypsin-Triton" aliquot B, and the "trypsin, non-detergent" aliquot C. Each aliquot contained 1 mg of microsomal protein. Aliquot B was treated with 0.1% Triton X-100 to lyse the microsomes. Then, a 1/100 part of trypsin (110 units/mg; Boehringer Mannheim) was added to each 1 part of aliquots B and C and incubated for 10, 20, 30, 60, and 120 min at 37°C. Samples containing 20 g of protein taken at different time points were analyzed by Western blot analysis using the anti-Tmp21 and anti-p24A antibodies.

RESULTS
Cloning of the Human and Rat cDNAs Encoding Tmp21 and p24A-In our previous studies on intracellular vesicle transport in rat pancreatic acinar cells, we had separated M r 21,000 proteins from the microsomal membrane fraction (Zeuzem et al., 1991. We have now isolated two of these proteins by preparative isoelectric focusing in the pH range 5.8 -6.1, followed by SDS-PAGE and microsequence analysis. Two NH 2terminal and nine internal peptide sequences were obtained (see underlined sequences in Fig. 1). Using degenerated primers deduced from peptide sequences, we were able to amplify, clone, and sequence cDNAs encoding for the tryptic and the NH 2 -terminal protein sequences of both M r 21,000 proteins. Using 32 P-labeled cDNA fragments as probes, positive clones were isolated from rat pancreatic, rat brain, and human brain cDNA libraries. The amino acid sequence information obtained from the screening procedure is summarized in Fig. 1 (hum-p24A, accession no. X92098; rat-p24A, accession no. X92097; hum-Tmp21-I, accession no. X97442; and rat-Tmp21-I, accession no. X97443). All peptide sequences identified by microsequencing of the M r 21,000/22,000 protein bands were encoded within the identified open reading frames (ORF) of the rat-Tmp21-I and rat-p24A cDNA screening clones. No tissue-specific differences could be detected between the rat pancreas and the rat brain Tmp21-I and p24A.
Amplification of a Tmp21-Variant Using Reverse Transcription-Polymerase Chain Reaction Technique-Using reverse transcription-polymerase chain reaction technique with degenerated primers, we have probed different human tissues with respect to tissue-specific expression of p24A and Tmp21 cDNAs. With this approach, we were able to characterize a human isoform of Tmp21, which we called Tmp21-II ( Fig. 1; accession no. X97444). The deduced primary sequence of this Tmp21-variant II shared 86.9% identity (92.5% DNA identity) to variant I. The Tmp21-II isoform was amplified from human brain, stomach, and colon. In all of these tissues, Tmp21-variant I was also found. The complete coding region of Tmp21-II has not yet been isolated. All of the 14 Tmp21 clones isolated from the different cDNA libraries were Tmp21-variant I. We have not found any variants for p24A.
Description of the Tmp21 and p24A Structure-The Tmp21 and p24A proteins display features typical for type I transmembrane proteins localized in intracellular membranes. We termed one of these proteins Tmp21, because it is a typical transmembrane protein and has a molecular weight of M r 21,000 in SDS-PAGE. The p24A is termed in analogy to the identical hamster homolog CHOp24, which has recently been cloned by Stamnes et al. (1995). Both proteins carry a NH 2terminal signal peptide, essential for the translocation to the ER. The NH 2 terminus of the mature rat-Tmp21-I starts with ISFH . . . , whereas the rat-p24A starts with YFVS . . . (Fig. 1).
Hydropathy plots indicate that both the Tmp21 and the p24A are anchored in the membrane by a hydrophobic domain of about 21 residues with ␣-helical character localized close to the COOH terminus (Fig. 2, A and B). The major portion of the mature Tmp21 and p24A, therefore, seems to be located in the lumen of the ER.
Both Tmp21 and p24A show highly conserved protein motifs, assumed to represent important functional sequence positions. The Tmp21-I has a dilysine ER retrieval signal at its short cytoplasmically exposed COOH-terminal tail (Jackson et al., 1990; Fig. 1). The diarginine motif of the protein p24A is homolog to the KKXX motif and has been described to maintain membrane proteins in the ER (Schutze et al., 1994). In the luminally exposed part, both proteins contain a short but conserved heptad repeat motif that supports the formation of an amphipathic coiled coil structure with high probability (Lupas et al., 1991; Fig. 2, C and D).
Comparison of the Tmp21s with Homologous Proteins-The rat-Tmp21-I (accession no. X97443) and the rat/hum-p24A (ac-FIG. 1. Multiple alignment of the p24 protein family. Aligned are the deduced primary sequences of hum-and rat-p24A (which define the p24 family), the yeast proteins Emp24p and YHR110w, hum-Tmp21-I and II, rat-Tmp21-I, the Xenopus laevis (X.lae) protein X1262, dog gp25L, and the human gp25L2. Identical amino acid residues among the p24 proteins are shown in black boxes. The predicted transmembrane region is indicated in the figure. The COOH-terminal KKXX consensus motif is marked by ‫.ءء‬ Residue ϩ1 represents the first amino acid of the mature proteins; the preceding amino acids represent the signal peptides. Underlined sequences correspond to those determined by sequencing the NH 2 terminus and tryptic fragments. Sequences which have not yet been determined are indicated by (n.d.). The mammalian p24 family can now be classified into the p24, the Tmp21, and the gp25L subfamilies. Emp24p is the yeast homolog to p24A, whereas X1262 represents the Xenopus homolog to Tmp21-I.
Comparison of the Tmp21 and p24A with sequences in the SWISS-PROT and the EMBL databank, using the FASTA and TBLASTN algorithm, revealed homology to some known proteins (Fig. 1). Weak but significant homologies (24 -36% identity) were found with the gp25L protein (accession no. X53592), a mammalian protein of rough microsomes (Wada et al., 1991), the human gp25L-variant gp25L2 (accession no X90872) and the Emp24p-protein, a yeast type I transmembrane component of ER-derived COPII-coated vesicles  accession no. X67317). Recently, a protein called CHOp24 (accession no. U26264) has been cloned, which is the hamster homolog to the rat and human p24A described here (Stamnes et al., 1995). A Tmp21 homologous cDNA clone, called X1262 (accession no. X90517), was identified in Xenopus laevis to be coordinately expressed with pro-opiomelanocortin in intermediate pituitary (Holthuis et al., 1995). The deduced X1262 protein is to 72.8% identical to rat-Tmp21-I. Structural similarity to translocon-associated type I proteins led Holthuis et al. (1995) to conclude that X1262 is a constituent of a transloconassociated protein complex.
Furthermore, ORFs encoding weak Tmp21/p24A homologs were identified in yeast (accession no. U00059, deduced YHR110w protein), rice (accession no. D46508), and Arabidopsis thaliana (accession no. T46508). Several short ORFs (until now more than 30 accession numbers) from sequencing projects submitted to the EMBL data base were found to be homologous or identical to our Tmp21/p24A cDNAs. These ORFs showed several obvious reading frame jumps and dislocations of base pairs, probably due to sequencing mistakes. A remarkable cDNA clone, called S31iii125, obtained from transcription mapping of the AD3 locus, associated to chromosome locus 14q24.3 in the aggressive form of Alzheimer disease, was submitted by Sherrington et al. (1995). This cDNA sequence shows several reading frame jumps, but apart from this, it could be identical to our human Tmp21-I transcript shown here.
Northern Analysis of the rat-Tmp21 and p24A Transcripts-For Northern analysis of the Tmp21 and p24A transcripts, we used total RNA isolated from different rat tissues and labeled them as described under "Experimental Procedures." Furthermore, we isolated RNA from the rat pancreatic acinar tumor cell line AR4 -2J (Jessop and Hay, 1980) and from the Ki-rasmutated human pancreatic duct cell line CAPAN (Kyriazia et al., 1982). Duct cells and acinar cells represent the main cellular portion of the pancreas. As shown in Fig. 3, p24A is represented by one transcript of about 1.6 kb, which is ubiquitously expressed in rat tissues and in the investigated pancreatic tumor cell lines. The situation for the Tmp21 transcripts is different. Hybridization with a hum-Tmp21-I probe (bp 30 -550) identifies two transcripts of about 1.4 and 3.5 kb. When reprobing the same blot and a control blot with a probe including 200 bases of the coding region and 700 bases of the 3Ј-untranslated region of rat-Tmp21-I, we obtained the same pattern. From our oligo(dT)-primed rat pancreatic cDNA library, we know that rat-Tmp21-I has a transcript length of 1.4 kb. The nature of the 3.5-kb transcript is not yet known. Since Tmp21 is expressed in two variants (see Fig. 1), the two Tmp21 isoforms are expressed from two different genetic loci. It is possible that the 3.5-kb transcript could be due to Tmp21-II.
Distribution of Tmp21 and p24A Proteins in Different Tissues-To determine the content of Tmp21 and p24A proteins in different rat tissues, protein homogenate and microsomes were prepared and immunologically analyzed. As compared to other tissues, both proteins are present mostly in pancreatic acinar cells (Fig. 4) and are enriched in the microsomal fraction.
Subcellular Distribution of Tmp21 and p24A Proteins in Rat Pancreatic Acinar Cells-For immunological characterization of the rat pancreatic Tmp21 and p24A, antibodies were raised in rabbits using synthetic peptides that consist of the 12 NH 2terminal amino acids in both Tmp21-I and p24A. To test the specificity of the antibodies and to confirm the identity of the cDNAs encoding for the microsomal proteins, we expressed the ORF for rat-Tmp21-I and the ORF encoding hum-p24A in the E. coli pMal expression system (New England Biolabs).
As shown in Fig. 5A, lanes 1 and 3, the antibodies raised against the NH 2 terminus of rat-Tmp21 and rat-p24A reacted  Kyte and Doolittle (1982) using the program PEPPLOT (Genetics Computer Group) and coiled-coil probability according to Lupas et al. (1991) using the PEPCOIL program (Genetics Computer Group) (C and D). The hydrophobicity plot includes the signal peptide sequence, whereas the coiled-coil analysis was calculated from residue ϩ1 of the primary sequence. The structure analysis indicates the hydrophobic signal peptides at the NH 2 -terminal end of Tmp21-I and p24A and a hydrophobic domain at the COOH-terminal part of the proteins, which is long enough to form an ␣-helical transmembrane domain. The PEPCOIL analysis identifies a heptad repeat motif in both proteins, which forms coiled coils with high probability (Tmp21-I, p ϭ 0.5; p24A, p ϭ 0.95). The proposed coiled-coil forming region is located at the NH 2 -terminal site close to the transmembrane domain.
specifically with the corresponding 65-kDa fusion proteins. The recombinant Tmp21-I and p24A were cleaved from the fusion proteins by treatment with factor Xa and were specifically recognized by the corresponding antibodies (Fig. 5A, lanes 2  and 4). Both the recombinant Tmp21/p24A proteins and the Tmp21/p24A from microsomal membranes showed the same mobility in SDS-PAGE, indicating that the native proteins were not likely to be glycosylated.
To study the cellular distribution of Tmp21 and p24A, different subcellular membrane fractions and cytosol of pancreatic acinar cells were prepared as described under "Experimental Procedures." As shown in Fig. 5B (left panel, lane a), antibodies raised against Tmp21 reacted specifically with a M r 21,000 protein in the homogenate. Tmp21 was enriched by a factor of 2.3 in the ER/Golgi microsomal fraction, by 1.6 in the plasma membrane, and by 1.8 in the zymogen granule membrane fraction as compared to the homogenate (Fig. 5B, left  panel, lanes a-d). The antibodies did not react, however, with any proteins from the cytosol (Fig. 5B, left panel, lane e). The antibodies raised against p24A recognized a M r 21,400 protein in the homogenate and showed about 2-fold enrichment in microsomal membrane fractions, about 1.3-fold enrichment in plasma membrane fractions, and about 1.5-fold enrichment in zymogen granule membrane fractions (Fig. 5B, right panel,  lanes a-d). No reaction was observed in the cytosol (Fig. 5B,  right panel, lane e).
To prove the purity of the different membrane fractions, we tested for the presence of marker proteins using specific antibodies. As shown in Fig. 5C (upper row, left), the ER protein SSR␣ is about 3-fold enriched in microsomal membranes as compared to homogenate, but its content is 4-and 2.5-fold reduced in plasma membranes and zymogen granule membranes, respectively. The second ER protein tested, calnexin, is only slightly enriched in microsomal membranes, but it is reduced in plasma membranes and has not been detected in zymogen granule membranes and in the cytosol (Fig. 5C, upper The recombinant Tmp21 proteins were expressed in the pMal-C2 vector system and cleaved from the maltose binding protein by treatment of the fusion protein with factor Xa. Uncleaved and cleaved recombinant Tmp21 proteins were separated by SDS-PAGE, transferred to nitrocellulose membranes, and probed with antibodies raised against Tmp21 (lanes 1 and 2) and p24A (lanes 3 and 4)  row, right). The marker proteins for plasma membranes and zymogen granule membranes, Ras and Rab3A/B, are highly enriched in their corresponding membranes (Fig. 5C, lower row). This indicates that plasma membranes and zymogen granule membranes, respectively, do not significantly contaminate other membrane fractions. The presence of the ER proteins SSR␣ and calnexin in membrane fractions other than the microsomal membranes is due to a contamination of these other membrane fractions with microsomes. However, this contamination is very low, as shown by the reduction of the ER proteins as compared to the homogenate. Comparing the enrichment factors for Tmp21 and p24A mentioned above in plasma membranes and in zymogen granule membranes and the concomitant decrease of anti-SSR␣ and anti-calnexin in these fractions make it unlikely that the presence of both anti-Tmp21 and anti-p24A in plasma and zymogen granule membrane fractions is due to a contamination of the latter by microsomal membranes.
Topology of the Tmp21 and p24A Proteins in Microsomal Membranes-Hydropathy analysis of Tmp21 and p24A revealed a stretch of about 21 amino acids close to the COOH terminus that is sufficiently long to span the lipid bilayer. This suggests that Tmp21 and p24A are type I transmembrane proteins with a large luminal domain and a short COOH terminal stretch of about 1 kDa, which is exposed to the cytosol. Consistent with this prediction, Tmp21 and p24A partitioned into the Triton X-114 phase, whereas the secretory protein amylase was found in the aqueous phase (Fig. 6). These data support the contention that Tmp21 and p24A are integral membrane proteins.
To confirm that the short COOH-terminal stretch of both Tmp21 and p24A is exposed to the cytosolic side, intact and solubilized pancreatic microsomal vesicles were treated with trypsin. If the NH 2 -terminal part of our protein were localized in the lumen of intact microsomes and the COOH-terminal part were exposed to the cytosol, the NH 2 -terminal recognition site of our antibodies should be protected against trypsin digestion, whereas the COOH-terminal stretch of 1 kDa should be trypsinized. Fig. 7 (left panels) shows that in intact microsomes treated with trypsin, a loss of about 1 kDa from Tmp21 and p24A was observed, as visualized by Western blot analysis. However, in Triton X-100-solubilized microsomes treated with trypsin, the NH 2 -terminal antibody recognition sites were also lost (Fig. 7,  right panels).
From the hydropathy plots (Fig. 2, A and B), the Triton X-114 extraction, and the trypsin digestion assay, we conclude that Tmp21 and p24A are integral membrane proteins with type I topology.

DISCUSSION
Here we describe the molecular characterization of two ubiquitously expressed rat type I transmembrane proteins with a molecular weight of M r 21,000 (rat-Tmp21 and rat-p24A) and the molecular cloning of the human homologs (hum-Tmp21-I, hum-Tmp21-II, and hum-p24A) (Fig. 1). Both the Tmp21 and p24A proteins show weak but significant homology to one another (23% identity). This homology is mainly mediated by conserved structural features, the NH 2 -terminal signal peptide, conserved cysteines and coiled coils in the luminal domain, a transmembrane region with ␣-helical character, and a highly conserved COOH-terminal tail. The proteins were immunologically localized in microsomal membranes, isolated zymogen granule membranes, and in the plasma membrane.
The Tmp21/p24A are homologous to proteins found in mammals, yeast and plants. Stamnes et al. (1995) recently published a sequence for a protein called CHOp24, which is the hamster homolog to the here shown rat-and hum-p24A. The authors defined the p24 homologous proteins as members of the p24 family. By analogy, we classified the Tmp21-variants as a p24 subfamily. The sequence alignment (Fig. 1) of the p24 homologs indicates an old but evolutionarily conserved family of intracellular transmembrane proteins, consisting of subfamilies and variants. The mammalian p24 family can now be classified into the p24A subfamily, the Tmp21 subfamily, and the gp25L subfamily. Emp24p represents the yeast homolog of the p24 subfamily.
The p24 family has highly conserved structural elements. With respect to the neutral theory of molecular evolution (Kimura, 1983), we conclude that these conserved structural elements represent important functional sequence positions. Besides Emp24p, all p24-like proteins carry a short but highly conserved heptad repeat signal in the luminal part of the protein with a predicted propensity to form amphipathic coiled coils (Lupas et al., 1991;Oas and Endow, 1994). This heptad repeat is followed by a COOH-terminal membrane anchor. Such structures are also observed in receptor-like proteins participating in vesicle targeting processes like the soluble N-ethylmaleimide-sensitive fusion (NSF) attachment protein receptor (v-SNARE; Dascher et al., 1991) and the SNARE-like protein Sft1p (Banfield et al., 1995), as well as the lectin-like protein Emp47p . The function of the  7. Effect of trypsin treatment for different times on Tmp21 and p24A in untreated and Triton X-100-treated microsomal vesicles of rat pancreatic acinar cells. Vesicles enriched in microsomal proteins (1 mg/ml) were permeabilized with 0.1% Triton X-100 or left untreated and subsequently incubated with trypsin (10 g/ml). At indicated times after trypsin addition, the reaction was stopped by the addition of 3 ϫ concentrated SDS-PAGE sample buffer and subsequent boiling. Proteins (20 g/ml) were separated by SDS-PAGE and transferred to nitrocellulose membranes as described under "Experimental Procedures." Nitrocellulose was probed with antibodies raised against Tmp21 (upper row) and p24A (lower row). heptad repeat structure in the receptor-like p24 family is not known.
The Tmp21 and some of the homologs (Fig. 1) have a dilysine ER-localization signal at their short cytoplasmically exposed COOH-terminal tail. This motif has been identified to be a retrieval motif that brings proteins from an unknown sorting compartment back to the ER (Jackson et al., 1990). It is shown that the KKXX-mediated retrieval can occur at several sites between the intermediate compartment and the trans-Golgi network (Jackson et al., 1993;Martire et al., 1996). COOHterminal dilysine-motifs bind specifically to the protein complex coatomer Letourneur et al., 1994;Lowe and Kreis, 1995), which is essential for the retrieval of dilysine-tagged proteins to the ER. We have co-isolated the dilysine-tagged Tmp21 and p24A with ER and Golgi-enriched microsomal membranes but surprisingly also with zymogen granule membranes and PMs. In the case of the dilysine-tagged ERGIC-53, a substantial portion of ERGIC-53 is localized outside the ER (Schweizer et al., 1988). Overexpression of ER-GIC-53 results in its appearance in PMs (Kappeler et al., 1994). Furthermore, it has been shown by substitutional mutagenesis that the motif KKFF at the COOH-terminal, cytosolic tail of ERGIC-53 acts as a signal for endocytosis (Itin et al., 1995). The high expression rate of Tmp21 and p24A in rat pancreas, as compared to other tissues (Fig. 4), is remarkable and might indicate that Tmp21 and p24A are involved in the special function of this organ to synthesize, transport, and exocytose digestive proteins at high rates. This could lead to a passive flow of Tmp21 and p24A through the ER and Golgi compartments to the zymogen granule membrane and the PM. The appearance of Tmp21 and p24A in zymogen granule membranes and PMs does not presuppose a function of these proteins in granule and plasma membranes, but it makes a sorting mechanism for KKXX-containing proteins at the PM necessary. The proteins ERGIC-53 and VIP36, which carry dilysine motifs, have been, in part, localized to coated pits (Fiedler et al., 1994), supporting the idea that the KKXX-sorting at the PM is mediated by the clathrin adapter complex. The p24A proteins carry a COOH-terminal diarginine motif. Diarginine motifs are similar to dilysine motifs and have been shown to maintain type II membrane proteins in the ER (Schutze et al., 1994).
Until now, the function of the mammalian p24 members is unclear. In yeast it was shown that knocking out the p24 homolog Emp24p (in that study called yp24A) in a vesicle fusion mutant resulted in a dramatic reduction in the accumulation of transport vesicles (Stamnes et al., 1995). In another Emp24p yeast mutant (disruption deletion of the luminally located amino acids 43-118), delivery of periplasmic invertase and glycosylphosphatidylinositol-anchored plasma membrane protein (Gas1p) to the Golgi apparatus was reduced. In contrast, the transport of ␣-factor, carboxypeptidase Y or acid phosphatase, and two vacuolar proteins were unaffected in this deletion mutant . The authors proposed that Emp24p is involved in the sorting and/or concentration of a subset of secretory proteins into ER-derived COPIIcoated transport vesicles .
Taken together, we suggest that the luminal domain of the p24 protein family is able to transduce information, like the cargo packaging of the transport vesicles, to the cytosol. The cytosolic domain of the p24 proteins contains the KKXX motif, which could build up the contact to coat structures such as the coatomer and help to direct COP-coated vesicles in anterograde and/or retrograde pathways. It is striking that of all tissues tested, the highest amounts of Tmp21 and p24A have been found in the pancreas, an organ whose main task is the synthesis, packaging, transport, and exocytosis of proteins. Therefore, we speculate that the proteins of the p24 family may play a receptor like role in sorting and directing proteins within the secretory pathway.