Mink Serum Amyloid A Protein* EXPRESSION AND PRIMARY STRUCTURE BASED ON cDNA SEQUENCES

The nucleotide sequences of two mink serum amyloid A (SAA) cDNA clones have been analyzed, one (SAAl) 776 base pairs long and the other (SAA%) 552 base pairs long. Significant differences were discovered when derived amino acid sequences were compared with data for apoSAA isolated from high density lipo-protein. Previous studies of mink protein SAA and amyloid protein A (AA) suggest that only one SAA isotype is amyloidogenic. The cDNA clone for SAA2 the “amyloid prone” isotype while SAA1 is only in serum. (32) using RNase H and DNA polymerase I. The products of the first and second strand reactions were analyzed by alkaline-agarose gel electrophoresis (33). After blunt end formation using T4 DNA polymerase, methylation of internal EcoRI sites, addition of EcoRI linker adapters, and subsequent EcoRI cleavage, the cDNA inserts were ligated into h ZAP11 arms. After packaging and plating, transformant plaques (4.4 X lo”) were screened with a Syrian hamster SAA cDNA insert, phSAA (23), and radiolabeled by nick translation as above. Positive hybridization signals were visual- ized by autoradiography. Selected clones were plaque-purified, and in uiuo excision of pBS SK (-) phagemids was achieved by the addition of R408 helper phage. Plasmid DNA was isolated (34, 35) prior to analysis by restriction endonuclease cleavage and nucleotide sequencing. Analysis of Mink SAA Clones-Nucleotide sequencing was per- formed using the dideoxy-chain termination method (36). DNA se- quences were analyzed using the Geneus computer program (version 6.0; Genetics Computer Group, Madison, WI). Prediction of the helical content of the derived amino acid sequences was done with Chou-Fasman analysis (37), and hydropathy was predicted by the method of Kyte and Doolittle (38).

The nucleotide sequences of two mink serum amyloid A (SAA) cDNA clones have been analyzed, one (SAAl) 776 base pairs long and the other (SAA%) 552 base pairs long. Significant differences were discovered when derived amino acid sequences were compared with data for apoSAA isolated from high density lipoprotein.
Previous studies of mink protein SAA and amyloid protein A (AA) suggest that only one SAA isotype is amyloidogenic. The cDNA clone for SAA2 defines the "amyloid prone" isotype while SAA1 is found only in serum. Mink SAA1 has alanine in position 10, isoleucine in positions 24, 67, and 71, lysine in position 27, and proline in position 105. Residue 10 in mink SAA2 is valine while arginine and asparagine are at positions 24 and 27, respectively, all characteristics of protein AA isolated from mink amyloid fibrils. Mink SAA2 also has valine in position 67, phenylalanine in position 71, and amino acid 105 is serine. It remains unknown why these six amino acid substitutions render SAA2 more amyloidogenic than SAAl. Eighteen hours after lipopolysaccharide stimulation, mink SAA mRNA is abundant in liver with relatively minor accumulations in brain and lung. Genes encoding both SAA isotypes are expressed in all three organs while no SAA mRNA was detectable in amyloid prone organs, including spleen and intestine, indicating that deposition of AA from locally synthesized SAA is unlikely.
A third mRNA species (2.2 kilobases) was identified and hybridizes with cDNA probes for mink SAA1 and SAAB. In addition to a major primary translation product (molecular mass 14,400 Da) an additional product with molecular mass 28,000 Da was immunoprecipitable.
Serum amyloid A (SAA)' was first recognized in serum * This work was supported by the Norwegian Research Council for Science and the Humanities. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. because of its cross-reactivity with antisera to the amyloid AA protein isolated from deposits of secondary amyloid (1, 2), and a precursor-product relationship between these two proteins has been established (3). SAA is a sensitive acute phase protein (4) found in circulation bound to HDL (apo-SAA) (5). Although the concentration during acute inflammation can reach 1 g/liter and the amount of SAA in the HDL fraction can reach 40% or more of total apoproteins (6), its function remains unclear. Interleukin-1, interleukin-6, and tumor necrosing factor released from activated macrophages during inflammation stimulate hepatic SAA production (7-9). Administration of LPS to mice produces a 2000-fold increase in the level of hepatic SAA mRNA (10) indicating that the regulation of SAA gene expression occurs at a pretranslational level. Recent studies indicate that SAA gene transcription increases during inflammation and is probably due to de nouo initiation of RNA polymerase II activity rather than enhanced elongation of nascent SAA primary transcripts (10). Heterogeneity of amino acid sequences in SAA proteins may confer properties which result in deposition of some SAA isotypes as AA protein in amyloidosis. Only SAA2 has been demonstrated in murine amyloidosis suggesting that various murine SAA isotypes may have different amyloidogenic potential, and the structure and expression of SAA genes have been most extensively studied in the mouse in which SAA is encoded by three non-allelic genes and a pseudogene (11,12). The amino acid sequences derived from murine SAA1 and SAA2 genes show close homology to two murine SAA proteins (13). A polypeptide product corresponding to the open reading frame of the murine SAA3 gene has not been identified despite the fact that this gene contributes 30% of the SAA mRNA transcript pool (10).
In man, several SAA genes have been characterized by nucleotide sequence analysis (14-16), but, in contrast to murine models, no obvious amyloidogenic human SAA sequences have been detected (17,18). However, in patients with amyloidosis and familial Mediterranean fever, the AA protein consists of 76 amino acids and has threonine at residue 69 (19), whereas all AA proteins characterized by sequence analysis have phenylalanine in this position. The latter is also the case with SAA from two single human individuals (18) as well as a pool of human sera (17). This phenylalanine/threonine substitution in familial Mediterranean fever amyloid arises from a complete codon substitution, as demonstrated by the structure of a corresponding human gene (16), and may be important in defining the amyloidogenicity of the altered SAA in familial Mediterranean fever. Of interest also is that the amino acid sequence deduced from the structure of one of the human SAA genes differs from any human AA protein so far studied (14). This could be another example of a non-amyloidogenic SAA gene.
Mink are susceptible to the development of AA amyloidosis, and the observation that only valine occurs in amino acid position 10 of mink AA (20) while mink apoSAA contains isoleucine or valine at that position (21) has supported the postulate that certain SAA isotypes may be amyloidogenic. Several other amino acid positions in mink SAA demonstrate heterogeneity, but the amino acid sequence of the amyloidogenie isotype has not been known for this animal. As expression of SAA genes has been described in tissues outside the liver in other species (22,23), it was also of importance to look for extrahepatic expression of amyloid prone SAA.

EXPERIMENTAL PROCEDURES
Materials-LPS was purchased from Difco. Molecular biological enzymes and Freund's adjuvant were obtained from New England Biolabs and Boehringer Mannheim. The cell-free translation system was purchased from Promega Biotec. The cDNA synthesis kit was supplied by Bethesda Research Laboratories. X ZAP11 arms and Gigapack Gold packaging kits were purchased from Stratagene. Sequenase from United States Biochemical Corp. was used for nucleotide sequence reactions, and oligonucleotides were synthesized using an Applied Biosystems 360B synthesizer.
Induction of SAA nRNA-Two mink (Must& uision) were injected subcutaneously with 1 mg/kg LPS (Escherichia coli 026:B6), while an equal volume of saline was administered to one control animal. All three animals were put to death after 18 h. Brain, lungs, heart, liver, spleen, and intestine were promptly frozen in liquid nitrogen and stored at -70 "C.
Antiserum fo Minh SAA-Protein SAA was isolated from mink serum after stimulation with LPS as described (29). Purified SAA (0.5 mg) in complete Freund's adjuvant was injected intradermally in New Zealand White rabbits, and booster injections (0.25 mg of SAA in incomplete Freund's adjuvant) were administered every 2 weeks. The animals were bled after 6 weeks, and the antiserum was absorbed with normal mink serum to a final concentration of 20% and with dry lyophilized liver powder from normal mink.
Analysis of Primary Translation Products-RNA aliquots were translated in a rabbit reticulocyte lysate cell-free system in the presence of [""Slmethionine (1072 Ci/mmol) and amino acids (30). The radiolabeled translation products were incubated overnight at 4 "C with rabbit anti-mink SAA antiserum. A 10% suspension of protein A in 5% bovine serum albumin, 0.5% deoxycholate, 1% Triton X-100, and 1% SDS was added to each translation and incubated at 4 "C for 60 min. After centrifugation pellets were washed extensively in 1% SDS, 1% Triton X-100, and 0.5% deoxycholate. Dissociation of protein-antibody-protein A complexes was achieved by boiling for 5 min in SDS samole buffer. After centrifuaation for 5 min at 12,000 x g the supernatants were removed and analyzed by SDS-polyacrylamide gel electrophoresis (31). Gels were stained with Coomassie Blue, destained, soaked in ENHANCE, and exposed on Kodak XAR film at -70 "C.
cDNA Library Construction and Screening-An acute phase mink liver cDNA library was constructed from poly(A+) RNA isolated from LPS-stimulated mink using cloned murine mammary tumor virus reverse transcriptase for first strand cDNA synthesis. The second strand DNA was synthesized using a modification of the method of GubIer and Hoffman (32) using RNase H and DNA polymerase I. The products of the first and second strand reactions were analyzed by alkaline-agarose gel electrophoresis (33). After blunt end formation using T4 DNA polymerase, methylation of internal EcoRI sites, addition of EcoRI linker adapters, and subsequent EcoRI cleavage, the cDNA inserts were ligated into h ZAP11 arms. After packaging and plating, transformant plaques (4.4 X lo") were screened with a Syrian hamster SAA cDNA insert, phSAA (23), and radiolabeled by nick translation as above. Positive hybridization signals were visualized by autoradiography. Selected clones were plaque-purified, and in uiuo excision of pBS SK (-) phagemids was achieved by the addition of R408 helper phage. Plasmid DNA was isolated (34, 35) prior to analysis by restriction endonuclease cleavage and nucleotide sequencing.
Analysis of Mink SAA Clones-Nucleotide sequencing was performed using the dideoxy-chain termination method (36). DNA sequences were analyzed using the Geneus computer program (version 6.0; Genetics Computer Group, Madison, WI). Prediction of the helical content of the derived amino acid sequences was done with Chou-Fasman analysis (37), and hydropathy was predicted by the method of Kyte and Doolittle (38).

RESULTS
Identification and Characterization of Mink SAA-specific cDNA Clones-When 4.4 x lo5 recombinants from the mink hepatic cDNA library were screened with the cross-reacting hamster SAA cDNA probe, 152 positive clones were identified. After phagemid release, 15 clones were selected for further analysis. Mapping of inserts with 19 restriction endonucleases indicated that the clones could be separated into two groups, each containing inserts of different sizes. The largest clone in each group, 776 base pairs (pmiSAA1) and 552 base pairs (pmiSAA2), respectively, were chosen for sequence analysis. 95% of pmiSAA1 and 100% of pmiSAA2 were determined from both strands. Sequence strategy and relevant restriction map data in the coding region are shown in Fig. 1. The nucleotide sequence and deduced amino acid sequences for mink SAA1 and SAA2 are shown in Fig. 2. The nomenclature for mink SAA1 and SAA2 was adopted according to the convention established in the murine system where SAA2 is amyloidogenic and SAA1 is not. For comparison, the published polypeptide sequence for mink SAA and AA (21,20) is also included (Fig. 2). The longest clone, pmiSAA1, has a 5'untranslated region of 28 nucleotides, a signal peptide region corresponding to 19 amino acids, a coding sequence for a mature SAA polypeptide containing 111 amino acids, and a 3'untranslated region of 361 base pairs. The derived polypeptide sequence for mink SAA1 has alanine in position 10, isoleucine in positions 24, 67, and 71, lysine in position 27, and proline in position 105. The predicted amino acid sequence for mink SAA2 differs from that of SAA1 by substitutions of valine in position 10, arginine in position 24, and asparagine at residue 27, all of which are characteristic of protein AA isolated from mink amyloid deposits (20) (Fig. 2). Mink SAA2 also has valine in position 67, phenylalanine in position 71, and serine in position 105. There are four additional nucleotide substitutions between SAA1 and SAA2 in the coding region, but these are at positions of codon redundancy and do not result in additional polypeptide sequence heterogeneity.
Since discrepancies in amino acid sequence data were noted when mink SAA sequences were compared with derived polypeptide sequences from pmiSAA1 and pmiSAA2, partial or complete nucleotide sequences were determined for 12 additional clones. Using synthetic primers special attention was directed to sequences in the 5' end of the clones since an isoleucine-specific codon had not been found in SAA1 or SAA2 at amino acid position 10, a residue previously reported to be isoleucine in the protein sequence (21). All additional clones were found to be shorter but com-pletely homologous to either pmiSAA1 or pmiSAA2. The 5' ends of two clones predicted isoleucine at position 24 and lysine at residue 27 and are thus probably truncated versions of pmiSAA1.
Six of the clones predicted alanine at position 10 (as in pmiSAA1) and four predicted valine (as in pmi-SAAZ). A codon for isoleucine was not found in any clone studied. Although residue number 64 in protein AA and apo-SAA is reported as glutamine (20, 21), the 10 cDNA clones studied in this region predict serine (codon TCT) in this position. The finding of isoleucine instead of phenylalanine at residue 6 could not be explained.
Expression of SAA in Different Tissues-Hybridization of mink hepatic RNA with the cross-reacting hamster SAA probe showed no detectable SAA mRNA in the sample from the control animal, while in both LPS-stimulated animals two SAA-specific signals were detected at approximately 0.6 and 0.8 kb when compared with radiolabeled HindIII-digested X DNA molecular weight standards (Fig. 3). Radiolabeled synthetic 1%mer oligonucleotides specific for mink SAA1 and SAA2 mRNA species (Fig. 2) identify the 0.8-kb mRNA as SAA1 and the 0.6-kb signal as corresponding to SAA2 (results not shown). Hybridization with radiolabeled pmiSAA1 and pmiSAA2 inserts also detected an inducible mRNA species at 2.2 kb in the liver (Fig. 4). No extrahepatic SAA mRNA was Mink Serum Amyloid detected when SAA1 or SAA2 inserts were radiolabeled by nick translation.
However, when single-stranded RNA probes were used, a small amount of SAA-specific mRNA was detected in lung and brain (Fig. 4), but no signal could be found in RNA samples isolated from heart, spleen, or intestine. The amount of mRNA corresponding to SAA1 and SAA2 in liver was found to be approximately equal while a much smaller amount was represented by the 2.2-kb band (Fig. 4). The higher molecular mass species was not detected in any other tissue studied. In the brain, accumulation of SAA1 and SAA2 mRNA appeared equal while in the lung samples SAA2specific mRNA was more abundant (Fig. 4). Using specific oligonucleotides for SAA1 and SAA2 (Fig. 2) no 2.2-kb signal could be detected in the organs studied. The sensitivity of these RNA blots using end-labeled oligonucleotides was, however, rather low.
Analysis of Primary Translation Products-Since an SAA mRNA species of 2.2 kb has not been described in any species, the translatability of mink SAA mRNA was studied by in. uitro translation, immunoprecipitation with specific mink anti-SAA antiserum, and SDS-polyacrylamide gel electrophoresis analysis. A 14,400-Da primary translation product was identified from the LPS-stimulated mink liver sample, while no such band was observed when hepatic RNA from an unstimulated mink was used. When 100 pg of total cellular RNA isolated from the LPS-stimulated mink liver was studied, an additional band at 28,000 Da was observed (Fig. 5). A third faint protein band at 50,pOO Da could not be explained as in some experiments, in contrast to the 28,000-Da band, it also was seen in samples from normal liver.

DISCUSSION
Amino acid sequence heterogeneity has been demonstrated for SAA proteins from all species studied. It is also evident that protein AA in man (18), mouse (13), horse (39), as well as in mink is more restricted in this respect. While published data for protein sequences of mink SAA have demonstrated heterogeneity at amino acid residues 10, 67, and 71 (21), no polymorphisms have been identified in mink protein AA suggesting that only one SAA isotype is the AA precursor in secondary amyloid deposits in this animal (20). Thirty-six hours after LPS stimulation 80% of mink serum SAA was characterized by isoleucine in position 10 (SAAl), and 20% had the same amino acid as AA (valine)  2) (21). We have studied 10 mink SAA cDNA clones from which information about amino acid residue 10, important for amyloidogenicity, could be derived. No codon predicting isoleucine at position 10 was identified, while four clones representing the "amyloid-prone" SAA2 (valine at position 10) were identified, and six clones encoded alanine at this position of SAAl. However, an isoleucine codon was found at amino acid residue 6 of all cDNA clones. Upon scrutiny of the amino acid composition of the BNPS-2 and the T-l fragment isolated from apoSAA (21), the prediction from the cDNA clones of alanine and valine at position 10 and isoleucine at position 6 suggests a sequence which is in better agreement with the actual amino acid compositions of these peptide fragments. This is, however, not the case for protein AA, where only one isoleucine is found in the total sequence of AA1 and none in AA2 (20). The pmiSAA1 and pmiSAA2 nucleotide sequences both encode polypeptides eight amino acids longer than the published apoSAA, and the derived amino acid sequences of mink SAA correspond in that respect to SAA from horse (39) and sheep (40), and to protein AA from duck (41).
The derivation of isoleucine at positions 67 and 71 from one cDNA clone set and of valine and phenylalanine at these respective positions in the other group has defined SAA1 and the "amyloid-prone" SAA2 primary structure for mink. The differences between SAA1 and SAA2 for positions 24,27, and 105 were not detected in apoSAA (Fig. 2) where no heterogeneity has been documented and may represent allelic variation. The six amino acid substitutions, including three in the NH,-terminal portion of the SAA molecule found in protein AA, indicate that SAA2 is the amyloidogenic isotype. Whether amyloidogenicity is related to altered binding to HDL, changed susceptibility to proteolysis, or different binding affinity to the tissue amyloid matrix is not clear. Prediction of secondary structure based upon the amino acid sequence of mink SAA1 and SAA2 did not reveal any major differences in helical content or hydrophilicity of these two polypeptides. Similar observations have been reported for murine SAA1 and SAAB, but circular dichroism studies indicate that significant structural dissimilarities between these two proteins occur when they are interacting with heparan sulfate in the presence of calcium (42). Although 18 h after LPS stimulation the hepatic mRNA species for mink SAA1 and SAA2 are equally abundant, SAA2 accounts for only 20% of apoSAA isolated from mink serum 36 h after LPS stimulation (21). This may be due to selective SAA2 removed from the circulation, as has been shown in the murine model (43), or may result from differential SAA isotype binding to HDL or differences in translation rate or post-translational modifications.
The discrepancies between mink SAA sequences obtained from protein analyses and those derived from cDNA sequences may reflect expression of a third SAA gene contributing to most of the SAA complexed to HDL 36 h after stimulation.
Such a gene may direct biosynthesis of another SAA isotype containing isoleucine at position 10 but was not evident in detailed studies of 10 cDNA clones. Anti-SAA antiserum immunoprecipitates a 28,000-Da primary translation product in addition to a major 14,400-Da species, suggesting that the 2.2-kb SAA mRNA may be the transcription product of a third SAA gene and not partially processed mRNA. Sequence analysis of cDNA clones generated from the higher molecular mass mRNA species will be necessary to resolve whether isoleucine occurs at position 10. High molecular weight SAA has not been identified in the HDL fraction although anti-AA reactivity at higher molecular weighm has been observed in gel chromatography (6) and Western blotting experiments (3). The relationship of the 28,000-Da primary translation product to SAA amyloidogenicity remains to be determined.
Our studies of expression of mink SAA1 and SAA2 18 h after stimulation by LPS show that liver is the site of most abundant SAA mRNA production. Minor amounts of SAA mRNA for both SAA1 and SAA2 were detected in lung and brain. The relative accumulation of SAA mRNA at extrahepatic sites compared with that in the liver was less in the mink than reported for the mouse (22) and hamster (23) although mink are equally susceptible to the development of amyloidosis. Mink SAA mRNA was not identified in amyloidprone organs such as spleen and intestine. These data support the observation that amyloid deposition does not arise from locally produced and degraded amyloid-prone SAA isotypes but rather from hepatically derived SAA. Amyloidosis is an important complication to chronic inflammation.
These studies of SAA in mink have detailed the amino acid sequences of SAA isotypes which are differentially involved in amyloid deposition.
In addition, these studies demonstrate a SAA-specific mRNA species which is considerably larger than those previously described. These observations suggest the importance of further studies of SAA and amyloidosis in mink.