Aberrant Processing of Human Proislet Amyloid Polypeptide Results in Increased Amyloid Formation

The cell lines β-TC-6 (mouse), AtT-20 (mouse), GH3 (rat), GH4C1 (rat), and COS-7 (monkey) were obtained from American Type Culture Collection (Manassas, VA) and cultured in RPMI-1640 (Sigma, Stockholm, Sweden) with 11 mmol/l glucose supplemented with 10% fetal bovine serum (Sigma), 100 IU/ml penicillin, and 100 μg/ml streptomycin in humidified air/5% CO₂ at 37°C.

Gene expression levels of PC2 and PC1/3 were compared in β-TC-6, GH3, and AtT-20 cells using semiquantitative PCR technique. mRNA was isolated from 10⁶ cells from each cell line using the QuickPrep Micro mRNA Purification Kit (Amersham Biosciences, Uppsala, Sweden). A 125-ng mRNA was used for first-strand cDNA synthesis using the First-Strand cDNA Synthesis Kit (Amersham Biosciences). For detection of prohormone convertases, PCR was performed using PC1/3 forward primer 5′-CGGGGTACCGAAGCAAACCCAAATCTC-3′ and reverse primer 5′-CTACTATTTAGCCATTTTCACCAT-3′ or PC2 forward primer 5′-CGGGGTACCACTCTGAGGCATTCTGGG-3′ and reverse primer 5′-CTACTAACATTCTTTCTTCTCAGGCACGTTTC-3′. Succinate dehydrogenase (mSDHA) with Gene Bank accession no. AF095938 was selected as the housekeeping gene and a 559 bp fragment was amplified by PCR using the forward primer 5′-GGGAACATGGAAGAGGACAA-3′ and the reverse primer 5′-TGGGGTGGAACTGAACAAAT-3′. The selected primer sequences are specific for both mouse and rat. The PCR was performed using 5 μl of the first-strand DNA preparation and 10 μmol/l of each primer under the following conditions: denaturation 95°C for 1 min, annealing 50°C for 1 min, and elongation 72°C for 1 min for 30 or 32 cycles. The products were analyzed on 1.8% agarose gel stained with ethidium bromide and were quantified with fluorescence imaging with FL-filter Y515-Di (LAS 1000, Fuji, Japan) and Image Gauge version 4.0 (Fuji).

DNA fragments corresponding to h-preproIAPP were generated by PCR from pBluescript II KS vector containing the cDNA for h-preproIAPP (31) using the forward primer with a KpnI restriction enzyme cleavage site 5′-ATTGGTACCGGCATCCTGAAGCTGCA-3′ and the reverse primer 5′-CTACTAAAGGGGCAAGTAATTCAGTGG-3′. mRNA was isolated from β-TC-6 mouse cells and first-strand cDNA was synthesized. Of this, 200 ng was used for production of mouse preproIAPP (m-preproIAPP) DNA fragments by PCR together with the forward primer with a KpnI restriction enzyme cleavage site 5′-ATTAGGTACCATGTGCATCTCCAAACTGCC-3′ and the reverse primer 5′-CTACTATTAAACGAGTAAGAAATCCAAGGATTCCC-3′. The fragments were cloned into the KpnI site of the mammalian expression vector pcDNA-3 (Invitrogen, San Diego, CA).

For transfection, 10⁶ cells were cultured overnight on 19-mm coverslips in 100-mm cell culture plates (Falcon; Labora, Stockholm, Sweden). Immediately before transfection, the culture medium was replaced with 9 ml serum-free medium. Transfection was performed with 40 μg DNA in 175 μl of 0.01 mol/l polyethylenimine (Sigma) (32) with an average molecular weight of 25,000 Da (pH 7.0) and 5% sucrose. Six hours after transfection, 1 ml FBS and 0.6 mmol/l selective antibiotics G418 were added. For studies of expression over time, coverslips were removed daily for 11 days and fixed in 2% paraformaldehyde in PBS for 30 min and stored at 4°C until analyzed. The transfection was repeated for a minimum of three times for each cell line.

Monoclonal antibodies (mAbs) specific for PC1/3 and PC2 were produced using recombinant expressed proteins as immunogen. mRNA was isolated from AtT-20 mouse pituitary cells and first-strand cDNA was synthesized. PCR was performed using the PC1/3-specific forward primer 5′-GGAAGCAAACCCAAATCTCACC-3′ and PC1/3 reverse primer 5′-CTACTAACATTCTTTCTTCTCAGGCACGTTTC-3′. When expressed, this product results in a 71–amino acid fragment corresponding to the amino acids 397–467 of PC1/3. mRNA was isolated from GH3 rat pituitary cells, and first-strand cDNA was amplified with the PC2-specific forward primer 5′-GACTCTGAGGCATTCTGGGACATC-3′ and reverse primer 5′-CTACTATTTAGCCATTTTCACCAT-3′. Expression of the amplified fragment produces an 81–amino acid product that corresponds to amino acids 376–456 of PC2. The DNA fragments were cloned into the pGEX KG vector (Amersham Biosciences). PC1/3 and PC2 constructs were transformed into Y1090 E. coli and grown under agitation (230 rpm) in +37°C until OD₆₀₀ reached 0.8. Protein expression was induced by the addition of isopropyl β-d-1-thiogalactopyranoside (Amersham Bioscience) to a final concentration of 3 mmol/l. Recombinant peptides were purified from the lysed bacteria on SDS-PAGE (15%) and extracted from gel slices with 0.15 mol/l sodium chloride. These GST-tagged peptides (50 μg) were mixed with 0.5 ml Freund’s adjuvant and used for immunization. Hybridomas were produced as described earlier (33). Isotypes of the antibodies to IgG2a κ for PC1/3 and IgM κ for PC2 were determined using the Serotec mouse monoclonal isotyping kit (Serotec, Oxford, U.K.). Specificity and lack of cross-reactivity between PC1/3 and PC2 were shown by Western blots with the use of recombinant peptides with the GST-tag and with recombinant peptides where the GST-TAG had been removed by cleavage with thrombin protease.

Rabbit antisera were raised against amino acids 20–29 of mature IAPP (A133), the intact NH₂-terminal IAPP processing site (A169), and the intact COOH-terminal IAPP processing site (A165) (Fig. 1). Antiserum A133 is specific for h-IAPP and detects proIAPP, the processing intermediates 1–48 and 12–67 of h-proIAPP, and mature IAPP. The antiserum A133 does not detect mouse/rat IAPP (34). The specificity of antisera A169 and A165 was verified with synthetic peptides corresponding to amino acids 1–9, 6–13, 46–55, and 52–67 of h-proIAPP (synthesized by Ulla Engström, Ludwig Institute, Uppsala, Sweden) and with recombinant-produced peptides corresponding to amino acids 1–48 and 12–67 of h-proIAPP. The recombinant peptides were expressed with the GST system, and the amino acid sequences of the recombinant peptides were verified with mass spectroscopy. In dot blot analysis, A169 reacted with recombinant peptide 1–48 of proIAPP and synthetic peptide 6–13 of proIAPP while A165 reacted with recombinant peptide 12–67 of proIAPP and synthetic peptide 46–55 of proIAPP. No reactivity occurred between A169 and recombinant peptide 12–67 or between A165 and recombinant peptide 1–48. Also, the antisera did not detect synthetic peptides localized outside the respective processing site (Fig. 2). Antiserum reactive to mIAPP 1–37 (A110) was used for studies of m-preproIAPP transfected cells. This antiserum also detects h-IAPP.

Cells on coverslips were fixed in 2% paraformaldehyde in PBS for 30 min. Rabbit antisera A165, A110, and A133 were used at a dilution of 1:200 and A169 was used at a dilution of 1:25. Sections were incubated overnight at 4°C. Cells evaluated for PC2 and PC1/3 expression were incubated overnight at 4°C in undiluted mAbs raised against the respective processing enzymes. Secondary antibodies were Alexa 488 conjugated goat anti-rabbit, Alexa 594 conjugated rabbit anti-mouse IgG, and Alexa 488 conjugated rabbit anti-mouse IgM (Molecular Probes, Eugene, OR). Secondary antibodies were used at 1:1,000 dilution, and the incubation was performed for 2 h at room temperature.

All antibody incubations and washing steps were performed in the presence of 0.1% Saponin (Amersham Bioscience) in BSS-HEPES buffer (137 mmol/l NaCl, 5.36 mmol/l KCl, 1.26 mmol/l CaCl₂, 811 μmol/l MgSO₄*7H₂O, 441 μmol/l KH₂PO₄, 1.4 mmol/l Na₂HPO₄, and 1% HEPES). The nuclei were stained with 25 μg/ml propidium iodide (Sigma) containing 250 μg/ml RNAse A (Amersham Bioscience). For each expression experiment and cell line, a minimum of 5,000 cells were analyzed.

Congo red dye binds to the amyloid fibril and stains amyloid deposits specifically without detecting nonfibrillar forms of the precursor protein. Cells on coverslips were incubated with Congo red solution A (NaCl-saturated 80% ethanol with 0.01% NaOH) for 10 min and thereafter in Congo red solution B (solution A saturated with Congo red) (Sigma) and incubated for 10 min (35). Slides were rinsed in absolute alcohol followed by xylene and mounted with Mountex (Histolab, Gothenburg, Sweden). Cells stained for amyloid were viewed in polarized light.

Immunolabeled cells were examined with a Nikon eclipse E600 microscope connected to a Nikon C1 confocal unit with argon 488 and HeNe 543 lasers (Nikon, Kawasaki, Japan). Digital pictures were taken with an EZ-C1 digital camera and software version 1.0 for Nikon C1 confocal microscopy. Cells stained with Congo red were examined with confocal microscopy using an argon HeNe 543 laser.

Normalized with respect to the housekeeping gene succinate dehydrogenase, PCR (32 cycles) was used for semiquantification of mRNA expression. Whereas β-TC-6 cells expressed both PC1/3 and PC2 mRNA, AtT-20 cells expressed only the former and GH3 cells only the latter mRNA (Fig. 3A). The expression level of PC1/3 in AtT-20 was 80% of that in β-TC-6 cells; PC2 mRNA in GH3 was only 20% of that in β-TC-6 cells. There was no detectable gene expression of either processing enzyme in GH4C1 cells (Fig. 3A). Protein expression followed the pattern of gene expression for those enzymes. Immunolabeling with mAbs revealed that β-TC-6 cells contained both PC1/3 and PC2 (Fig. 3B and F), whereas AtT-20 cells (Fig. 3D and H) contained only the former and GH3 cells (Fig. 3C and G) only the latter protein. Neither protein was expressed in GH4C1 cells (Fig. 3E and I).

We analyzed the h-IAPP expression rate over 11 days in the different cell lines and found that there was a decrease of expression after 3–4 days in all cells except for the β-TC-6 cells. Therefore, we chose to study the effects of expression in cells 48 h after transfection.

The transfection efficiency with polyethylenimine varied among cell lines based on the percentage of cells immunoreactive for h-IAPP. Approximately 40–50% of COS-7 cells were transfected in different experiments, whereas only 10–25% of AtT-20, 10–15% of GH4C1 cells, and ∼5–15% of β-TC-6 and GH3 cells were transfected. The percentage of cells transfected with m-preproIAPP was analogous to that of cells transfected with the human construct (Table 1). The transfection rate of β-TC-6 cells with m-preproIAPP could not be evaluated due to the endogenous mIAPP synthesis. Despite the low level of transfection in some cell lines, the use of polyethylenimine was still an advantage in comparison with calcium phosphate transfection, Dotap liposomal transfection (Boehringer Mannheim, Mannheim, Germany), and electroporation, which were all compared in an earlier pilot study (not shown here). These three different methods either resulted in negative transfection in some of the cell lines or led to an unacceptable degree of cell death. In contrast, polyethylenimine transfection is associated with low toxicity, and only a few dead cells were observed after transfection.

Immunolabeling with antiserum A110 after transfection with m-preproIAPP gave an evenly distributed fluorescence throughout the cytoplasm of all transfected cells (Fig. 4A, C, E, G, and I). Congo red staining did not reveal any deposited amyloid.

The expression of h-IAPP in cells was shown by immunoreactivity with the human-specific A133 antiserum. In transformed β-TC-6 cells, this antiserum uniformly labeled the cytoplasm (Figs. 4B and 5A). In this cell type, the proIAPP molecule is expected to be processed to IAPP by the prohormone convertases. Thus, immunoreactivity with antibodies directed against the processing sites (antisera A169 and A165) was almost completely absent in all cells studied (Fig. 5B and C) (Table 2). Immunolabeling of processing sites reflects an ongoing event. Therefore, a weak background is not unexpected. Traces of labeling are designated <0.1% (Table 2). After staining for amyloid, no Congophilic material was found in any β-TC-6 cells (Fig. 5D).

In h-preproIAPP–transfected GH4C1 and COS-7 cells, where no processing of proIAPP is expected to occur, the h-IAPP–specific antiserum labeled large aggregates in the close vicinity of the nucleus (Figs. 4D and F, 5E and I). Both antisera detecting the NH₂- and COOH-terminal processing sites of proIAPP (A165 and A169) labeled GH4C1 and COS-7 cells in the same perinuclear regions (Fig. 5F, G, J, and K) (Table 2). The same intracellular locations that labeled with IAPP antibodies stained positive for amyloid with Congo red and revealed red fluorescent signal when analyzed with HeNe 543 lasers (Fig. 5H and L). This Congophilic material exhibited a bright-green birefringence (not shown) when examined with polarized light, thus confirming the presence of amyloid.

In transfected AtT-20 cells that express PC1/3 (which removes the COOH-terminal flanking peptide), h-IAPP reactivity was observed with the A133 antiserum. Areas of immunoreactivity were observed as smaller aggregates in close contact with the nucleus (Figs. 4H and 5M). In some AtT-20 cells, the same cellular h-IAPP immunoreactive locations stained positive for amyloid with Congo red (Fig. 5P). Specific reactivity was also found with antiserum A169 (detecting the NH₂-terminal processing site) (Fig. 5N) (Table 2) while reactivity against the COOH-terminal processing site was absent (Fig. 5O) (Table 2).

The level of PC2 expression in GH3 cell was determined to be 20% of the level detected in β-TC-6 cells. Despite this lower level of expression, we could not identify any reactivity with antibody A169, which recognizes the unprocessed NH₂-terminal processing site (Fig. 5R) (Table 2). In these GH3 cells, proteolytic cleavage also occurred at the COOH-terminal processing site because no reactivity with antiserum A165 was seen (Fig. 5S) (Table 2). We have also used a rabbit antibody raised against the COOH-terminal flanking peptide of proIAPP without recognizing IAPP 1–37. When used on transfected GH3 cells, this antibody revealed a more even cytoplasmic labeling pattern (not shown) different from the A133 labeling. This dissimilarity in labeling between the two antisera shows that the COOH-terminal flanking peptide had been removed from the proIAPP. This processing pattern mirrors the situation in the PC1/3 null mice, in which PC2 by itself has the capacity to process proIAPP completely (15). The transfection yield of GH3 cells was at an average of 5% in the different transfections, and up to 25% of these transfected GH3 cells contained aggregates immunoreactive to A133 as seen in Fig. 5Q. These aggregates did not stain for amyloid with Congo red and therefore are not amyloid (Fig. 5T).

Amyloid deposits selectively stained by Congo red elicit green birefringence when examined with polarized light and red fluorescence when studied with ultraviolet light (36). Methodologically, it is not possible to evaluate immunofluorescence and Congo red fluorescence in the same cells. Instead, we compared the pattern and localization of the aggregates with two different techniques on separate cells. In GH4C1 cells, which are rat pituitary cells without PC1/3 and PC2, many of the amyloid aggregates revealed a ring-shaped structure located in close contact to the cell nucleus. Virtually identical localization of amyloid with Congo red and h-IAPP with A133 was also seen in COS-7 cells and in AtT-20 cells. Transfection of proIAPP into COS-1 cells, a cell line in which PC processing enzymes are absent, has previously been reported to be associated with intracellular amyloid formation (37). A subsequent ultrastructural study of amyloid localization in COS-1 cells revealed the endoplasmic reticulum to be an amyloid storage compartment (38). In our present study, the percentage of amyloid-laden GH4C1 and AtT-20 cells constituted 3% of the total number of transfected cells at 48-h posttransfection. The percentage of amyloid-affected COS-7 cells was higher, and in this cell line, amyloid was found in 5% of the transfected cells on average. We do not know whether this reflects differences in transfection efficiency or is a result of detachment of less viable or dead amyloid-containing cells during the extensive rinsing procedures used during immunolabeling. Our pilot study showed that there was a rapid decrease of transfected cells after 3 days, and after 4–5 days (dependent of the transfection method) no h-IAPP–positive cells could be found. This also supports our negative experience in establishing stable transfectants from proIAPP expressing AtT-20, GH4C1, and COS-7 cells. While there were marked differences in the efficiency of preproIAPP transfection among cell types, we do not believe this accounted for the differences in amyloid formation observed in the different cell lines. Amyloid was detected in COS-7, GH4C1, and AtT-20 cells that had hpreproIAPP transfection efficiencies of 40, 6.3, and 25% respectively. Thus, amyloid formation in this system appears not to be purely a result of the magnitude of h-preproIAPP expression.

In this study, we have examined the role of proIAPP processing in the development of IAPP-derived amyloid. Our results show that expression of h-proIAPP in AtT-20, GH4C1, and COS-7 cells, where processing of proIAPP into IAPP is incomplete or absent, can lead to rapid formation of intracellular amyloid. In contrast, expression of human proIAPP in β-TC-6 cells derived from β-cells of murine pancreatic islets, where complete processing of proIAPP into IAPP is expected to occur, does not result in amyloid formation within the studied time limit. In the murine β-TC-6 cells and GH3 cells, both capable of processing proIAPP into IAPP, amyloid was absent irrespective of the choice of transfection method. Previously, proIAPP has been synthesized and shown to assemble into amyloid-like fibrils in water solutions (39). However, when the amyloidogenicity of IAPP was compared with that of proIAPP, the latter was found to be less amyloidogenic. The reason for this discrepancy is not known, but the intracellular environment may be important.

In contrast to h-IAPP, mIAPP is not amyloidogenic. As expected, no amyloid formation occurred in any of the studied cell lines after transfection with m-preproIAPP. It is interesting to note that the percentage of GH4C1, AtT-20, and COS cells expressing h-IAPP, i.e., cells in which amyloid deposits occurred, decreased more rapidly than those expressing mIAPP. It is reasonable to believe that this difference depends on toxic effects of aggregated (pro)IAPP.

In vitro, h-IAPP 1–37 has an intrinsic capacity to assemble into amyloid-like fibrils, and h-IAPP has been said to be one of the most amyloidogenic peptides known. This amyloidogenic property of IAPP does also exist in highly chaotropic conditions such as in the presence of 5 mol/l guanidine hydrochloride (pH 5) (P. Westermark, unpublished result). Therefore, it is remarkable that h-IAPP does not form amyloid in normal human islets. This has led to the suggestion that there are natural inhibitors present in β-cells. In vitro studies on IAPP 1–37 fibrillogenesis have shown that insulin is a potent inhibitor for amyloid-like fibril formation (40–43). This inhibitory effect is concentration dependent and independent of IAPP concentrations. IAPP has binding sites for insulin (43), and the lack of fibril formation could depend on the occurrence of complexes between IAPP and insulin. Amyloid does not occur after expression of proIAPP in β-TC-6 cells in which insulin is produced, and thus could act as an inhibitor in these cells. In a thioflavin T assay used for studies on fibril formation, we showed that mIAPP can act as an inhibitor of fibril formation by h-IAPP (23). However, neither insulin nor mIAPP is present in the pituitary-derived GH3 cells and yet, no amyloid occurs. However, these cells do synthesize growth hormone and prolactin. These latter polypeptide hormones would be expected to be present in the same compartments as IAPP, but we do not know if these hormones can act as inhibitors for h-IAPP fibril formation.

Human islet amyloid has mainly been studied in pancreata recovered at autopsy. In this material, where amyloid has accumulated progressively over an extended time period, amyloid is observed extracellularly (44). However, intracellular amyloid has been described to occur in insulinoma (26), in human islets transplanted into nude mice (25), and in cultured islets isolated from h-IAPP transgenic strains (45). In our transgenic mice overexpressing h-IAPP, amyloid was found intracellularly in a few cells, but the major deposits were present extracellularly. The present study indicates that the first amyloid formed intracellularly may consist of incompletely processed proIAPP. In biochemical analyses of IAPP amyloid, only completely processed IAPP was found (7, 8, 46).

In vitro studies on fibril formation show that this process is preceded by a lag phase that can be dramatically shortened by the addition of preformed fibrils (so-called niduses) (47). In a poorly understood way, these act as a template for further fibril formation. Also amyloid fibrils formed by other peptides may have the capacity to trigger fibril formation (48). If the nidus theory is correct, once the amyloidogenic process is initiated, it will be a self-driven process. The result of our present study supports the possibility that aberrant processing of proIAPP may lead to the initial nidus for islet amyloid formation by promoting aggregation of proIAPP or its intermediates.

In conclusion, our study indicates that insufficient processing of proIAPP to mature IAPP may be important in initial islet amyloid formation. It must be noted, however, that other factors influence that event because alterations in PC2 and/or PC1/3 by themselves did not produce amyloid in a large percentage of cells transfected with h-proIAPP. One such factor is oxidative stress, which is a product of chronically elevated glucose (49). Islets of type 2 diabetic individuals do, in fact, display increased levels of protein markers for oxidative damage to DNA and phospholipids, as well as decreased levels of the antioxidant enzyme Cu, Zn-superoxide dismutase (50). Increased oxidation of membrane lipids increases their negative charge density (51), which greatly accelerates islet amyloid fiber formation (i.e., fibrillogenesis) in vitro by enhancing the binding of h-IAPP to lipid membranes (52). As proposed by Knight and Miranker (52), h-IAPP molecules aggregate on the lipid membrane of insulin secretory granules, forming protofibrillar β-sheets, an end of which can dissociate from the membrane and twist to form a protofibril. Knight and Miranker also argue that higher concentrations of h-IAPP accelerate fibrillogenesis under conditions of elevated negative lipid charge density. Since more complete h-proIAPP processing would raise the concentration of h-IAPP, such processing might be expected to promote fibrillogenesis. In the unstressed conditions of the present study, however, absence of h-IAPP with presumably higher levels of precursor molecules promoted fibrillogenesis. The contribution of prohormone convertases to islet amyloid fibrillogenesis may therefore, to some extent, be dependent on oxidative stress levels. Further studies are needed to evaluate that possibility.

This work was supported by the Swedish Research Council (project no. 14040-03A), the Novo Nordic Insulin Foundation, the Ollie and Elof Ericsson Foundation, the Magnus Bergvall Foundation, and the Östergötland County Council Medical Research Foundation.

We thank Kenneth H. Johnson and Per Westermark for fruitful discussions.