Disparate Intracellular Processing of Human IL-12 Preprotein Subunits: Atypical Processing of the P35 Signal Peptide¹

Interleukin-12 is critically involved in the post-thymic development of Th1/Th2-type immune responses (1). It is unique among numbered cytokines in that it exists as a heterodimer of two glycosylated polypeptides (termed p35 and p40) that are encoded by genes on separate chromosomes (2). The p35 shares structural similarities with the cytokines IL-6 and G-CSF (3), whereas p40 is structurally related to the IL-6R (4). Heterodimers (p35/p40) are required for IL-12 bioactivity (5). In addition to forming heterodimers with p35, p40 forms homodimers (p40₂), which exhibit biologic activities antagonistic to those of bona fide IL-12 (6). Homodimers of p35 have not been reported to date; however, p35 may heterodimerize with a second cellular protein, EBV-induced gene 3 (7), in addition to p40. What function this additional heterodimeric form may perform in IL-12 biology is unknown at this time. We (8) and others (9) have shown that p35 expression is the limiting factor controlling production of IL-12 in APC, although, even under optimal conditions, p40₂ may predominate. In fact, recent studies suggest that the IL-12 response is a carefully orchestrated temporal balance between antagonistic p40₂ and IL-12 heterodimers, with p40₂ appearing first, followed by IL-12 (10).

To better understand the regulation of IL-12 production during immune responses, we investigated the intracellular protein processing of each of the IL-12 subunits. Signal peptides target proteins for secretion and are thought to direct transport across the endoplasmic reticulum (ER)⁴ membrane by a cotranslational mechanism (reviewed in Ref. 16). Signal peptides have a common structure comprised of a short, positively charged amino-terminal region (n-region), a central hydrophobic region of approximately 10 aa (h-region), and a more polar carboxyl-terminal region of 4–6 aa (c-region) that is typically the site of cleavage by signal peptidase (11). All the information necessary to direct processing of the signal peptide by signal peptidase is thought to be contained within the signal peptide (12). In eukaryotes, but not prokaryotes, partial glycosylation may occur concomitant with translocation. The functional activity of signal peptides may be examined in vitro using a combination of protein translation in the presence or the absence of microsomes that perform many, if not all, of the translocation and cleavage functions of the ER in vitro. Alternatively, cDNA constructs may be transfected transiently into cells in vitro and monitored for the presence of the protein encoded by the cDNA within the cell or secreted into the culture media. Both of these approaches were employed to evaluate the post-translational processing and secretion of IL-12 p35 and p40 proteins. Key aspects of the process were also tested by examining the stable processing intermediates of the native gene products in monocytes/macrophages following gene induction of IL-12. Our results indicate that while p40 follows this model, the p35 preprotein does not function as a typical secretory peptide as would be predicted by its amino acid sequence (12). Instead, processing of the p35 secretory signal peptide involves two sequential cleavages, which occur after migration of the intact preprotein into the ER. The first cleavage is located within the hydrophobic portion of the predicted signal peptide. The second cleavage occurs at a later stage, possibly coupled to secretion and complex glycosylation, and releases the amino-terminus of the secreted p35 protein. Discrimination of this alternate pathway is not confined solely to the p35 signal sequence, but appears to also be influenced by other regions of the p35 preprotein. Our results demonstrate that the IL-12 subunits are processed by disparate protein processing pathways and suggest new therapeutic intervention points for modulation of IL-12 production.

Human peripheral blood monocytes were purified from single-donor leukapheresis preparations by centrifugal counterflow elutriation. Monocytes were <95% pure by Giemsa and nonspecific esterase staining. Cells were cultured in six-well tissue culture plates (Costar, Cambridge, MA) in RPMI 1640 supplemented with l-glutamine, gentamicin sulfate, HEPES, and 10% FBS (Life Technologies, Gaithersburg, MD). All culture reagents were tested and were free of detectable endotoxin.

COS-7 cells were obtained from American Type Culture Collection (Manassas, VA) and grown in DMEM (Life Technologies) containing 10% FCS, l-glutamine (2 mM), nonessential amino acids (0.1 mM), and gentamicin (0.1 mg/ml) in an atmosphere of 6% CO₂.

The cDNAs of both human p35 and p40 in Bluescript (Stratagene, La Jolla, CA) were provided by U. Gubler (Hoffmann-La Roche, Nutley, NJ; GenBank accession no. M65271 and M65272). Two forms of the p35 open reading frame (ORF) were constructed, representing each of the two potential forms of the signal peptide fused to the ORF encoding the secreted portion of p35. The first, termed p35₋₅₇, contains two potential initiator methionines at positions −57 (M₋₅₇) and −22 (M₋₂₂) relative to the first amino acid (R) of the secreted form of p35. This form corresponds to the ORF described in lymphoblastoid cells (13). The second, termed p35₋₂₂, corresponds to ORF resulting from transcription arising from the TATA-like element described for IFN-γ/LPS-stimulated monocytes/macrophages (14). This latter form was constructed by removing nucleotides 1–156 from the p35 cDNA. Both constructs were excised from Bluescript by NotI/KpnI treatment and ligated into pcDNA3.1(+) (Invitrogen, San Diego, CA), which had been digested with NotI/KpnI. In vitro transcription was performed from the T7 promoter. The p40 cDNA was excised from Bluescript by XhoI/NotI treatment and ligated into pcDNA3.1(+) following treatment with XhoI/NotI.

Mutations within the signal peptide (Table II) were constructed using PCR-mediated site-directed mutagenesis (15) and p35₋₅₇ as template. Briefly, the left PCR product was generated with the forward priming p35₋₅₇-specific oligonucleotide gene-specific primer (GSP) 61 (5′-AGCGGTACCTTATAAAAATGTGGCCCCCT-3′; sequence complementary to bases 60–79 and containing a novel KpnI site) and the reverse GSP containing the base change for the particular mutant (see Table I). The right PCR product was generated using the corresponding forward gene-specific primer (Table I) and vector-specific primer (VSP) 59 (5′-ACGGGCCCTCTAGACTCGAGCGGCCGC-3′; complementary to bases 978-1004 of pcDNA3.1(+)). To generate the complete sequence, corresponding left and right PCR products were gel purified, added in equimolar amounts to a PCR reaction, annealed to each other, and then amplified using the external oligonucleotides, GSP61 and VSP59, and Advantage Klentaq polymerase (Clontech, Palo Alto, CA). The resultant PCR products were gel purified, digested with KpnI/NotI, and ligated into pcDNA3.1(+), which had been linearized with KpnI/NotI. This was then transfected in DH5α cells (Life Technologies), selected, and purified using Plasmid Maxi kits (Qiagen, Santa Clarita, CA). A double mutant (L_-10→R/R₊₁→G) was generated using the primers listed to generate mutant L₋₁₀→R, but the template was R₊₁→G. All DNA manipulations were confirmed by DNA sequence analysis using Sequenase version 2.0 (Amersham, Arlington Heights, IL).

Plasmids were linearized with NotI (p35) or XhoI (p40). The RNA was transcribed in vitro with T7 polymerase according to instructions supplied by the manufacturer (Stratagene, La Jolla, CA). Rabbit reticulocyte lysate and canine pancreatic microsomes were obtained from Promega (Madison, WI) and used to translate RNA in the presence of 20 μCi of [³⁵S]methionine as recommended by the manufacturer. Proteinase K treatment of translation reactions was performed at a final concentration of 0.1 mg/ml (with added 10 mM CaCl₂) for 45 min on ice with or without 0.1% Triton X-100. The reaction was terminated by addition the protease inhibitor 4-(2-aminoethyl)-benzenesulfonyl-fluoride to 10 mM.

COS-7 cells were transfected with Lipofectin reagent (Life Technologies) according to the manufacturer’s instructions, and 48 h later cells were labeled with 100 μCi of [³⁵S]methionine for the indicated time periods. Labeled cultures were washed twice in PBS and extracted with lysis buffer (150 mM NaCl, 50 mM Tris (pH 7.5), 50 mM NaF, 0.5 mM Na₄P₂O₇, and 1% Triton X-100) supplemented with the protease inhibitors 4-(2-aminoethyl)-benzenesulfonyl-fluoride (1 mM; ICN Biomedicals, Aurora, OH), aprotinin (2 μg/ml; Sigma, St. Louis, MO), leupeptin (1 μg/ml; Sigma), and pepstatin A (1 μg/ml; Sigma). In some experiments, transfected COS-7 cells were treated with the metabolic inhibitors tunicamycin (1 μg/ml; Roche, Indianapolis, IN) or benzyl- 2-acetamido-2-deoxy-α-d-galactopyranoside (2 mM, Sigma) for 24 h before and during the [³⁵S]methionine labeling period.

Human elutriated monocytes were plated at a density of 3.3 × 10⁶/ml and cultured overnight in the presence of IFN-γ (100 ng/ml; Genentech, South San Francisco, CA). The following morning LPS (1 μg/ml; catalogue no. L2755, Sigma) was added to the cultures for 4 h. Cells were then washed in PBS, placed in methionine-free culture medium containing LPS (1 μg/ml) and 100 μCi of [³⁵S]methionine, and incubated for either 4 h or overnight, at which time the cells and culture fluids were harvested.

Immunoprecipitations were performed in final volumes of 500 μl. When necessary, in vitro translation mixtures were diluted to the final volume with lysis buffer. Cellular supernatants were supplemented with the above-mentioned protease inhibitors and concentrated using Centricon concentrators (Amicon, Beverly, MA). Samples were treated first with protein G-Sepharose beads before incubation with either mouse anti-human p35 Ab mixture or mouse anti-human p40 Ab (anti-p35, catalogue no. 20501D, anti-p40, catalogue no. 20512D, isotype control, catalogue no. 03171; PharMingen, San Diego, CA). Immunocomplexes were collected with protein G-Sepharose beads (Amersham-Pharmacia, Uppsala, Sweden), heated to 95°C for 5 min, and analyzed by 12% SDS-PAGE and autoradiography.

Immunocomplexes were resuspended in 50 μl of 0.2 M phosphate buffer (pH 6.0), containing 0.2% SDS and 2% 2-ME, and heated at 95°C for 5 min, then cooled on ice. After this, Nonidet P-40 (2%) and AEBSF (10 mM) *were added, and the samples were incubated overnight at 37°C individually or in combination with the following enzymes: 20 mU endoglycosidase (Endo) H (Roche), 250 mU Endo F/N-glycosidase F-free (Roche), 3 mU of O-glycosidase (Roche), and 100 mU neuraminidase (Sigma). Samples were then subjected to SDS-PAGE analysis following treatment for 5 min at 95°C in reducing sample buffer.

We first examined processing of the p40 signal peptide and found that p40 processing conformed to the accepted model of signal peptide function (Fig. 1). In vitro translation of p40 RNA resulted in a preprotein of 42.5 kDa. Addition of microsomes to the translation reaction resulted in a size reduction to approximately 41.5 kDa, representing both cleavage of the signal peptide as well as partial glycosylation, which was confirmed by removal of carbohydrates with Endo H to yield a 39.5-kDa protein (Fig. 1,A). Thus, approximately 3 kDa was removed from the nascent p40 protein during translocation into the microsomes, consistent with removal of the entire 22-aa signal peptide (13) by signal peptidase. Transfection in COS-7 cells showed that the intracellular and secreted forms were similar in size to the 41.5-kDa form observed in vitro, suggesting that p40 was not subjected to complex glycosylation (Fig. 1,B). This was confirmed by subjecting the intracellular and secreted forms of p40 to deglycosylation with Endo H and Endo F. The initial stage of protein glycosylation within the ER results in the addition of high mannose sugars that are susceptible to removal by either Endo H or Endo F. However, once within the Golgi, additional carbohydrate modifications may occur that result in loss of susceptibility to Endo H, but not to Endo F. Both intracellular and secreted forms of p40 were sensitive to deglycosylation with Endo H, demonstrating that both contained only noncomplex carbohydrates. Secreted p40 did not undergo complex glycosylation, because the results with Endo F, which cleaves both complex and noncomplex carbohydrates, were identical with those with Endo H. When p40 transfected cells were treated with tunicamycin, an inhibitor of lipid-linked oligosaccharide formation, a single band of 39.5 kDa was immunoprecipitated from intracellular and secreted, indicating that glycosylation was not necessary for secretion (Fig. 1 C). Taken together, these studies demonstrate that the p40 signal peptide is processed in vitro and in living cells consistent with removal of the signal peptide by signal peptidase.

The p35 gene can be transcribed from one of two possible promoters: either a CpG-like element functional in EBV-transformed lymphoblastoid cells (13) or a downstream TATA-like element following LPS stimulation of IFN-γ-primed monocytes (14). Depending upon which promoter is used, the resultant ORF may contain either one or two initiator methionines (diagrammed in Fig. 2 a). Transcripts of p35 in IFN-γ/LPS-stimulated monocytes (p35₋₂₂) contain only the second initiator methionine (M₋₂₂) and are predicted to encode a typical signal peptide of 22 aa and a predicted preprotein size of 25 kDa. Lymphoblastoid cell-derived p35 mRNA (p35₋₅₇) contains an additional in-frame initiator methionine (M₋₅₇) encoding an additional 34 aa between M₋₅₇ and M₋₂₂ with a predicted preprotein size of 28 kDa. Both forms contain the same predicted signal peptide sequence between M₋₂₂ and R₊₁. The functional significance of these two potential forms of p35 is unknown at present (13).

We first determined the functionality of each of the potential initiator methionines to direct in vitro translation (Fig. 2,B). RNA transcribed from p35₋₂₂ and p35₋₅₇ cDNAs encoded proteins of either 30 kDa (p35₋₅₇) or 27 kDa (p35₋₂₂), which displayed apparent molecular sizes on SDS-PAGE approximately 2 kDa larger than the predicted theoretical sizes and suggested that the first initiator methionine encountered was used by in vitro translation. Each form also functioned similarly, in that comparable amounts of protein were produced from equivalent amounts of mRNA. In the presence of microsomes, both transcripts yielded additional products approximately 5 kDa larger. These larger species were glycosylated forms of the proteins, since they were susceptible to treatment with the glycosidase Endo H. However, upon deglycosylation the proteins were identical in size to their counterparts translated in the absence of microsomes, indicating that the signal sequences were not removed from either p35₋₅₇ or p35₋₂₂ during translocation into the microsomes (Fig. 2 B).

To determine whether failure to remove the signal peptides of p35 preproteins in vitro reflected integration into the membrane of the microsome or failure to traverse into the lumen of the microsome, we employed a proteinase K protection assay (Fig. 2 C). If p35 protein were completely translocated across the microsomal membrane, it should be protected from digestion by proteinase K by the lipid bilayer of the microsome. If p35 were a transmembrane protein, only those domains located within the lumen of the microsome vesicle should be protected. Following treatment with proteinase K, the glycosylated forms of form of p35₋₂₂ (32 kDa) and p35₋₅₇ (35 kDa) were protected from proteolysis, indicating that each form of the p35 secretory leader was able to direct translocation into the lumen of the microsome. In contrast, the unglycosylated forms of p35₋₂₂ (27 kDa) and p35₋₅₇ (30 kDa) were not protected, indicating that these forms were not translocated. This is not completely unexpected, as translocation of proteins into microsomes by the in vitro system is not quantitative. However, taken together, these results demonstrated that both forms of the p35 signal peptide were capable of directing translocation across the ER, but were not concomitantly removed during this process.

To determine whether similar events occurred in living cells, COS-7 cells were transiently transfected with cDNA constructs encoding each potential form of the preprotein, and the resultant proteins were characterized (Fig. 3). Regardless of whether cells were transfected with p35₋₅₇ or p35₋₂₂, immunoprecipitates from cell lysates contained a predominant species of 31 kDa rather than the 32-kDa (p35₋₂₂) or 35-kDa (p35₋₅₇) forms observed following in vitro translation in the presence of canine microsomes. The different size suggested that additional processing occurred within the endoplasmic reticulum regardless of the form of p35 signal peptide used (Fig. 3,A). A second band, approximately 5 kDa larger than the predominant band, appeared in immunoprecipitates from p35₋₅₇ transfected cell lysates. This 36-kDa protein had a half-life of 30 min and disappeared over a prolonged time period, whereas the 31-kDa species had a half-life of approximately 3 h regardless of whether it was encoded by p35₋₅₇ or p35₋₂₂ (data not shown). We interpret this finding as representing initiation of protein synthesis from M₋₅₇ or p35₋₅₇, which was then subsequently processed within the ER to yield the more stable 31-kDa species. However, the 31-kDa intracellular form does not represent the secreted form, because immunoprecipitated p35 protein from transfected COS-7 cell culture supernatants contained a single diffuse band of 36 kDa (Fig. 3 A) regardless of whether the cells were transfected with p35₋₅₇ or p35₋₂₂, demonstrating that additional processing events were required.

To clarify whether our findings for p35 signal peptide processing in COS-7 cells were reflective of events in human monocytes/macrophages, the primary biological source of secreted IL-12, we evaluated p35 biosynthesis under conditions shown previously to induce large amounts of bona fide IL-12 (8). Proteins from IFN-γ- and LPS-stimulated monocytes were metabolically labeled with [³⁵S]methionine and immunoprecipitated from cell lysates or culture medium supernatants with anti-p35 Ab (Fig. 3 B). The stable intracellular form of p35 in monocytes/macrophages was identical in size to the 31-kDa species observed in COS-7 cells transfected with each cDNA. Likewise the secreted form of p35 was 36 kDa. In addition, a 42.5-kDa protein corresponding in size to p40 was also observed in anti-p35 immunoprecipitates from cell lysates, suggesting that heterodimerization had occurred between the partially processed p35 preprotein and the fully processed p40. A small amount of the 31-kDa species was also observed in the supernatant at 24 h, possibly due to cell death/leakage during the prolonged metabolic labeling period. Similar results were seen in COS-7 cells cotransfected with p35 and p40 (data not shown). Taken together, these data demonstrate that the results obtained by transient transfection of cDNA in COS-7 cells were reflective of IL-12 processing in monocytes/macrophages.

Our studies suggested that the most stable form of the p35 preprotein translocated into the ER was processed to a 31-kDa form, which was subsequently secreted as a 36-kDa protein regardless of the initiation methionine used (M₋₂₂ or M₋₅₇) or the cell type assayed. To clarify what role glycosylation played in this observation we determined the effects of the deglycosylation enzymes Endo H, Endo F, O-glycosidase, and neuraminidase on intracellular and secreted forms of p35. The initial stage of protein glycosylation within the ER results in the addition of high mannose sugars that are susceptible to removal by either Endo H or Endo F. However, once within the Golgi, additional carbohydrate modifications occur that result in loss of Endo H susceptibility, but not that of Endo F. O-linked glycosylation, which is thought to also occur within the Golgi compartment, is susceptible to removal by O-glycosidase. Addition of N-acetylglucosamine, which occurs after exiting the ER, but not before, is removed by neuraminidase.

When cell lysates from COS-7 cells transfected with each p35 form were digested with either Endo H or Endo F, the stable intracellular 31-kDa form of p35 was reduced in size to 25 kDa. No difference could be discerned between the products resulting from treatment with either Endo H or Endo F (Fig. 4,A, upper panel), consistent with the presence of only high mannose sugars. In contrast, no change in the size of the stable 31-kDa intracellular form ofp35 was observed when the immunoprecipitate was digested with O-glycosidase and/or neuraminidase (data not shown). Moreover, only the 25-kDa species was observed when cocktails of these enzymes were used, and no difference was found from the use of either Endo H or Endo F alone. These results confirm that the stable intracellular 31-kDa form of p35 was subjected only to noncomplex glycosylation and, based upon the sensitivity to Endo H, revealed that this form resided within the ER and not within the Golgi. In contrast, the secreted 36-kDa form p35 was resistant to Endo H (Fig. 4,A, lower panel), suggesting that additional, complex glycosylation occurred during later stages of the secretion process. This was confirmed by treatment of secreted p35 with Endo F, which resulted in a size reduction of approximately 10 kDa (from 36 to 26.5 kDa; Fig. 4,A, lower panel). The 1.5-kDa disparity between the intracellular and secreted deglycosylated forms of p35 (25 vs 26.5 kDa, respectively) suggested that additional post-translational protein processing may occur. Neuraminidase treatment alone reduced the size of secreted p35 by approximately 3 kDa (Fig. 4,A). However, O-glycosidase treatment had no effect either by itself (data not shown) or in a mixture with Endo F and neuraminidase, as the observed reduction in size was the same as that after treatment with Endo F alone (Fig. 4 A, lower panel). These studies demonstrated that the 31-kDa intracellular form of p35 contained noncomplex carbohydrates, while the secreted form contained complex, N-linked carbohydrates. However, since the 31-kDa glycosylated species was sensitive to both Endo H and Endo F, we concluded that this stable intracellular form of the preprotein must reside within the ER, since complex glycosylated proteins become insensitive to the effects of Endo H, but not Endo F, as they migrate out of the ER and are modified by further glycosylation within the Golgi (17).

We further investigated the role of glycosylation in p35 protein processing using inhibitors of glycosylation (Fig. 4,B). Benzyl 2-acetamido-2-deoxy-α-d-galactopyranoside, an inhibitor of O-glycosylation, did not affect the size of the 31-kDa intracellular form or affect secretion, confirming the results obtained with O-glycosidase. However, treatment with tunicamycin, a metabolic inhibitor of N-glycosylation, completely blocked secretion and resulted in the appearance of 29- and 24-kDa intracellular forms for p35₋₅₇ (Fig. 4 B). In a parallel experiment only a single species of 24 kDa was observed for p35₋₂₂ (data not shown). These findings suggested that the 24-kDa species resulting from tunicamycin treatment most likely represented the unglycosylated counterpart of the 31-kDa stable intracellular form of p35, while the 29-kDa species most likely represented the unglycosylated counterpart to the transient 36-kDa species. The 1-kDa difference observed between the 25-kDa Endo F-treated intracellular product and the 24-kDa product after tunicamycin treatment may be due to the different modes of action of these agents: removal of oligosaccharide additions by Endo F vs prevention of the first step in lipid-linked oligosaccharide precursor formation by tunicamycin. Together, these results demonstrated that 1) p35 was a complex glycosylated protein; 2) complex glycosylation was an integral part of the p35 secretory process; 3) primary cleavage of the preprotein occurred regardless of the state of glycosylation; but 4) subsequent protein processing and/or secretion required glycosylation.

To date we have shown that the stable intracellular intermediate in p35 biosynthesis is a 31-kDa glycoprotein that is observed with p35₋₅₇ or p35₋₂₂ or in monocytes/macrophages following induction by IFN-γ and LPS. To exclude formally the possibility that the 31-kDa protein observed for p35₋₅₇ in COS-7 cells might be due to translation initiation arising internally at M₋₂₂, mutational analysis was performed (Table II and Fig. 5). Substituting L for M₋₂₂ (M₋₂₂→L) did not abrogate production of the 31-kDa band, indicating that the 31-kDa protein was not generated from M₋₂₂, and supported our interpretation that partial cleavage of the preprotein occurred after translocation into the ER (Fig. 5 A). Occasionally in eukaryotes, leucine may initiate protein translation; however, there is a substantial decrease in the amount of protein translated relative to that when initiation occurs at a methionine residue (18). Since we could not detect any difference in translation efficiency between p35₋₅₇ and M₋₂₂→L, we concluded that translation initiation at M₋₂₂ was not responsible for the 31-kDa preprotein.

To identify the approximate position of signal peptide cleavage within the 31-kDa form, we compared the m.w. of deglycosylated p35₋₅₇ contained in COS-7 cells with that of p35₋₂₂ translated in vitro (Fig. 5,B). Treatment of p35₋₅₇ protein with Endo H revealed a protein approximately 2 kDa smaller than the protein generated by in vitro translation from M₋₂₂ in the absence of microsomes (25 vs 27 kDa). This suggested that cleavage occurred distal to M₋₂₂, within the signal peptide between M₋₂₂ and R₊₁ (see Fig. 2,A). Complementing this result was the observation that approximately 5 kDa of leader sequence was removed from p35₋₅₇ (Fig. 2,C), whereas the predicted size difference between M₋₅₇ and M₋₂₂ was 3.4 kDa. This prediction was shown to be correct, since there was approximately a 3-kDa difference between in vitro products using either M₋₅₇ or M₋₂₂ (Fig. 2,B). Based upon these findings, we postulated that the initial cleavage site must occur within a region between 10–17 aa downstream of M₋₂₂, corresponding to the middle portion of the h-region of the signal peptide (diagrammed in Fig. 2,A). To evaluate the potential function of this region, aa −13 to −6 (V₋₁₃ATLVLLD₋₆, Δh, Table II) within the h-region of the signal peptide were deleted. The resultant p35 protein detected in COS-7 cells using anti-p35 Ab was a single band of 28 kDa (Fig. 6 A), corresponding to the expected size of the unmodified preprotein translation product (29 kDa minus approximately 1 kDa of the Δh region). This form of the p35 preprotein was consistent with the failure to translocate into the ER and undergo subsequent glycosylation to produce the 31-kDa form.

Within the Δh region, three L were conserved among 10 animal species of p35 examined. Thus, we decided to mutate each L in turn to determine its possible function in this process. Mutating L₋₁₀ to either R (L₋₁₀→R) or Q (L₋₁₀→Q) resulted in a 29-kDa protein, suggesting that the h-region of the signal peptide was involved with entry into the ER. Mutating L₋₈ (L₋₈→R) or L₋₇ (L₋₇→R) yielded a mixture of the 29-, 31-, and 36-kDa proteins (Fig. 6 A). These results suggested that mutation of L₋₁₀ abrogated entry into the ER, while mutation of either L₋₈ or L₋₇ only partially inhibited this process. Thus, the mutations Δh, L₋₁₀→R, L₋₁₀→Q, and L₋₁₀→R/R₊₁→G appeared to disrupt the ability of the p35 peptide to translocate from the cytosol into the ER.

To examine further the role of L₋₁₀ in vitro translation in the presence or the absence of microsomes was performed (Fig. 6,B). In the absence of microsomes, 30-kDa proteins were produced for each form of p35₋₅₇ (L₋₁₀→R, L₋₁₀→R/R₊₁→G, and p35₋₅₇). However, when microsomes were present, only the unmodified p35₋₅₇ proteins had increased in m.w., suggesting that only this protein was translocated and subsequently glycosylated. As the 29-kDa proteins produced in COS-7 cells by Δh, L₁₀→R, and L₋₁₀→R/R₊₁→G were identical in size to the proteins observed in tunicamycin-treated COS-7 cells transfected with wild-type p35₋₅ (shown in Fig. 4,B), we compared results obtained from tunicamycin-treated cells transfected with either p35₋₅₇ or L₋₁₀→R (Fig. 6 C). Tunicamycin treatment of p35₋₅₇ transfected cells resulted in 29- and 24-kDa proteins, whereas COS-7 cells transfected with the L₋₁₀→R mutation of p35₋₅₇ produced a single 29-kDa species, regardless of treatment with tunicamycin. It is unlikely that the introduced mutations somehow affected preprotein glycosylation, which, in turn, inhibited cleavage, since our studies with tunicamycin demonstrated that primary cleavage within the signal peptide occurred regardless of the glycosylation state. Taken together, these results demonstrated that processing of the p35 secretory peptide within the ER resulted in cleavage within the mid-h-region of the signal peptide and not at the site predicted for cleavage of a signal peptide by signal peptidase.

The amino terminus of secreted p35 has been mapped to an R (12, 19), designated +1 (see Table II and Fig. 2,A). Since cleavage within the mid-h-region of the secretory peptide would not release a protein with an amino terminus comparable to that of the mature protein, we hypothesized that a second cleavage must have occurred after exiting the ER. Additional sequence analysis revealed that the c-region of the signal peptide resembled a monobasic protease recognition site, with cleavage predicted for the N-terminus of R₊₁: 1) cleavage occurs at R; 2) there is H at the −5 position; and 3) L and A are present immediately preceding the single R cleavage site (20, 21). Therefore, we examined this possibility using site-directed mutagenesis (Fig. 7,A) of essential elements within this sequence. As expected, replacing R₊₁ with G (R₊₁→G) or L (R₊₁→L) did not affect cleavage within mid-h to produce the stable 31-kDa p35 preprotein contained within the cell lysate. However, subsequent p35 protein secretion was abolished by R₊₁→G and was greatly reduced by R₊₁→L (Fig. 7 B). Taken together, these results confirmed that removal of the p35 signal peptide required two sequential cleavages. Translocation of the preprotein into the ER was accompanied by the addition of 7 kDa of carbohydrate. The first cleavage arose after entry into the ER, within the mid-h-region of the signal peptide. The second cleavage, which generated the amino terminus of the secreted protein, occurred either immediately before or upon exiting the ER into the Golgi, where an additional 5 kDa of carbohydrate was added. The latter steps appeared to be quite rapid, as none of these forms could be detected intracellularly.

In this study we show that post-translational processing of IL-12 subunits occurs by disparate pathways. Processing of the p40 subunit, which can homodimerize or heterodimerize with p35, conforms to the conventional model of signal peptide function. By this model the signal peptide is removed cotranslationally during translocation into the ER through a single cleavage of the signal peptide mediated by signal peptidase. This pathway is thought to be used for secretion of many cytokines, chemokines, and growth factors. In contrast, processing of the p35 signal peptide does not follow this pathway, requiring instead two sequential cleavage steps, possibly in distinct intracellular compartments. Unlike signal peptide processing by signal peptidase, translocation of the p35 protein into the ER is not accompanied by proteolytic cleavage of the signal peptide. Within the ER, an intermediate cleavage occurs within the mid-h-region of the signal peptide, possibly around L₋₁₀, and is accompanied by noncomplex glycosylation. This form of the protein is rather stable, having a half-life of approximately 3 h (a transient form corresponding in size to the unprocessed primary preprotein is also observed, but with a half-life of ∼30 min). Upon exiting the ER into the Golgi compartment, a second cleavage generates the amino terminus of the mature protein. This step appears to be critical, because mutation of the cleavage site abrogates secretion, but does not affect processing within the ER. This process must also be quite rapid, as none of this form of the protein can be detected intracellularly by immunoprecipitation. Unlike p40 and many other secreted proteins (22, 23, 24), p35 appears to require glycosylation for secretion, since inhibition of N-linked glycosylation by tunicamycin completely inhibits secretion.

Our current understanding of signal peptide function suggests that the signal sequence of a nascent protein emerges from the ribosome and binds to a signal sequence binding site on the methionine-rich domain of SRP54, one of the six different polypeptides that together with one 7S RNA molecule make up the signal recognition particle (SRP). This complex of ribosome-nascent protein and SRP then interacts with SRP receptor on the ER membrane, facilitating translocation of the peptide for further processing into the secretory pathway (16, 25). Similar to the findings of other investigators (26), we have shown that disruption of the hydrophobic core of the signal peptide inhibits translocation, suggesting that a similar process is required for p35. In particular, a single amino acid substitution at the −10 position completely inhibited this process, whereas similar substitutions at positions −8 and −7 were less inhibitory. Since the first cleavage site and the signal recognition site appear to be so closely related, we are unable to precisely identify the position of the first cleavage by the methods employed.

How the p35 signal peptide manages to bypass signal peptidase is unknown at this time. We do not think that the p35 signal peptide mediates this process by itself, because when the p35 signal peptide was used to direct secretion of p40 in vitro, it behaved as a typical secretory peptide, much as the p40 signal peptide. However, when we used the p40 signal peptide to direct secretion of p35 in vitro, the p40 signal peptide was not removed upon entry into microsomes (data not shown), indicating that the ability to evade signal peptidase was associated or at least influenced by some portion of the main body of the protein and was not mediated solely by the signal peptide.

What is the mechanism for generation of the amino terminus of secreted p35? The first amino acid residue of the secreted p35 peptide was shown previously by amino acid sequencing to be an R (13, 19). Consequently, the second site of cleavage must occur amino terminal to this residue. Upon closer examination of the amino acid sequence around R₊₁, we noted that the sequence conformed to a monobasic cleavage recognition site, as proposed by Benoit et al. (20) and more recently by Devi (21); namely, there is H at the −5 position and L and A immediately precede the single R cleavage site. Using site-directed mutagenesis we showed that reducing the length and pK_a of the basic amino acid side chain (L) diminished, while removing the functional group altogether (G) abrogated, secretion, demonstrating the importance of this functional group in cleavage and/or site recognition. Proteases cleaving amino terminal to monobasic sites are often metalloproteases (27, 28, 29). We are currently investigating a role for metalloproteases in cleavage at R.

Proteases recognizing monobasic cleavage sites appear to have specific subcellular distributions (including the Golgi compartments), and different enzymes have been reported for every preprotein examined (21). This raises the possibility that specific inhibition of injurious IL-12 production, as in the case of toxic shock (30), might be achievable by blocking the function of a p35 monobasic-specific endopeptidase without globally affecting immune responses. Moreover, we do not know at this time the tissue distribution of the p35 monobasic-specific endopeptidase or whether all cell types can process and secrete IL-12. This raises the concern that in vivo delivery of IL-12 by gene transduction methods to some tissue types may result in atypical production of one or both of the subunits, with possible unintended effects.

In summary we demonstrated that p40 is processed in a manner consistent with the signal peptide/signal peptidase model, but that p35 is processed atypically. Initial cleavage of p35 occurred within the hydrophobic domain of the signal peptide and was accompanied by but not dependent upon partial glycosylation within the ER. Cleavage to release the amino terminus of the secreted form of the protein appeared to be a rapid, transient process, with no detectable cellular accumulation, and appeared concomitant with additional complex glycosylation. Glycosylation appeared to be requisite for p35, but not for p40, secretion.

This work is dedicated to the memory of Dr. David S. Finbloom, whose intellectual acumen, thoughtful advice, and moral support were invaluable to the pursuit of excellence in science at the Center for Biologics Evaluation and Research. We thank Drs. Eda Bloom, Kathleen Clouse, Blair Fraser, Edward Max, Jack Stein, and Giovanna Tosato for helpful discussions and critical review of the manuscript, and Ms. Valerie Calvert for purification of human monocytes.