Heterogeneous Nuclear Ribonucleoprotein P2 Is an Autoantibody Target in Mice Deficient for Mer, Axl, and Tyro3 Receptor Tyrosine Kinases

Abstract

Deficiencies in clearance of apoptotic cells predispose to the development of autoimmune disease. This is evident in mice lacking the receptor tyrosine kinases Tyro3, Axl, and Mer. Deficient mice exhibit an increased abundance of apoptotic cells in tissues and manifest diverse autoimmune conditions. To test these mice for the presence of autoantibodies to apoptotic cells, we generated spontaneous splenic B cell hybridomas and used a novel microscopy screen to detect Ab binding to apoptotic Jurkat cells. From hybridomas secreting IgG Abs reactive with apoptotic cells, we selected one that recreated the major serum specificity for apoptotic cells. The Ab LHC7.15 bound to an Ag that is differentially distributed between the nucleus and the cytoplasm in live and apoptotic cells. In late apoptotic cells, the Ag coalesces into aggregates that bleb from the cell surface. Immunopurification of the Ag, followed by mass spectrometry, identified a protein of 69 kDa whose partial sequence matched heterogeneous nuclear ribonucleoprotein P2. This multifunctional protein binds DNA, RNA, and several known ribonucleoprotein autoantigens. Our observations indicate that a ribonucleoprotein complex, formed and translocated to the cell surface in apoptosis, represents a potent stimulus for breaking tolerance and inducing systemic autoimmunity in mice with defective clearance of cell remnants.

Inefficient clearance of apoptotic cells predisposes to autoimmunity. This paradigm is firmly rooted in observations from diverse experimental systems (reviewed in Refs. ^1–3). Defects in serum proteins that serve in the recognition of apoptotic cells, such as complement components and proteins from the pentraxin and collectin families, are risk factors for systemic lupus erythematosus (4) and for lupus-like conditions in mouse models for the human disorder (5–7). Defects in enzymes that assist in the efficient disposal of apoptotic cells delay apoptotic cell clearance and increase the incidence of autoimmunity (8, 9). Delayed clearance provides time for the execution of the late stages of apoptosis, allowing the release of substances suspected of stimulating the immune response (10–13). Moreover, late stages of apoptosis are characterized by the redistribution of nucleoprotein complexes to the cell surface (14–16), thereby providing an opportunity for a direct stimulation of B cell responses by antigenic complexes that become externalized on apoptotic cells (17).

Inactivation or deletion of phagocyte receptors that mediate apoptotic cell clearance promote the development of autoimmunity (18–20). Thus, mice with homozygous defects in each of the three receptor tyrosine kinases, Tyro3, Axl, and Mer (TAM^−/− mice),³ exhibit an increased abundance of apoptotic cells in their tissues, a pleomorphic activation of the immune system, and signs of autoimmunity affecting multiple organs (18). The clinical picture in TAM^−/− mice includes splenomegaly, lymphocytic tissue infiltrates, glomerulonephritis, arthritis and bone deformities, micro-capillary leakage, strokes, and seizures (18). By ELISA, autoanti-bodies that are characteristic of various autoimmune disorders, including anti-DNA, anti-phospholipid, and rheumatoid factor, are detected. However, it is not clear whether and how most of the autoimmune and clinical findings are caused by the defective clearance of apoptotic cells.

To test whether defective clearance in TAM^−/− mice gives rise to autoantibodies that recognize specific features of apoptotic cells, we screened sera and monoclonal autoantibodies for binding to cells in apoptosis. We used a novel microscopy screen to detect Ab binding to autoantigens in apoptotic Jurkat cells (16, 17). Sera from TAM^−/− mice showed a consistent and characteristic reactivity with apoptotic cells. To elucidate the molecular basis of this reactivity, we analyzed mAbs from this murine model of systemic autoimmunity. Spontaneous hybridomas were readily obtained from TAM^−/− mice and many were reactive with apoptotic cells. We focused our initial work on a hybridoma secreting an Ab that recreates the main characteristics of the serum reactivity in these mice. The Ab LHC7.15 binds nuclei in a speckled pattern reminiscent of the binding of anti-spliceosomal autoantibodies, and to sites in the cytoplasm of live cells. Upon induction of apoptosis, LHC7.15 binds intensely to nuclei before nuclear fragmentation, yet is unreactive against nuclear fragments. In later stages of apoptosis, LHC7.15 binds large cytoplasmic granules that form near the plasma membrane, occupy a subset of surface blebs, and separate as apoptotic bodies from the remainder of the apoptotic cell. Molecular analysis revealed that LHC7.15 is specific for heterogeneous nuclear ribonucleoprotein (hnRNP) P2.

Materials and Methods

Mice

TAM^−/− mice were generated and maintained as previously described (18, 21). Mice, ranging in age from 4 to 10 mo, were bled, and sera were prepared following standard procedures. All procedures involving mice were approved by the University of Tennessee Health Sciences Center and the Salk Institute Institutional Review Boards.

Hybridoma fusions

A female TAM^−/− mouse was used at 8 mo of age for the generation of B cell hybridomas. Splenic B cells were prepared and used immediately for fusion to SP2/0 cells following a standard procedure (22). After incubation at 37°C for 1 h, cells were plated in 12 96-well plates. After colonies reached an appropriate size, supernatants were tested for H chain isotype and 195 of the most vigorous IgG-producing wells were expanded to 24-well plates. All 195 supernatants were tested for binding to apoptotic Jurkat cells by confocal microscopy, and cells from 27 wells were further sub-cloned for use in additional tests.

Induction of apoptosis

Jurkat cells (clone E6-1) were treated with 2.0 μM camptothecin (Sigma-Aldrich) for 3–6 h. Apoptosis was assessed by light microscopy, and cells were harvested after at least 25% of cells in the population exhibited surface blebs.

Confocal microscopy

Ab binding to apoptotic or viable cells was evaluated as described (16, 17, 23). Briefly, cells were washed in HBSS (supplemented to 3 mM CaCl₂) and fixed for 15 min in ice-cold 6% paraformaldehyde (Electron Microscopy Sciences). The fixative was freshly prepared in the same buffer. Fixed cells were washed and blocked by suspension in wash buffer (HBSS containing 3 mM CaCl₂, 3% FBS, and 0.02% azide) for 5 min. Cells were suspended in buffer containing sera at a 1/200 dilution or purified Abs at 20 μg/ml. After incubation with the primary Abs, cells were washed in wash buffer, pelleted as above, and incubated in a mixture of Alexa Fluor 647 rabbit anti-mouse IgG antisera (1/100 dilution), SYTOX Orange DNA stain (1/10,000 dilution), and Alexa Fluor 488 Annexin V (1/70 dilution). All secondary reagents and stains were obtained from Molecular Probes.

Samples were viewed on a Zeiss LSM 510 laser scanning microscope (Carl Zeiss) by using a ×40 (or a ×100) Plan-Apochromat oil-immersion lens, and laser excitation at 488, 543, and 633 nm. Detection channels were set to record fluorescence emission above 650 nm for Alexa Fluor 647, between 560 and 615 nm for SYTOX Orange, and between 505 and 530 nm for Alexa Fluor 488. Consecutive images were collected at optimized intervals of between 0.4 and 0.8 μm to assemble complete three-dimensional representations of treated and untreated cells.

Apoptotic cell extracts

Cytoplasmic and nuclear cell extracts were prepared as described (17), starting from camptothecin-treated and control cultures preincubated with z-VAD-fmk (Enzyme System Products) before induction of cell death. Briefly, cells were harvested by centrifugation, washed in HBSS, and incubated at 10⁸ cells per milliliter in hypotonic buffer containing 10 mM HEPES (pH 7.4), 10 mM NaCl, 1 mM EDTA, and 1 mM DTT for 15 min on ice. Subsequently, Nonidet P-40 was added to a final concentration of 0.5%, cells were vigorously vortexed for 30 s, and the soluble cytoplasmic fraction was recovered as the supernatant after centrifugation for 5 min at 13,000 rpm. The nuclear pellets were extracted separately in 2×Laemmli gel loading buffer.

Immunopurification

The cytoplasmic and nuclear fractions prepared at various times after induction of cell death and controls were used for immunoprecipitations. The LHC7.15 IgG2a was adsorbed on protein A-Sepharose (Zymed Laboratories) according to suppliers’ recommendations. Adsorbed Ab was incubated with the cytoplasmic extracts prepared at 4 h after induction of apoptosis. After a 1-h incubation at 4°C, depleted extracts were collected, beads were washed three times with 10-fold excess of lysis buffer, and proteins were eluted in 2×gel loading buffer.

Immunoblotting

Eluates from the LHC7.15 affinity purification and control fractions were analyzed on duplicate 10% denaturing SDS-PAGE. One of the gels was used for Coomassie blue staining, the other for Western blotting. Proteins were transferred from the gels to nitrocellulose in a semidry blotter (Owl Separation Systems) with transfer buffer (48 mM Tris, 39 mM glycine, 0.04% SDS, and 20% methanol) at 0.8 mA/cm² for 90 min. Membranes were air-dried and blocked in PBS buffer containing 2% BSA, 3% FCS, 2.5 mM EDTA, and 0.25% Tween 20. Blocked membranes were incubated with 0.5 μg/ml LHC7.15 for 1 h. All subsequent steps and washes were in 0.15 M NaCl, 50 mM Tris (pH 7.4), 0.2% Tween 20. Alkaline phosphatase-labeled secondary reagents were used according to manufacturers’ recommendations. Immunoreactive bands were visualized by using the chromogenic substrate in the AP color development kit (Bio-Rad).

Mass spectrometry

Tryptic peptides were extracted from polyacrylamide, as described (24). Briefly, the protein bands were excised from SDS-PAGE and placed in 25 mM ammonium bicarbonate to swell. Proteins were reduced in 10 mM DTT at 55°C for 10 min and alkylated with 10 mM iodoacetamide at room temperature. Proteins were digested in the gel with sequencing-grade trypsin (Promega) for 16 h at 37°C. The resulting peptides were extracted into acetonitrile:ammonium bicarbonate and subjected to analysis by mass spectrometry.

Mass spectra were recorded on a Bruker Ultraflex MALDI-TOF/TOF reflecting time-of-flight mass spectrometer (Bruker Daltonics). Matrix-related ions and trypsin autolysis products were used for internal spectra calibration. Delayed ion extraction results in peptide masses with better than 50 parts per million mass accuracy on average, limited mainly by ion statistics in the smaller peaks. Locally installed PROWL for Intranet software (Proteometrics) was used to search a nonredundant protein sequence database (National Center for Biotechnology Information) with a list of peptide masses.

Results

Serum Abs from TAM triply deficient mice bind to apoptotic cells

The impaired apoptotic cell clearance in TAM^−/− mice is considered a direct cause for the development of the autoimmune phenotype (18). Persistence of apoptotic cells in vivo may lead to an activation of the adaptive immune system by self-Ags released from cell corpses. To examine TAM^−/− mouse sera for reactivity with apoptotic cells, we induced apoptosis in Jurkat cells by treatment with the topoisomerase I inhibitor camptothecin. We applied a confocal microscopy technique used previously to detect nucleosome core particle exposure to anti-nucleosome autoantibodies at the cell surface in apoptosis (17). The advantage of this technique is that it employs fixed cells that are not exposed to denaturants or detergents. Binding of serum Abs to apoptotic cells was assessed with fluorescent anti-mouse IgG Abs (Fig. 1). The location of the Ag relative to the nucleus was determined by visualizing DNA by binding to Sytox Orange. The cell surface was identified by the binding of annexin V to phosphatidylserine in the plasma membrane.

FIGURE 1.

TAM^−/− mouse sera bind to apoptotic cells. Jurkat cells, treated with camptothecin or left untreated, were fixed and incubated with dilutions of mouse sera before detection of bound Abs with anti-mouse IgG (displayed in red). Untreated cells (A) display Ab binding in the nucleus and cytoplasm. Nuclear DNA, bound by Sytox Orange, is displayed in blue; plasma membranes bound by annexin V are displayed in green. Nuclear binding of IgG is localized to speckles and excluded from nucleoli. The binding to one of the nuclei is greatly reduced. Cytoplasmic binding is directed to granules of heterogeneous size. After induction of apoptosis (B), nuclear Ab binding of increased intensity is noted in nuclei with reduced diameter. Cytoplasmic granule binding is observed. C, Serum IgG-reactive surface blebs (arrowheads) form on cells with condensed nuclear DNA and partially immunoreactive nuclei. Cell in advanced apoptosis, identified by nuclear fragmentation and membrane blebbing (D), shows absence of Ab reactivity in nuclear fragments and a prominent surface bleb that reacts with serum IgG (arrowhead). Two cells with condensed nuclear DNA and partially immunoreactive nuclei were observed in the same frame as one Ab-reactive subcellular particle (arrow in E). Eight TAM^−/− mouse sera gave equivalent IgG binding patterns. Cells incubated in the absence of primary Abs or with control littermate sera did not yield appreciable Ab binding. The images represent individual cross-sections from stacks of serial sections providing complete three-dimensional cell reconstructions. The larger image in each set represents a merged image composed of the three separately acquired fluorescent signals shown at one-third size.

Binding of serum IgG produced a constellation of diverse, yet specific and reproducible binding patterns. In live cells (Fig. 1A), binding of IgG to nuclei resulted in a fine grainy pattern that excluded nucleoli, although other cells within the same population showed very little or no nuclear IgG binding. Live cells also showed cytoplasmic granules of various sizes reacting with IgG. A series of consistent and reproducible binding patterns emerged upon analysis of sufficient numbers of apoptotic cells. Distinct stages of apoptosis were characterized by distinct distributions of the Ag(s), suggesting a dynamic and evolving function in apoptosis. After induction of apoptosis with camptothecin (Fig. 1, B–E), nuclear Ab binding increased in intensity, concomitant with the condensation of nuclear chromatin and the decrease in the diameter of the nucleus (Fig. 1B). Subsequently, IgG reactivity may depart the nucleus in a poleward fashion, as cells with IgG binding that was limited to one nuclear hemisphere were noted (Fig. 1C). In cells with fragmented nuclei, strikingly little or no IgG binding to nuclear fragments was observed. Instead, the predominant binding was to large cytoplasmic Ag granules, often located at or near the cell surface (Fig. 1D). As shown previously, blebs, surface protrusions containing large, Ag-dense granules (arrowheads in Fig. 1, C and D), may detach from the remainder of the apoptotic cells (25). Accordingly, subcellular particles that intensely bound TAM^−/− serum IgG were observed in apoptotic cell cultures (arrow in Fig. 1E).

Multiple experiments with eight TAM^−/− sera yielded identical IgG binding patterns in our assay. In each case, the full spectrum of IgG binding to live and apoptotic cells was observed. An equal number of MRL/lpr/lpr sera generated a more varied set of binding patterns and, notably, none resembled the TAM ^−/− pattern. A comparison with single Mer^−/− knockout mouse sera revealed that only approximately half of the Mer tyrosine kinase-deficient mice produced IgG reactive with apoptotic cells. Importantly, the sera that reacted with apoptotic cells resulted in binding that was distinct from the TAM^−/− pattern (data not shown). Thus, our data indicate that TAM^−/− IgG consistently react with Ag(s) with multiple functions in the nucleus and cytoplasm of live and apoptotic cells. The Ag(s) exhibit a dynamic behavior in apoptosis, accumulating at the cell surface, and being shed from the cell in the form of distinct subcellular particles.

One mAb reproduces the major serum specificity of TAM^−/− mice

To examine the possibility that a single Ab reproduces the predominant binding of TAM^−/− mouse sera, we generated a panel of monoclonal IgG Abs and screened them for binding to apoptotic Jurkat cells. Of the nearly 200 spontaneously activated, IgG-producing B cells that were immortalized in our fusion, 27 bound to Ags expressed in apoptotic Jurkat cells, as assessed by confocal immunomicroscopy. Four of these reproduced the main characteristics of TAM^−/− serum IgG binding. One of the four, LHC7.15, was selected for further analysis.

The LHC7.15 hybridoma produces an IgG2a/κAb whose binding was visualized in the manner used to analyze serum IgG (Fig. 2). Indeed, each component of the complex serum binding pattern could be reproduced by using this single Ab. In live cells (Fig. 2A), binding to cytoplasmic granules and to fine, punctate nuclear determinants was observed. In apoptotic cells (Fig. 2B), binding to nuclei increased in intensity as the nucleus condensed, yet binding to nuclei was greatly reduced or absent after nuclear fragmentation. In contrast, cytoplasmic granules were fewer yet larger, in more advanced apoptosis, and the large Ag-containing aggregates were predominantly localized at or near the cell surface (Fig. 2B). In accord with the TAM^−/− serum binding, cultures of late apoptotic cells contained small apoptotic bodies that intensely reacted with our Ab. Therefore, LHC7.15 recapitulates most or all of the distinct binding patterns that were initially observed with the TAM^−/− mouse sera.

FIGURE 2.

The LHC7.15 mAb binds to apoptotic cells. Ab binding to rapidly proliferating Jurkat cells (A) results in nuclear binding that is observed as fluorescent nuclear speckles and excludes nucleoli. Binding to nuclear speckles varies in intensity among cells. Binding to Ag in the cytoplasm gives rise to granular fluorescence, with granules of variable size localizing throughout the cytoplasm. Five hours after addition of camptothecin (B), a variety of apoptotic cell phenotypes are observed. Nuclear Ab binding varies dramatically, being most intense in cells with nuclei of reduced size and absent in fragmented nuclei.

LHC7.15 IgG binds a 69-kDa protein

To explore the molecular target of LHC7.15 binding, we prepared cellular extracts of live and apoptotic Jurkat cells and examined Ab binding by immunoblotting. We observed one specific band that migrated at a relative mobility of 69 kDa in both viable and apoptotic cell extracts (Fig. 3A). No differences in migration were observed between cytoplasmic or nuclear extracts for time points ranging from 0 to 6 h after addition of camptothecin (our unpublished data), suggesting the Ag is not attacked by site-specific proteases.

FIGURE 3.

Detection and purification of the autoantigen recognized by LHC7.15. Cytoplasmic extracts were prepared from viable cells (lane 1) or cells incubated in the presence of camptothecin for 5 h (lane 2). Extracts were separated by SDS-PAGE, transferred to nitrocellulose, and probed with LHC7.15 (A). One intensely reactive protein band is observed to migrate at ~69 kDa. Dashes indicate relative mobilities of reference proteins whose masses are noted. Cytoplasmic extracts prepared 4 h after camptothecin addition were examined before (lane 3) and after reacting with the LHC7.15 mAb adsorbed on protein A-Sepharose (lane 4). The starting and depleted extracts were compared with proteins dissociated from the LHC7.15 affinity matrix by wash buffer (lane 5) or by SDS (lane 6). Note bands that correspond to the IgG2a/κH and L chains. One band of ~69 kDa is observed in the SDS eluate by Coomassie blue staining. Its migration coincided with the migration of an immunoreactive protein that was detected by Western blotting with LHC7.15 in the starting extract and the SDS eluate. The stained gel shown in B and the Western blot shown in C contained equivalent amounts of proteins and were run in parallel.

To purify the Ag for protein identification, we examined the efficiency of immunoaffinity purification using LHC7.15 adsorbed on protein A-Sepharose. The immunoreactive band, a minor protein species in the total cytoplasmic extract (Fig. 3B), was selectively depleted by adsorption to LHC7.15 (Fig. 3C). Control reactions indicated that this band did not adsorb to protein A-Sepharose alone (our unpublished data). After removal of the majority of contaminating proteins, the specific Ag was released from the affinity matrix with a strong denaturant (Fig. 3, B and C). Electrophoretic analysis of the eluted fraction revealed a highly purified immunoreactive protein migrating at 69 kDa.

Mass spectrometric identification of the hnRNP P2 autoantigen

The purified Ag was excised from polyacrylamide and subjected to trypsin digestion. MALDI-TOF of the tryptic digest resolved peptides whose masses were used to query a protein database. The best match was with hnRNP P2, also known as FUS/TLS (fused by translocations in liposarcomas) (ENTREZ accession number: P35637). Overall, 9 peptides, ranging in size from 6 to 19 aa, were matched (Table I), representing a 17% sequence coverage of P2 (89 of a total of 525 amino acids; Fig. 4). Two of the peptides matching the P2 sequence were identified in previous studies, including the study by Calvio et al. (26) who established the identity of hnRNP P2 with FUS/TLS, a protein whose coding sequence is interrupted by chromosomal breakpoints in liposarcomas (27). The second peptide, TGQPMINLYTDRETGK (Table I), previously was used to identify FUS/TLS in ribonucleoprotein (RNP) assemblies organizing the formation of focal adhesions (28). The mass of an additional peptide, GGRGGYDRGGYRGR (Table I), matched the predicted size for a peptide of that sequence, provided all three internal arginines are dimethylated. Clearly, the presence of dimethylarginines will have to be confirmed by peptide sequencing. Indirect support for this possibility rests on the resistance of these arginines to trypsin digestion and data reported by Rappsilber et al. (29) who identified dimethylated arginines in this part of hnRNP P2.

Table I.

Tryptic peptides of hnRNP P2 identified by mass spectrometry

Measured Mass	Computed Mass	Error (ppm)	Missed Cuts	Peptide Sequence
777.387	777.388	−1	1	RADFNR
835.212	835.295	−99	0	NECNQCK
1407.649	1407.681	−23	0	TGQPMINLYTDR
1423.618	1423.676	−41	0	TGQPMINLYTDR (+O@MET)a
1438.818	1438.743	53	2	VSFATRRADFNR
1538.734	1538.616	77	0	GGPMGEGGYGGGGSGGGGR (+O@MET)
1822.824	1822.887	−35	1	TGQPMINLYTDRETGKb
1893.839	1893.925	−45	1	AAIDWFDGKEFSGNPIKc
2001.835	2001.918	−41	3	GGFGPGKMDSRGEHRQDR (+O@MET)
1566.673	1566.812	−88	3	GGRGGYDRGGYRGR (3DMRs)d

^aMethionine oxidation (+O@Met) is a modification that is introduced by in vitro manipulation of a protein.

^bThis peptide was identified by de Hoog et al. (28) in proteins from spreading initiation centers.

^cThe sequence of this peptide was used to confirm the identity between hnRNP P2 and FUS/TLS (26).

^dThis peptide lies within a larger peptide shown by Rappsilber et al. (29) to contain dimethylated arginines (DMR) at the underlined positions.

FIGURE 4.

Trypsin digestion of the LHC7.15 autoantigen releases peptides that match peptides of hnRNP P2. Peptides listed in Table I are identified in the context of the protein sequence by underlining the corresponding sequences in hnRNP P2. Arginines (R) and lysines (K) are highlighted in bold typeface and the putative dimethylarginines are circled.

Several characteristics of the P2 amino acid sequence (Fig. 4) make this protein a less-than-ideal candidate for trypsin cleavage and mass spectrometry. For example, the first arginine is at position 212 and the first lysine is at position 263, thereby making 40% of the P2 amino-terminal sequence inaccessible to trypsin cleavage. Moreover, many of the 51 lysines and arginines in P2 are clustered, giving rise to short peptides or partial cleavage products. Both of the latter features increase random matches with entries in the database and thus lower significance of scoring. In addition, many of the arginines and lysines are followed by negatively charged amino acids, placing them in a sequence context that is suboptimal for trypsin proteolysis. Despite these cumulative factors, the match between our protein and hnRNP P2 was established with remarkably robust statistical significance.

To confirm independently P2 as the Ag recognized by LHC7.15, we compared the binding of LHC7.15 to the results obtained with a commercial Ab to TLS/FUS (BD Pharmingen, clone 15). Both Abs exhibited equivalent binding on Western blots and gave rise to identical immunofluorescence binding patterns (our unpublished data). Therefore, we identified hnRNP P2 as the autoantigen for LHC7.15 by independent and highly specific criteria.

Discussion

We report that mice with the triple homozygous deletions of the Mer, Axl, and Tyro3 receptor tyrosine kinases and delayed apoptotic cell clearance produce autoantibodies reactive against P2, a member of the hnRNP group of proteins. Moreover, P2 itself, complexes between P2 and other RNP, and particles shed in apoptosis that contain P2 are consistent targets of TAM^−/− autoantibodies (Fig. 1). The distinct and specific RNP autoreactivity of TAM^−/− IgG is drastically different from the reactivity of autoantibodies to the nucleosome core particle (common in MRL/lpr/lpr and NZB ×NZW F₁ mice), yet both types of autoantigens exhibit a dynamic redistribution during apoptosis (17). Our observations are consistent with the idea that P2 provides a strong and penetrating stimulus to break tolerance and induce autoimmunity. In this study, we discuss functions of P2 that may contribute to the induction of an autoimmune response.

Autoimmunity to one component of an RNP complex, over time, expands to include additional components of the complex (30, 31). Therefore, proteins that interact with P2 are expected to become targets of autoantibodies. Indeed, preliminary results indicate that TAM^−/− mice also express autoantibodies to other RNP (data not shown). This is consistent with the fact that several of the hnRNP that copurify with primary transcripts synthesized by RNA polymerase II (32) are autoantigens in lupus (33, 34), rheumatoid arthritis (35), and in neurodegenerative disorders (36). It is likely that anti-P2 autoantibodies arise by a related pathogenic mechanism, as TAM^−/− mice mirror the human disorders and develop immune complex deposits in kidneys, joint inflammation, and neurodegeneration (18). However, more has yet to be learned about the mechanisms that induce autoantibodies to P2.

P2 associates with other autoantigens, both in the nucleus and at the cell membrane. Opportunities for interactions arise because P2 belongs to those hnRNP that shuttle between the nucleus and the cytoplasm (32). In the nucleus, P2 transiently associates with the RNA polymerase at specific gene promoters but prefers to bind the nascent RNA transcript (37). Purified P2 exhibits sequence-specific RNA binding (38, 39) and modulates the selection of splice sites in transcripts that are substrates for alternative splicing (40). Bound to mRNA, P2 transits to the cytoplasm where it contributes to the spatially restricted translation of particular messages. For example, the translation-coupled assembly of focal adhesions at the cell membrane is regulated by hnRNP P2 (28). Notably, both nuclear hnRNA processing (41, 42) and localized protein synthesis at the plasma membrane (28) provide opportunities for P2 to join Sm and SR proteins, well-known RNP autoantigens (43, 44).

Our observations suggest a likely series of events that contribute to anti-hnRNP P2 autoimmunity. Of particular importance is the behavior of P2 in apoptosis, as apoptotic cells are the likely source of Ags that incite autoimmunity. The immunoreactivity of P2 is most intense just before nuclear fragmentation (Figs. 1 and 2). Nuclear fragmentation, a hallmark of apoptosis, remains poorly understood. Because the meshwork of nuclear lamins provides a scaffold that supports the nuclear envelope, caspase-mediated degradation of lamins is considered a requirement for nuclear fragmentation (45). Moreover, chromatin is much more densely packed just prior to and after nuclear fragmentation. Therefore, it has been suggested that nucleases, activated in apoptosis, contribute to nuclear chromatin condensation (46, 47). However, it is not clear how DNA cleavage might enhance the packing density of nuclear DNA.

We propose that chromatin condensation and nuclear fragmentation may require the removal of RNA from the apoptotic nucleus. Electron microscopy reveals that RNP particles depart from the apoptotic nucleus and accumulate in large clusters at the cell surface (48). P2 is a component of these clusters because it is detected within large, immunoreactive granules that exit from the nucleus and form large protrusions at the cell surface (Fig. 1, C–E). Presumably, P2 associates with Sm and SR proteins in RNP granules that bleb from the cell surface (49). Studies with various autoantibodies have demonstrated that Sm (50), as well as other RNA-binding proteins (51), become accessible at the cell surface in apoptosis. Exposure of autoantigens at the cell surface is considered an important requirement for the induction of autoimmune responses (1, 17). Moreover, P2-containing blebs, or the apoptotic bodies derived from these blebs, may have the potential for breaking tolerance and inducing autoimmunity. Most autoantigens are modified during apoptosis, some by proteolytic cleavage, others by various types of posttranslational modifications (52). Posttranslational modifications of P2 may enhance the potential of P2 to serve as autoantigen. In particular, deimination of arginine residues in apoptosis (53) may enhance immunogenicity of P2.

Autoimmunity to P2 may be induced more easily if apoptotic cells discharge a tissue-specific variant of the protein. For example, the distinct manner of pyruvate dehydrogenase processing in apoptotic cholangiocytes may be a reason for the production of Abs to this enzyme in primary biliary cirrhosis (54). Is there evidence for a tissue-specific isoform of hnRNP P2 that may stimulate an autoimmune response? At first glance, this possibility seems remote, as P2 is abundantly expressed in most or all tissues (55). However, a unique and essential role for P2 in homologous DNA recombination was uncovered by constructing mice with P2 germ-line deletions (56). The lack of P2 caused reproductive sterility because mutants were not able to replace P2 in one essential step of DNA strand invasion (57). It seems plausible that the unique function of P2 in meiotic germ cells is accomplished by an antigenically distinct isoform of the protein. If this idea is correct, it follows that apoptotic germ cells could release a unique isoform of P2. This scenario is consistent with the fact that autoimmune manifestations in TAM^−/− mice arise after the onset of sexual maturation (18), a time characterized by greatly increased rates of germ cell apoptosis in these mice (21). The release of a tissue-specific form of P2 may trigger a systemic autoimmune response in a manner that is analogous to the autoimmunity elicited by administration of a heterologous, structurally conserved protein (58). In summary, we expect that TAM^−/− mice will continue to provide a useful model for autoimmune pathogenesis, and that studies on the mechanisms causing anti-P2 autoimmunity will contribute to a deeper understanding of immune responses to RNP.