Two Distinct Pathways Leading to Nuclear Apoptosis

One of the hallmarks of apoptosis is the degradation and concomitant compaction of chromatin. It has been generally assumed that caspases as well as downstream effectors such as caspase-activated DNase (CAD) and Acinus are rate limiting for the development of nuclear apoptosis 1,2,3,4,5. Accordingly, the inactivation of the caspase-3 gene 6,7 and that of the caspase activator Apaf-1 8 can delay cell death and largely abolish the type of chromatin condensation observed in normal control cells treated with apoptosis inducers such as staurosporine (STS) or etoposide. Similarly, the inactivation of CAD prevents advanced chromatin condensation in different cell types 9. However, caspases and CAD are not the only effectors causing nuclear apoptosis. Thus, chromatin condensation has been observed in lymphoid cells treated with STS, anti-CD2 10, anti-CD4, or anti-CXCR4 11, as well as in fibroblasts overexpressing PML 12, even when caspase activation is inhibited. Partial chromatin condensation is found in thymocytes undergoing apoptosis in the presence of the pan-caspase inhibitor Z-VAD.fmk 13. Recently, apoptosis-inducing factor (AIF), a mitochondrial intermembrane flavoprotein, has been found to translocate from mitochondria to nuclei in a caspase-independent fashion. When added to purified nuclei, recombinant AIF causes caspase-independent large scale (∼50 kb) DNA fragmentation and a type of peripheral chromatin condensation that resembles the first stage of nuclear apoptosis (stage I) observed in intact cells undergoing apoptosis 14,15. Here, we examined the question of whether several independent pathways may lead to nuclear apoptosis.

Mouse embryonic fibroblasts (MEFs) obtained from caspase 3^−/− 7, Apaf-1^−/− 8, or control mice were cultured with STS (2 μM), etoposide (100 μM), cisplatin (150 μM), arsenite (50 μM; Sigma-Aldrich), and/or Z-VAD.fmk (50 μM; Enzyme Systems). Cells were microinjected (pressure 150 hPa; 0.4 s; reference 16) with the following: buffer only; recombinant AIF; an inactive deletion mutant of AIF (Δ1-351; reference 14); horse cytochrome c (Cyt-c; Sigma-Aldrich); recombinant active caspase-3 17; or inactive inhibitor of CAD (ICAD)/CAD or active CAD (generated by digestion of the 250 nM ICAD–CAD complex with 3 U of caspase-3 in 10 μM of CAD buffer; 30 min at room temperature, followed by addition of 100 μM Ac-DEVD.fmk; Enzyme systems). After microinjection, cells were cultured for 180 min and stained for 10 min with the mitochondrial transmembrane potential (ΔΨ_m)-sensitive dye CMXRos (100 nM) and the DNA-intercalating dye Hoechst 33342 (1.5 μM; reference 16). Microinjected viable cells (100 per session; two to three independent sessions of injection) were identified by inclusion of 0.25% (wt/vol) FITC-dextran (green fluorescence) in the injectate. Only the blue and red fluorescence was recorded.

Fixed and permeabilized MEFs were stained for AIF and Cyt-c as described 14,15. A rabbit polyclonal antiserum, CM1 (revealed as anti-AIF), was used to detect the p18 subunit of cleaved caspase-3 18. Unfixed cells were incubated for 15 min with 1.2 μM ΔΨ_m-sensitive JC-1 (Molecular Probes). Confocal microscopy was performed on a Leica TC-SP equipped with an ArKr laser mounted on an inverted Leica DM IFBE microscope with an 63 × 1.32 numerical aperture oil objective. Two stages of nuclear apoptosis were distinguished by staining with 25 nM Sytox-green (Molecular Probes) for 15 min at room temperature. Stage I was characterized by rippled nuclear contours and a rather partial chromatin condensation, and stage II by a more pronounced pattern of chromatin condensation 14,15.

Cells were fixed in pellets for 1 h at 4°C with 1.6% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.4, washed three times, and then post-fixed in 1% osmic acid in phosphate buffer before scrapping, dehydration, and embedding. Ultrathin sections mounted on 200 mesh grids were examined in a JEOL 1200 EX electron microscope.

Oligonucleosomal DNA fragmentation was detected by agarose gel electrophoresis 19. For pulse field gel electrophoresis, DNA was prepared from agarose plugs (10⁶ cells; reference 20) and analyzed in a Bio-Rad CHEF-DR II (1% agarose; TBE; 200 V; 24 h; pulse wave 60 s; 120° angle; Bio-Rad Laboratories).

Cytosols from MEFs stimulated for 24 h with STS (2 μM), etoposide (100 μM), or cisplatin (150 μM) were prepared in cell-free system buffer (50 μl/10⁶ cells) supplemented with 50 μM Z-VAD.fmk, as previously described 21. Immunodepletion of AIF (or sham immunodepletion) was performed using an anti-AIF antiserum (or preimmune serum) and protein A/G coupled to agarose (Santa-Cruz Biotechnology, Inc.; reference 14). Purified HeLa cell nuclei were exposed to cytolic extracts (2 μg/μl protein), AIF 14, CAD, ICAD 4, caspase-3 17, and/or the AIF inhibitor para-chloromercuriphenylsulfonic acid (PCMPS; Sigma-Aldrich; reference 22) in cell-free system buffer 23, and nuclear DNA content was quantitated by cytofluorometry after staining with propidium iodide 14.

Apaf-1^−/− or caspase-3^−/− cells, as well as control MEF, responded to four different apoptosis inducers (STS, etoposide, cisplatin or arsenite) by manifesting a decrease in the ΔΨ_m, translocation of AIF from mitochondria to nuclei, and translocation of Cyt-c from mitochondria to the cytoplasm (Fig. 1A and Fig. B). As expected, at no time point did Apaf-1^−/− or caspase-3^−/− MEFs stain with an antibody specific for activated caspase-3 (Fig. 1 A). The nuclear phenotype manifested by Apaf-1^−/− or caspase-3^−/− cells appeared clearly distinct from that of control cells. Apaf-1^−/− or caspase-3^−/− cells (Fig. 1A,Fig. c) only manifested a minor peripheral chromatin condensation (stage I), as was also found in control cells after short-term incubation with STS. However, Apaf-1^−/− or caspase-3^−/− cells failed to develop the more advanced chromatin condensation (stage II) of control cells (Fig. 1A,Fig. c). None of the mitochondrial parameters nor the pattern or kinetics of chromatin condensation of Apaf-1^−/− or caspase-3^−/− cells (Fig. 1A and Fig. B) were influenced by addition of the pan-caspase inhibitor Z-VAD.fmk. However, Z-VAD.fmk arrested the nuclear apoptosis of control wild-type cells at stage I (Fig. 1A and Fig. B). At the ultrastructural level, nuclei from apoptotic Apaf-1^−/− or caspase-3^−/− cells demonstrated a rather partial chromatin condensation, with patches of chromatin abutting to the apparently intact envelope and without nucleolar degradation (Fig. 1 C). As a biochemical correlation of the morphological features of caspase-independent apoptosis, Apaf-1^−/− or caspase-3^−/− cells developed large scale DNA fragmentation to ∼50 kb (Fig. 1 D), yet failed to show the (presumably CAD-mediated; references 1, 2, 9) oligonucleosomal DNA fragmentation (Fig. 1 E).

Wild-type, Apaf-1^−/−, or caspase-3^−/− cells were microinjected with Cyt-c, active caspase-3, CAD, or AIF. As expected 8,24, Cyt-c alone induced signs of apoptosis (ΔΨ_m loss, chromatin condensation) in wild-type cells, but not in Apaf-1^−/− nor in caspase-3^−/− cells (Fig. 2A and Fig. B). In contrast, microinjection of active caspase-3 provoked full-blown apoptosis (stage II) in all three cell types (Fig. 2 B). Similarly, AIF and CAD induced apoptosis in all cell types (Fig. 2A and Fig. B). AIF induced a peripheral type of chromatin condensation (similar to the caspase-independent stage I, Fig. 1 A), whereas CAD (and its activator caspase-3) provoked a more advanced pattern of nuclear compaction (similar to the caspase-dependent stage II, Fig. 1 A). The differential effect of AIF and CAD was confirmed in a cell-free system. When added to purified HeLa nuclei, AIF caused peripheral chromatin condensation (Fig. 3A and Fig. B), whereas CAD induced a much stronger type of chromatin compaction accompanied by a reduction in nuclear size (Fig. 3A and Fig. B). Moreover, AIF alone caused large scale DNA fragmentation (Fig. 3 D), yet was unable to provoke the “ladder type” oligonucleosomal chromatin digestion (Fig. 3 E). In contrast, CAD provoked the digestion of DNA in two steps, first into ∼50 kb (at low doses; Fig. 3 F) and then into mono- and oligomers of ∼200 bp (at high doses; Fig. 3 E). Moreover, CAD could act on AIF-pretreated nuclei (which lack oligonucleosomal DNA fragmentation, Fig. 3 E) to induce oligonucleosomal fragmentation (Fig. 3 G). In conclusion, AIF and CAD cause two morphologically and biochemically distinct types of nuclear apoptosis.

As described above, AIF and CAD induced nuclear apoptosis independently from each other, both upon microinjection into intact cells (Fig. 2) and upon addition to purified nuclei (Fig. 3). ICAD inhibited CAD, yet did not antagonize AIF (Fig. 3C–E). Cytosolic extracts from wild-type MEFs undergoing apoptosis in response to STS, etoposide, or cisplatin were found to contain a biological activity which caused nuclear DNA loss in vitro upon addition to purified HeLa cells. Addition of ICAD failed to block this activity (Fig. 4A and Fig. B). Similarly, immunodepletion of AIF had no major inhibitory effect on such extracts (Fig. 4A and Fig. B). However, if AIF immunodepletion was combined with ICAD, apoptotic DNA loss was abolished (Fig. 4A and Fig. B). Cytosolic extracts from Apaf-1^−/− and caspase-3^−/− cells treated with different apoptosis inducers also contained an activity which induced nuclear apoptosis in the cell-free system (Fig. 4 B). In contrast to wild-type extracts, AIF depletion from Apaf-1^−/− and caspase-3^−/− extracts sufficed to inhibit the apoptosis-inducing activity, indicating that AIF is the principal factor responsible for nuclear apoptosis in such cells.

Apoptosis-associated chromatin condensation and degradation may be expected to serve two purposes, namely to facilitate shrinkage (which in turn facilitates heterophagic removal of the apoptotic cells) and to prevent the DNA of the dead cell, which may include viral genomes or mutated genes including oncogenes, from being incorporated into adjacent cells. As shown here, at least two parallel and redundant pathways lead to nuclear apoptosis. One of these pathways involves caspases, ICAD, and CAD, and leads to oligonucleosomal DNA fragmentation and advanced chromatin condensation. The second, caspase-independent, pathway involves AIF and leads to large-scale DNA fragmentation and peripheral chromatin condensation (Fig. 1,Fig. 2,Fig. 3). These pathways may be activated in a sequential fashion, as the AIF-induced phenotype of nuclear apoptosis (stage I) normally precedes that induced by CAD (stage II; Fig. 1 and Fig. 5) and large scale DNA fragmentation precedes internucleosomal DNA cleavage in several models of apoptosis 15,25,26,27. Moreover, both pathways can act in a redundant fashion as suggested by the fact that, in a cell-free system involving cytoplasmic extract as apoptosis inducer, nuclear apoptosis is only prevented when both CAD and AIF are inhibited (Fig. 4 and Fig. 5). Recently, the existence of another chromatin condensation factor, Acinus, which requires activation by caspase-3 and another, unknown, protease has been reported 5. It may be speculated that Acinus is not fully activated in the cytosolic extract system used herein or, alternatively, that Acinus requires activation by AIF and/or is inhibited by ICAD.

In apparent contrast to the data reported here and in reference 4, a caspase-resistant mutant of ICAD completely prevents STS-induced DNA fragmentation in Jurkat cells, even at the level of large-scale DNA fragmentation 28. This may imply that the AIF pathway was not activated (or blocked?) in this cell type. The fact that the contribution of AIF (or other factors) to nuclear apoptosis varies is also suggested by variations in the observed effect of caspase inhibitors on apoptotic nuclear ultrastructure in different experimental systems, ranging from suppression of most signs of chromatin condensation 13 to no inhibition at all 10,11,12. AIF is expressed in an ubiquitous fashion 14, suggesting that factors other than AIF expression, e.g., endogenous AIF inhibitors or the absence of AIF targets, must explain these differences. Microinjection of an anti-AIF antibody prevents chromatin condensation induced by atractyloside or staurosporine 14, whereas depletion of AIF from cytosolic extracts (of wild-type cells) does not suffice to prevent chromatin condensation induced by such extracts (Fig. 4). This apparent paradox may reflect the fact that AIF is released from mitochondria before Cyt-c, correlating with stage I of apoptosis 15. Thus, when choosing a suitable time frame in short-term microinjection experiments, neutralization of AIF prevents the appearance of stage I of apoptosis. Moreover, AIF is a factor that causes the release of Cyt-c 14. Thus, AIF neutralization may be expected to retard the release of Cyt-c and the subsequent caspase/CAD activation that leads to stage II chromatin condensation.

Irrespective of these theoretical considerations, our data indicate the existence of at least two pathways leading to chromatin condensation and degradation during apoptosis. Why the process of apoptotic chromatin condensation is so complex and whether these pathways are connected at the molecular level by a common action on sessile nuclear proteins with proapoptotic potential remains an open question for future investigation.

We thank Anu Srinivasan (Idun Pharmaceuticals, San Diego, CA) for the CM1 caspase-3 antibody and Christine Schmitt for expert technical assistance.

L. Ravagnan and K.F. Ferri received Ph.D. fellowships from the French Ministry of Science & Technology; M. Loeffler received a postdoctoral fellowship from the Austrian Science Foundation; and P. Costantini received a fellowship from the Fondation pour la Recherche Medicale (FRM). This work has been supported by a special grant of the Ligue Nationale contre le Cancer, as well as by grants from Agence National de Recherche sur la SIDA, FRM (to G. Kroemer), Assistance Publique-Hôpitaux de Paris and Caisse Nationale Assurance Maladie (CANAM; contract 98006 to E. Daugas), and the Wellcome Trust (to W.C. Earnshaw).