APOPTOSIS II: BEYOND CYTOCHROME C; THE OTHER MITOCHONDRIAL PROTEINS

Contents

• Cytochrome c
• SMAC/Diablo
• HtrA2/Omi
• Apoptosis Inducing Factor
• Endonuclease G
• References

Since the pioneering work of Wang and coworkers identified Cytochrome c as a key component of the apoptosis pathway,¹ mitochondria have been central to apoptosis research.

Figure 1. Pro-apoptotic stimuli elicit apoptosis via activation of pro-apoptotic Bcl-2 family members. These proteins act on the mitochondria causing it to release several proteins from the intermembrane space into the cytosol or nucleus. These pro-apoptotic mitochondrial proteins include Cytocrome c, SMAC/Diablo, HtrA2/Omi, AIF and Endo G. Cytochrome c along with APAF-1 form the apoptosome, which functions to activate Caspase-9, which subsequently activates Caspase-3. SMAC/Diablo and HtrA2/Omi function to neutralize the inhibitory effect of IAPs on Caspases thereby allowing Caspase activity. AIF binds chromosomal DNA and causes chromatin condensation and remodeling, which facilitates DNA fragmentation by nucleases such as Endo G.

However, Cytochrome c is not the only mitochondrial protein that plays a vital role in the regulation of cell death. SMAC/Diablo, HtrA2/Omi, AIF, and Endonuclease G have also been identified as proteins with important pro-apoptotic functions.^2,3

Cytochrome c

Cytochrome c has been studied extensively, not only for its role in electron transport, but also for its role in apoptosis.² In brief, Cytochrome c is released from the mitochondria in response to specific apoptotic stimuli via Bcl-2 family-regulated mechanisms. Each of the Bcl-2 family members is either pro- (e.g. Bad, BAK, Bax, tBID, BIM, etc.) or anti-apoptotic (e.g. A1, Bcl-2, Bcl-w, Bcl-xL, etc.) in function and it is the balance between these factors that influences Cytochrome c release and therefore apoptosis.⁴ Once released, Cytochrome c binds to apoptosis protease activating factor-1 (APAF-1) complexed with dATP to form the oligomeric apoptosome complex. Procaspase-9 is recruited to the apoptosome, via its caspase recruitment domain (CARD), where it is activated and released.^5-7 Caspases are a family of cysteine proteases expressed as latent zymogens, requiring cleavage for activation. Active initiator Caspases, such as Caspase-9, can amplify the apoptotic cascade through the cleavage and activation of other effector caspases, such as Procaspase-3 and -7, and initiate orderly dismantling of the cell through proteolytic cleavage of other cellular substrates (Figure 1).⁸

SMAC/Diablo

Strict regulation of caspase activation is necessary to prevent inadvertently committing a cell to apoptosis. Once activated, caspase function is further regulated by a family of specific inhibitory proteins. Inhibitor of apoptosis proteins (IAPs) possess one or more BIR (baculoviral IAP repeat) domains and directly bind and inhibit active caspases. XIAP, cIAP, and cIAP-2 each possess 3 BIR domains (BIR1, BIR2, and BIR3) as well as a RING domain (cIAP-1 and -2 also have a CARD domain). XIAP binds and inhibits Caspase-9 via its BIR3 domain and Caspase-3 and -7 via its BIR2 and intervening linker domains.^8,9 The IAPs, however, may be inhibited themselves, thereby restoring caspase activity. This neutralization of caspase IAP inhibition is accomplished by a family of proteins first identified in Drosophila as Reaper,¹⁰ HID,¹¹ Grim,¹² and Sickle.^13-15 These IAP inhibitors possess a common N-terminal, 4-amino acid (aa) sequence called a Reaper or IAP-binding motif (Figure 2). The IAP-binding motif specifically interacts with the BIR domains of IAPs to facilitate IAP/IAP inhibitor complex formation, and thereby prevent IAP/Caspase complex formation.¹⁶ An IAP-binding motif has also been identified in Caspase-9 (Figure 2). IAP BIR domain binding to caspase and to IAP inhibitor is mutually exclusive suggesting that the two types of complexes, one anti-apoptotic (IAP/Caspase) and the other pro-apoptotic (IAP/IAP inhibitor), are in equilibrium with each other.¹⁷ While true mammalian homologs of Drosophila IAP inhibitors have yet to be identified, two functional analogs have been described.¹⁸

The first mammalian IAP inhibitor identified was the mitochondrial protein, SMAC/Diablo. The search for SMAC/Diablo began with the observation that detergent-isolated cell extracts possessed a greater ability to activate Caspase-3 than non-detergent extracts. However, this detergent-soluble factor could only activate Caspase-3 in the presence of APAF-1, Cytochrome c, and Procaspase-9. Purification of this factor from solubilized membrane extracts led to the identification of human SMAC (second mitochondria-derived activator of caspases).¹⁹ The mouse homolog of SMAC was identified by co-immunoprecipitation with XIAP and named Diablo (direct IAP binding protein with low pI).²⁰ SMAC/Diablo is expressed as a 239 (237 in mouse) aa precursor protein with an N-terminal mitochondrial localizing sequence (MLS). Upon translocation to the mitochondria, the N-terminal 55 aa are proteolytically removed to generate mature, 25 kDa SMAC/Diablo.^19,20 However, the apparent molecular weight of mature SMAC/Diablo is 100kDa upon size exclusion chromatography.¹⁹ Subsequent crystallographic analysis revealed that the mature species forms an elongated, symmetric homodimer through hydrophobic interactions causing it to exhibit a much greater apparent size than a globular protein of similar molecular weight.²¹

Removal of the N-terminal MLS upon SMAC/Diablo entry into the mitochondria reveals an N-terminal sequence (AVPI) that has significant similarity to the IAP binding motif found in the Drosophila IAP inhibitor proteins Reaper, HID, Grim, and Sickle and mammalian Caspase-9 (Figure 2).^21,22 Exposure of the intact IAP-binding motif is an absolute requirement for SMAC/Diablo apoptotic function. Immature, but intact^19,20 SMAC/Diablo and mature, but N-terminally mutated²¹ or deleted²² SMAC/Diablo are incapable of caspase activation. SMAC/Diablo interacts directly with XIAP, cIAP-1, and cIAP-2 BIR domains via its IAP binding motif.^21,22 While SMAC/Diablo can interact with both XIAP BIR2 and BIR3 domains, it has strongest affinity for XIAP BIR3.^22,23 The interaction of SMAC/Diablo and XIAP BIR3 domain was further investigated by NMR²³ and co-crystallization²⁴ studies, which indicated that the N-terminal 4 aa of the SMAC/Diablo IAP-binding motif specifically interacts with a surface groove on XIAP BIR3. SMAC/Diablo dimer formation is also critical for function as introduction of missense mutations affecting the dimer interface disrupted dimer formation and decreased interaction with XIAP.²¹ Since XIAP inhibits Caspase-9 via its BIR3 domain and Caspase-3 via its BIR2 domain and since SMAC/Diablo inhibits XIAP by binding its BIR2 and BIR3 domains, it is not surprising that SMAC/Diablo allows Caspase-9 and -3 activity.^21,22

XIAP, cIAP-1, and cIAP-2, however, are equipped with the ability to eliminate SMAC/Diablo binding. The IAP RING domain can function as an E3 ubiquitin-protease ligase with specificity for SMAC/Diablo. The IAPs cause the ubiquitination of SMAC/Diablo thereby targeting it for proteosomal degradation.^25,26 This mechanism is observed in Drosophila as well. Drosophila IAP (DIAP) can ubiquitinate Reaper, HID, and Grim causing their degradation and preventing apoptosis.^26,27 This mechanism may represent a means of protecting cells from undergoing apoptosis in response to accidental mitochondrial damage.²⁵

SMAC/Diablo translocates from the mitochondria to the cytosol in response to apoptotic stimuli via a mechanism regulated by pro-apoptotic Bcl-2 family members.^19,20 The precise mechanism and kinetics of SMAC/Diablo release remain unclear. While some results indicate that Cytochrome c and SMAC/Diablo are released simultaneously from the mitochondria,²⁸ other data suggest that there is a 4-fold longer rate of SMAC/Diablo release relative to Cytochrome c (Figure 1).²⁹

HtrA2/Omi

The second mammalian IAP inhibitor to be discovered was the mitochondrial protein HtrA2, also known as Omi. It was identified by five independent groups as a protein that could specifically bind XIAP.^30-34 Human HtrA2/Omi had been described previously as a protein homologous to the bacterial high temperature requirement protein A, HtrA.^35-37 Bacterial HtrA functions both as a chaperone protein and as a serine protease responsible for cleavage of denatured proteins under elevated temperature.³⁸ Similarly, HtrA2/Omi is involved in the cellular response to thermal and oxidative stress. It is expressed as a 50 kDa precursor protein with an N-terminal MLS. Serine protease activity allows HtrA2/Omi to autocatalytically remove its N-terminal 133 aa generating the mature 36 kDa protein.³⁷ HtrA2/Omi is expressed as a 50 kDa precursor protein with an N-terminal MLS. Upon translocation to the mitochondria, HtrA2/Omi cleaves to generate the mature 36 kDa protein.^30-34 Like Cytochrome c and SMAC/Diablo, mature HtrA2/Omi localizes to the intermitochondrial membrane space³¹ and translocates to the cytosol in response to a diverse set of apoptotic stimuli via pro-apoptotic Bcl-2 family members.^30-34

Cleavage of the MLS exposes an N-terminal sequence (AVPS) with significant similarity to the IAP-binding motif of Drosophila Reaper, HID, and Grim, and mammalian Caspase-9 and SMAC/Diablo (Figure 2).^30-34 Only the mature form of HtrA2/Omi can bind XIAP.^30,32,33 Additionally, mutation^30,31,39 or deletion^31,32,39 of the IAP-binding motif renders HtrA2/Omi incapable of binding XIAP. HtrA2/Omi interacts directly with cIAP-1 and -2, and with the BIR2 and BIR3 domains of XIAP.^31,33 In contrast to SMAC/Diablo, the affinity of HtrA2/Omi for the BIR2 domain of XIAP is greater than the BIR3 domain.^33,39

At least three pieces of evidence suggest that the serine protease activity of HtrA2/Omi may be of equal or greater importance than IAP binding in mediating its apoptotic function. First, HtrA2/Omi mutants that are unable to bind XIAP are still capable of inducing apoptosis.^30,31,33,39 Second, the bovine homolog of HtrA2/Omi does not possess an IAP-binding motif and yet is apparently functional.³⁹ Third, HtrA2/Omi is still able to induce cell death in the presence of Caspase inhibitors, and APAF-1 and Caspase-9 null mutations.³¹ Upon subsequent examination of the serine protease activity of HtrA2/Omi, it was discovered that this quality is required for its pro-apoptotic effects. Mutants that cannot bind XIAP, but have intact serine protease activity can induce apoptosis,^30,31,33,39 while mutants that can bind XIAP, but are proteolytically inactive cannot.^33,39 Crystallographic structural analysis shows that HtrA2/Omi forms a homotrimer with an apparent molecular weight of 110 kDa.³⁹ The active site serine residue is located in the center of the folded HtrA2 molecule. Mutation of the active site serine residue in the HtrA2/Omi homotrimer does not affect XIAP binding^31,39 and XIAP binding does not affect HtrA2/Omi homotrimer serine protease activity.^30,33 However, disruption of the homotrimer interface destroys its serine protease activity and renders HtrA2/Omi incapable of inducing apoptosis.³⁹ Taken together, these findings suggest that while HtrA2/Omi is capable of binding IAPs, it is not necessary for its pro-apoptotic function. Rather, HtrA2/Omi induces apoptosis through an alternative mechanism as well that is serine protease activity-dependent and caspase-independent.

Recently, a novel mechanism for HtrA2/Omi-IAP interaction was reported. During p53-dependent apoptosis in response to DNA damage, HtrA2/Omi cleaves cIAP-1.⁴⁰ cIAP-1 cleavage is inhibited by a general serine protease inhibitor, but not by a pan caspase inhibitor suggesting that the activity is serine protease- specific and caspase-independent.^40,41 HtrA2/Omi is capable of cleaving other IAPs including XIAP.⁴¹ Cleavage of cIAP-1 by HtrA2/Omi is partially dependent on the N-terminal IAP-binding motif as an 8 aa deletion reduced cleavage efficiency.⁴¹ HtrA2/Omi RNA interference resulted in a loss of cIAP-1 cleavage and decreased sensitivity to apoptosis-inducing stimuli. Overexpression of a cleavage resistant cIAP-1 had a similar effect.⁴¹ cIAP-1 cleavage significantly and irreversibly reduces its ability to inhibit and ubiquitinate caspases.^40,41 Collectively, these data suggest that this serine protease-specific, caspase- independent mechanism of IAP neutralization by cleavage may be the principle means of HtrA2/Omi pro-apoptotic function (Figure 1).

Apoptosis Inducing Factor

Apoptosis is characterized by changes in the nucleus including condensation, fragmentation, and laddering of chromosomal DNA mediated in part by nucleases. Nuclear DNA is first cut at A/T-rich sites in nuclear scaffold regions to form variably large (50 to 300 kb) fragments. These fragments are subsequently cut at internucleosomal spacer regions to form small, similar-sized (~180 bp) pieces in process known as DNA laddering.⁴²

Apoptosis Inducing Factor (AIF) was identified as a mitochondrial protein that is associated with initial chromatin condensation and high molecular weight DNA fragmentation, but not DNA laddering.⁴³ Mature AIF is a flavoprotein with C-terminal domain sequence and structural similarity to bacterial oxidoreductases.^43-45 In particular, the C-terminal domain of AIF is similar to BphA4, a bacterial NADH-dependant ferredoxin reductase.^44,45 AIF possesses three distinct domains: a FAD-binding domain (D1), an NADH-binding domain (D2), and a C-terminal domain (D3) that in mammalian species includes a long insertion that forms an open loop structure that may be involved in turnover and/or protein-protein interactions.^44,45 Truncation and mutational studies indicate that while FAD binding is required for the oxidoreductase activity,⁴⁶ neither FAD nor oxidoreductase activity are necessary for the pro-apoptotic function of AIF.^43,46 These data suggest that like Cytochrome c, AIF is a bifunctional protein that acts as both a mitochondrial electron transferase and an effector of apoptosis.^43-45

AIF is synthesized as a 67 kDa protein with an N-terminal MLS, which is proteolytically removed upon translocation into the mitochondria to generate the mature 57 kDa protein.⁴³ AIF localizes to the intermitochondrial membrane space and, like Cytochrome c, SMAC/Diablo, and HtrA2/Omi is released from the mitochondria in response to specific apoptotic stimuli via pro-apoptotic Bcl-2 family members.^43,47 However, unlike Cytochrome c, SMAC/Diablo, and HtrA2/Omi, mature AIF has a nuclear targeting sequence and hence translocates to the nucleus in a caspase-independent manner where it is involved in initial stages of chromatin condensation and large-scale DNA fragmentation.^43,47,48

Since AIF itself possesses no nuclease activity,^43,48 the exact mechanism by which it exerts its pro-apoptotic effects is unknown. One possibility was raised by the observation of a potential DNA-interaction domain. While no such domain was observed in the mouse AIF structure,⁴⁴ a large portion of the surface of human AIF (between D1 and D3) forms a groove with a positive electrostatic charge suggestive of a sequence-independent DNA-binding domain.⁴⁵ Mutation of individual positively charged residues contributing to this surface results in reduced or lost AIF DNA binding capability. Further, AIF mutants defective in DNA-binding, fail to induce apoptosis. It has been hypothesized that AIF may cause initial chromatin condensation by binding DNA, displacing chromatin-associated proteins, and subsequently causing DNA fragmentation actively, by recruiting nucleases, and/or passively, by increasing DNA susceptibility to nucleases (Figure 1).⁴⁵

Still, many aspects of AIF-associated apoptotic function remain unclear. There is some evidence that AIF may also possess anti-apoptotic qualities. Its redox-active region may act as a free radical scavenger preventing oxidative stress-induced apoptosis.⁴⁹ Further, it appears that Heat shock protein 70 (HSP70) can bind to AIF, perhaps at its C-terminal loop insertion, and prevent chromatin condensation.^45,50 HSP70 was already known as an inhibitor of apoptosis via its ability to bind APAF-1 thereby preventing formation of the apoptosome and activation of Procaspase-9.^51,52 However, overexpression of HSP70 in APAF-1- cells still prevents apoptosis, indicating HSP70 can also block apoptosis via another mechanism.⁵⁰

Endonuclease G

One of the nucleases responsible for both high molecular weight DNA fragmentation and DNA laddering is Caspase-activated deoxyribonuclease (CAD), also known as DNA fragmentation factor 40 (DFF40). CAD/DFF40 forms a heterodimer in the cytosol with inhibitor of CAD (ICAD), also known as DFF45. ICAD/DFF45 is important not only for the inhibition of CAD/DFF40 nuclease activity, but also for maintaining CAD/DFF40 stability in the cytosol. Upon apoptosis initiation, Caspase-3, cleaves ICAD/DFF45, freeing the CAD/DFF40 nuclease, which subsequently translocates to the nucleus and elicits DNA fragmentation and laddering.⁴² However, this caspase-dependent pathway is not the only means by which DNA is fragmented during apoptosis. Indeed, large-scale DNA fragmentation still occurs in caspase-resistant ICAD/DFF45 expressing mice⁵³ and in CAD/DFF40- suggesting the existence of an additional apoptosis-associated, but caspase-independent nuclease.

The initial observations leading to the identification of such a factor included the ability of mitochondrial supernatants to induce apoptotic morphological changes in isolated nuclei,⁵⁵ and the ability of mitochondria to release factors with DNase activity upon pro-apoptotic Bcl-2 family member treatment.^56,57 Subsequently, a mitochondrial DNase was purified and identified as Endonuclease G (Endo G).^56,57 Endo G is released from the mitochondria in a pro-apototic Bcl-2 family-dependent and caspase-independent manner after which it translocates to the nucleus where it cleaves DNA into large (50 to 300 kb) fragments.^56,57 Further examination of the action of Endo G on various nucleic acid substrates confirms that it elicits DNA fragmentation and can also contribute in some contexts to DNA laddering, likely due to cooperation with DNase I,⁵⁸ a known agent of DNA laddering.⁵⁹

Endo G had been identified previously in nuclei and mitochondria of chicken erythrocytes,⁶⁰ bovine heart^61,62 and thymus,^63,64 and rat liver⁶⁵ as a 26 kDa DNase that forms a homodimer and prefers long stretches of G/C sequences as substrates. The mouse and human homologs of Endo G were later identified as nuclear-encoded proteins that possess MLSs, which are cleaved upon translocation into the mitochondria where the mature enzymes then form dimers.⁶⁶ Due to the apparent lack of a nuclear localizing sequence in Endo G and the G/C-rich nature of the mitochondrial DNA replication origin, Endo G was originally thought to be involved in mitochondrial DNA replication.^64,66,67 However, several lines of evidence dispute such a function. For instance, the apoptosis-induced release of Endo G from the mitochondria,^56,57 suggests that it is compartmentalized in the intermembrane space and not in the matrix where mitochondrial DNA is located. Indeed, closer examination of Endo G localization reveals that it is restricted to intermembrane space and is not present in the matrix.⁶⁸

There is some evidence, at least in C. elegans, that Endo G and AIF work together to affect the nuclear aspects of apoptosis. The C. elegans Endo G homolog, Cps-6, is also a mitochondrial protein that causes DNA fragmentation when translocated into the nucleus.⁶⁹ Csp-6 acts synergistically with Wah-1, the caspase-dependent C. elegans homolog of AIF,⁷⁰ in promoting DNA degradation.⁶⁹ Mammalian Endo G may work synergistically with AIF to elicit the nuclear effects of apoptosis as well. In this model, AIF binds DNA to initiate chromatin condensation, thereby, directly or indirectly allowing Endo G to process the DNA into large fragments (Figure 1).

References

1. Liu, X. et al. (1996) Cell 86:147.
2. Wang, X. (2001) Genes Dev. 15:2922.
3. van Gurp, M. et al. (2003) Biochem. Biophys. Res. Commun. 304:487.
4. Scorrano, L. & S.J. Korsmeyer (2003) Biochem. Biophys. Res. Commun. 304:437.
5. Rodriguez, J. & Y. Lazebnik (1999) Genes Dev. 13:3179.
6. Salvesen, G.S. & V.M. Dixit (1999) Proc. Natl. Acad. Sci. USA 96:10964.
7. Zou, H. et al. (1999) J. Biol. Chem. 274:11549.
8. Shi, Y. (2002) Mol. Cell 9:459.
9. Deveraux, Q.L. & J.C. Reed (1999) Genes Dev. 13:239.
10. White, K. et al. (1994) Science 264:677.
11. Grether, M.E. et al. (1995) Genes Dev. 9:1694.
12. Chen, P. et al. (1996) Genes Dev. 10:1773.
13. Srinivasula, S.M. et al. (2002) Curr. Biol. 12:125.
14. Wing, J.P. et al. (2002) Curr. Biol. 12:131.
15. Christich, A. et al. (2002) Curr. Biol. 12:137.
16. Claveria, C. & M. Torres (2003) Biochem. Biophys. Res. Commun. 304:531.
17. Srinivasula, S.M. et al. (2001) Nature 410:112.
18. Vaux, D.L. & J. Silke (2003) Biochem. Biophys. Res. Commun. 304:499.
19. Du, C. et al. (2000) Cell 102:33.
20. Verhagen, A.M. et al. (2000) Cell 102:43.
21. Chai, J. et al. (2000) Nature 406:885.
22. Srinivasula, S.M. et al. (2000) J. Biol. Chem. 275:36152.
23. Liu, Z. et al. (2000) Nature 408:1004.
24. Wu, G. et al. (2000) Nature 408:1008.
25. MacFarlane, M. et al. (2002) J. Biol. Chem. 277:36611.
26. Hu, S. & X. Yang (2003) J Biol. Chem. 278:10055.
27. Olson, M.R. et al. (2003) J Biol. Chem. 278:4028.
28. Madesh, M. et al. (2002) J. Biol. Chem. 277:5651.
29. Springs, S.L. et al. (2002) J. Biol. Chem. 277:45715.
30. Suzuki, Y. et al. (2001) Mol. Cell 8:613.
31. Hedge, R. et al. (2002) J. Biol. Chem. 277:432.
32. Martins, L.M. et al. (2002) J. Biol. Chem. 277:439.
33. Verhagen, A.M. et al. (2002) J. Biol. Chem. 277:445.
34. van Loo, G. et al. (2002) Cell Death Differ. 9:20.
35. Faccio, L. et al. (2000) J. Biol. Chem. 275:2581.
36. Gray, C.W. et al. (2000) Eur. J. Biochem. 267:5699.
37. Savopoulos, J.W. et al. (2000) Protein Expr. Purif. 19:227.
38. Speiss, C. et al. (1999) Cell 97:339.
39. Li, W. et al. (2002) Nat. Struct. Biol. 9:436.
40. Jin, S. et al. (2003) Genes Dev. 17:359.
41. Yang, Q.-H. et al. (2003) Genes Dev. 17:1487.
42. Nagata, S. et al. (2003) Cell Death Differ. 10:108.
43. Susin, S.A. et al. (1999) Nature 397:441.
44. Mate, M.J. et al. (2002) Nat. Struct. Biol. 9:442.
45. Ye, H. et al. (2002) Nat. Struct. Biol. 9:680.
46. Miramar, M.D. et al. (2001) J. Biol. Chem. 279:16391.
47. Daugas, E. et al. (2000) FASEB J. 14:729.
48. Susin, S.A. et al. (2000) J. Exp. Med. 192:571.
49. Klein, J.A. et al. (2002) Nature 419:367.
50. Ravagnan, L. et al. (2001) Nat. Cell Biol. 3:839.
51. Beere, H.M. et al. (2000) Nat. Cell Biol. 2:469.
52. Saleh, A. et al. (2000) Nat. Cell Biol. 2:476.
53. McIlroy, D. et al. (2000) Genes Dev. 14:549.
54. Samejima, K. et al. (2001) J. Biol. Chem. 276:45427.
55. Jiang, Z.F. et al. (2000) Cell Res. 10:221.
56. Li, L.Y. et al. (2001) Nature 412:95.
57. van Loo, G. et al. (2001) Cell Death Differ. 8:1136.
58. Widlak, P. et al. (2001) J. Biol. Chem. 276:48404.
59. Oliveri, M. et al. (2001) Eur. J. Immunol. 31:743.
60. Ruiz-Carrillo, A. & J. Renaud (1987) EMBO J. 6:401.
61. Cummings, O. et al. (1987) J. Biol. Chem. 262:2005.
62. Gerschenson, M. et al. (1995) Nucleic Acids Res. 23:88.
63. Cote, J. et al. (1989) J. Biol. Chem. 264:3301.
64. Cote, J. & A. Ruiz-Carrillo (1993) Science 261:765.
65. Low, R.L. et al. (1988) Nucleic Acids Res. 16:6427.
66. Low, R.L. et al. (1987) J. Biol. Chem. 262:16164.
67. Ohsato, T. et al. (2002) Eur. J. Biochem. 269:5765.
68. Prats, E. et al. (1997) DNA Cell Biol. 16:1111.
69. Parrish, J. et al. (2001) Nature 412:90.
70. Wang, X. et al. (2002) Science 298:1587.