Centre for Cellular and Molecular Biology, Uppal Road, Hyderabad 500 007, India

*Corresponding author (Fax, 91-40-7171195; Email, kas@ccmb.ap.nic.in).

This article reviews the research on the inner nuclear membrane protein lamin B receptor (LBR). It focuses on the biochemical and immunological evidence for an LBR; the cloning of chicken, rat and human LBR cDNAs and genomic sequences; the lamin B-, chromatin- , DNA- and NLS-binding properties of the N-terminal domain and its phosphorylation by different kinases; the sterol C-14 reductase activity of the C-terminal domain; the use of yeast two-hybrid screens and co-immunoprecipitation to identify interacting proteins; and the probing of nuclear assembly and disassembly in living cells with LBR-GFP fusion proteins. The article concludes by considering a scenario whereby LBR levels might even regulate gene expression.

1. Introduction

The lamin B receptor (LBR) is an integral protein of the inner nuclear membrane of metazoan cells. In the literature it is also referred to by its acronym LBR, and by its apparent molecular weight in SDS-PAGE, p58. However not all usages of p58 refer to the LBR, (e.g., Ferrini et al 1994). The first paper on LBR appeared a little over ten years ago (Worman et al 1988). Today, LBR is recognized as possibly playing a central role in the mitosis-related disassembly and reassembly of the nuclear envelope (for a review see Gant and Wilson 1997). This article will focus on the discovery of diverse properties of LBR and conclude by speculating on whether LBR might also regulate gene expression.

The nuclear envelope is comprised of two membranes; the outer and inner nuclear membranes. The space between the two is called the perinuclear space. Numerous transcisternal connections are formed between these two membranes by the nuclear pores. The outer nuclear membrane is continuous with the endoplasmic reticulum (ER) and, like the ER, it is studded with ribosomes on its cytoplasmic surface. The perinuclear space is thus continuous with the ER lumen. The inner nuclear membrane is lined on its nucleoplasmic side by a filamentous meshwork, the nuclear lamina, whose major constituents are the intermediate filament proteins called lamins. There are essentially two kinds of lamins; the A- and C-lamins are neutral and the B-lamins are acidic. Lamins A and C arise from the same gene by alternative RNA splicing (Lin and Worman 1993); whereas there are two different lamin B proteins in somatic mammalian cells, lamin B1 and lamin B2, that are encoded by different genes. During mitosis the A- and C-lamins become soluble but the B-lamins remain associated with the intracellular membrane vesicles that are thought to be the remnants of the nuclear membranes. This suggested that a receptor might couple the B-lamins to the inner nuclear membrane. The nuclear lamina is a discontinuous structure and where it is disrupted the chromatin appears to be directly associated with the inner nuclear membrane. It turns out that LBR can bind to chromatin as well, thus LBR is both well positioned and functionally equipped to mediate interactions at the chromatin–lamina–membrane interface of the nucleus.

2. The biochemical evidence for an LBR

Avian erythrocytes are the tissue of choice for the isolation of nuclear membrane proteins because they contain nuclei but few other intracellular membranes. Worman et al (1988) extracted a turkey erythrocyte nuclear envelope preparation with 8 M urea to remove the peripherally bound lamins and used the lamin-depleted nuclear envelopes in binding assays with purified [¹²⁵I]-labelled lamin B. The nuclear envelopes bound the labelled lamin B in a saturable and specific manner. This binding was significantly greater than that of purified [¹²⁵I]-labelled lamin A and it was competitively inhibited by unlabelled lamin B but not by unlabelled lamin A. The plasma membrane did not show specific binding with labelled lamin B. Scatchard analysis revealed that labelled lamin B bound to the lamin-depleted nuclear envelopes with a Kd of approximately 0·2 mM and the LBR concentration was approximately 50% of the envelope protein.

The labelled lamin B bound to a 58 kDa protein (p58) in Western blots of the urea-extracted nuclear envelope proteins. This protein was absent from the urea- and alkaline-extracts of the nuclear envelopes but it could be solubilized by a mixture of 2% Triton X-100 and 2 M KCl. This showed that p58 is an integral membrane protein. Antibodies raised against purified p58 blocked the binding of labelled lamin B to the lamin-depleted nuclear envelopes and thereby confirmed that p58 and LBR are indeed the same protein. Indirect immunofluorescence microscopy with the anti-p58 antibodies showed a nuclear rim staining indicating that p58 is primarily localized in the inner nuclear membrane. The presence of a membrane receptor for lamin B provided an explanation for why lamin B unlike lamins A and C remains membrane bound during mitosis.

Incubation of turkey erythrocytes with ³²Pi followed by analysis of the proteins in the urea extracted nuclear envelopes by SDS-PAGE and autoradiography revealed a 58 kDa phosphoprotein that comigrated with the p58 antigen in a parallel immunoblot (Appelbaum et al 1990). This indicated that p58 is phosphorylated in vivo. To examine whether dephosphorylation of p58 affected its binding to lamin B, nuclear envelope proteins were separated by SDS-PAGE and transferred to replica nitrocellulose membranes. One membrane was dephosphorylated in situ with calf intestinal alkaline phosphatase and the other was mock treated and then both were incubated with [¹²⁵I]-lamin B. The p58 in the alkaline phosphatase treated filter bound less labelled lamin B relative to the mock treated filter. To verify that this reduction was not due to proteolysis of p58 during incubation with alkaline phosphatase the p58 bands were eluted from the filters, re-electrophoresed in a gel and immunoblotted with anti-p58 antibodies.

To further explore the regulation of the membrane-lamina interaction by phosphorylation, Courvalin et al (1992) immunoprecipitated in vivo labelled LBR from interphase and mitosis-arrested chicken cells. They found that LBR was phosphorylated only on Ser residues in interphase cells whereas in mitotic cells it was phosphorylated on Ser and Thr residues, and some Ser residues phosphorylated in interphase were not phosphorylated in mitosis. Additionally LBR was shown to be a substrate for a p34^cdc2-type mitosis specific kinase. This strengthened the hypothesis that mitotic phosphorylation of LBR might contribute to the dissociation of the lamina from the inner nuclear membrane. A subsequent section describes an LBR kinase that was co-immunoprecipitated with LBR. Phosphorylation by this kinase affects the association of LBR with other nuclear proteins.

3. The immunological evidence for an LBR

Human patients with autoimmune disorders often contain in their sera antibodies to nuclear antigens. Primary biliary cirrhosis (PBC) is a chronic autoimmune disease of the liver and is characterized by progressive inflamatory destruction of the intrahepatic bile ducts with eventual progression to cirrhosis. A subset of patients with PBC were found to harbour antibodies to turkey erythrocyte LBR (Courvalin et al 1990). This implied that the autoantigen, human LBR, is conserved between birds and mammals. Subsequently Lassoued et al (1991) used ELISA to show that the anti-LBR autoantibodies inhibited the binding to lamins of anti-lamin B autoantibodies but did not affect the binding to lamins of anti-lamin A and C antibodies. This provided additional evidence that LBR is indeed a specific receptor for lamin B and suggested that anti-LBR antibodies bear the internal image of the binding site of lamin B to LBR.

4. Cloning of LBR cDNAs and genomic sequences

Degenerate oligonucleotides were designed based on the N-terminal amino acid microsequence of purified turkey erythrocyte LBR and used to perform RT-PCR with chicken liver poly A enriched RNA (Worman et al 1990). An amplified product of the expected size was obtained and the amino acid sequence derived from it matched that of the turkey derived microsequence. A synthetic oligonucleotide corresponding to a portion of the sense strand of the PCR product was used to screen a chicken liver cDNA library. Overlapping cDNA clones were obtained and their nucleotide sequences were determined to obtain the primary structure of chicken LBR. The human cDNA clones were subsequently isolated by probing a HeLa cell cDNA library with the chicken LBR cDNA at reduced stringency (Ye and Worman 1994). The chicken and human cDNAs encoded proteins of 637 and 615 amino acid residues, respectively, that shared 68% amino acid sequence identity throughout their entire length.

The human cDNA was used to isolate genomic clones to determine the structure of the human LBR gene (Schuler et al 1994). This gene spans 35 kb and has 13 coding exons. Fluorescence in situ hybridization (FISH) localized the LBR gene to cytogenetic location 1q42·1 (Wydner et al 1996). This location is close to FRA1H a common 5-azacytidine fragile site.

5. LBR is also a nuclear localization signal
binding protein

The cDNA for rat LBR was cloned somewhat serendipitously by a group studying nuclear envelope proteins that could bind to nuclear localization signals (NLS). Haino et al (1993) resolved a rat liver nuclear extract by SDS-PAGE and ligand blotted it with [¹²⁵I]-labelled nucleoplasmin, a protein with a strong NLS. A 60 kDa protein was detected and designated NBP60 (for NLS binding protein 60 kDa). NBP60 also bound to the NLS of SV40 large T-antigen (T-peptide) conjugated to human serum albumin. The binding of NBP60 to nucleoplasmin-sepharose was specific because it was inhibited by the wild-type T-peptide and not by a mutant T-peptide in which an essential Lys residue was replaced with Thr. NBP60 could be extracted with 2% Triton X-100, 1 M KCl but not with 1 M KCl-2 M urea or 2% Triton X-100 and it partitioned into the lower layer in a 2-phase system using Triton X-114. These results showed that NBP60 is an integral membrane protein. NBP60 was purified by nucleoplasmin-sepharose affinity chromatography and hydroxyapatite HPLC and on the basis of its partial amino acid sequence PCR primers were designed to isolate clones from a cDNA library. The nucleotide sequence of the cDNA revealed that NBP60 is a 620 amino acid protein with 79% and 63% identity to human and chicken LBR (Kawahire et al 1997). In other words, NBP60 is
rat LBR.

6. LBR is comprised of two structural domains with different functions

All three LBR proteins for which the deduced primary structure is known (chicken, rat and human) have two identifiable domains. The amino terminal one-third of the protein (approximately 200 amino acid residues) is hydrophilic and basic whereas the carboxyl terminal two-thirds (approximately 400 residues) is quite hydrophobic and contains 8 putative transmembrane segments. The amino terminal domain (NTD) contains consensus sites for phosphorylation by protein kinase A, the mitosis specific kinase p34^cdc2, and a stretch that is rich in Arg-Ser dipeptides (RS). It also contains DNA binding motifs that are common to several gene regulatory proteins and histones. Remarkably, the carboxyl terminal domains (CTD) are highly similar (~ 40% amino acid sequence identity) to the ergosterol biosynthetic enzyme sterol C-14 reductase of fungi and yeast (Papavinasasundaram and Kasbekar 1994; Schuler et al 1994; Ye and Worman 1994; D P Kasbekar, unpublished results).

The hydrophilic nature of the LBR NTD made it a more convenient candidate (compared to the CTD) for study as a recombinant fusion protein with glutathione S transferase (GST). GST-NTD fusion proteins were expressed in Escherichia coli and were purified by affinity chromatography on glutathione-sepharose followed by elution with reduced glutathione (Smith and Johnson 1988). The human LBR NTD-GST fusion protein was used to affinity purify the anti-LBR autoantibodies from PBC sera, and nuclear rim staining with these antibodies was confirmed by immunofluorescence microscopy (Ye and Worman 1994). The fusion proteins were also used to concentrate rat lamin B from a dilute solution. In other experiments the fusion proteins were shown to retard the electrophoretic migration of double stranded M13 DNA in agarose gels. South-western blot analysis of fusion proteins containing different stretches of the NTD with [³²P]-labelled l DNA revealed that the stretch between Pro 2 and Ala 100 is required for DNA binding.

As mentioned earlier rat LBR was identified in a search for NLS binding proteins. To verify that human LBR also possesses NLS binding activity Kawahire et al (1997) tested GST-NTD fusion proteins for binding to labelled T-peptide. In this way the NLS-binding activity was localized to residues 1–89 which contains the RS (Arg-Ser) rich region. This binding was competitively inhibited by unlabelled T-peptide but not by a mutant T-peptide. The significance of LBR’s NLS-binding activity remains to be determined. As described later the RS rich region is the target for LBR kinase and for binding to the splicing factor 2 (SF2)-associated protein p34/p32.

7. The LBR CTD is a sterol C-14 reductase

The LBR CTD cannot be studied in aqueous buffers because of its hydrophobic nature. However its homology with fungal sterol C-14 reductases suggested that it might function as a sterol C-14 reductase. Prakash et al (1998, 1999) examined the CTD of human LBR for complementation of the Neurospora crassa sterol C-14 reductase mutant erg-3. Recombinant genes encoding proteins chimeric for different amino acid sequences of Neurospora sterol C-14 reductase and human LBR CTD were constructed and transformed into the erg-3 mutant strain. All the chimeric constructs tested were able to complement the biochanin A sensitive phenotype of the erg-3 mutant (see figure 1) and UV spectrophotometry confirmed that the transformants could synthesize ergosterol. This demonstrated that the LBR CTD is indeed a sterol C-14 reductase. Independently, Silve et al (1998) showed that human LBR can function as a sterol C-14 reductase in yeast. Novel questions emerge from these findings: Can sterol C-14 reductase inhibitors interfere with LBR function? Are sterol changes involved in nuclear envelope breakdown and reformation?

8. TM7SF2 and DHCR7 encode paralogues of the LBR CTD

Recently two genes, TM7SF2 and DHCR7, were identified on human chromosome 11q13 on the basis of the similarity of their predicted proteins to the LBR CTD (Holmer et al 1998). The DHCR7 protein contains 475 amino acid residues and it has been shown to be the

cholesterol biosynthetic enzyme 7-dehydrocholesterol reductase (Moebius et al 1998). Mutations in DHCR7 cause the Smith–Lemli–Opitz syndrome (Fitzky et al 1998; Wassif et al 1998; Waterham et al 1998), which is the second most common autosomal recessive genetic disorder (after cystic fibrosis) in the US white population (carrier frequency 1/200). The TM7SF2 protein is comprised of 418 amino acid residues and its function is yet to be determined. The TM7SF2 protein shows about 40% amino acid sequence identity with the Neurospora erg-3 protein and recent results from our laboratory showed that a protein chimeric for TM7SF2 and erg-3 sequences could complement the Neurospora erg-3 mutant (A Prakash and D P Kasbekar, unpublished results). This suggests that TM7SF2 is also a sterol C-14 reductase. A variant TM7SF2 protein of 590 amino acid residues was reported by Lemmens et al (1998). Unless a mistake has been made while sequencing, this variant might represent a rare alternatively spliced transcript of TM7SF2.

9. Identification of interacting proteins:
Yeast two-hybrid screens

Ye and Worman (1996) cloned the cDNA for human LBR NTD into the GAL4 DNA binding domain fusion vector and co-transformed the resulting plasmid into yeast cells together with a library of ~ 10⁶ recombinants of HeLa cell cDNA in the GAL4 activation domain fusion vector (pGADGH). Two co-transformants that showed the yeast two-hybrid interaction (Fields and Sternglanz 1994), were isolated and the inserts in their pGADGH-derived plasmids were sequenced. One encoded the protein HP1^Hsa and the other encoded the majority of a novel protein with 65% sequence similarity to HP1^Hsa. The latter protein was designated HP1^Hsg. Both HP1^Hsa and HP1^Hsg are homologous to the Drosophila melanogaster heterochromatin-associated protein HP1 which is a suppressor of position effect variegation (James et al 1989; Eissenberg et al 1990). The two HP1-like proteins were synthesized in an in vitro transcription-translation system and shown to bind specifically to LBR NTD-GST coupled glutathione-sepharose. The in vitro binding of LBR-NTD to the HP1-like proteins confirmed the authenticity of the interactions detected in the yeast two-hybrid assay.

An algorithm called hydrophobic cluster analysis (HCA) was performed on the amino acid sequences of the HP1-type proteins and the LBR NTD (Ye et al 1997). This revealed that the HP1 proteins are comprised of
two homologous but distinct globular domains. The N-terminal domain corresponded to the chromodomain originally identified on the basis of the homology between Drosophila HP1 and Polycomb, a protein involved in down-regulating homeotic genes during development (Paro and Hogness 1991). The C-terminal domain was designated the chromo-shadow domain. The chromo- and chromo-shadow domains are separated by a hinge region of ~ 70 amino acids. Experiments using the GST-HP1^Hsa fusion proteins or the yeast two-hybrid assay showed that the chromo-shadow domain mediates the self associations of HP1-type proteins as well as their binding to LBR NTD.

HCA of the LBR NTD also revealed two globular domains separated by a hinge region. The first domain is between amino acid residues 1 and 60 and the second between 105 and 210. The globular domains of the LBR NTD are distinct from the chromodomain and chromo-shadow domains of HP1-type proteins. Yeast two-hybrid assays and binding assays of in vitro translated fragments of the LBR NTD to GST-HP1^Hsa fusion proteins showed that the stretch bearing residues 97–174, which contains a portion of the second globular domain, mediates the binding of LBR to HP1^Hsa. Interestingly the Thr residue reported to be phosphorylated in mitosis (Courvalin et al 1992) is in the second globular domain. It was suggested that this phosphorylation may disrupt the binding between LBR and the HP1-type proteins at the onset of mitosis when the inner nuclear membrane dissociates from the chromatin. And conversely, its dephosphorylation at the end of mitosis might function in the targeting of LBR-bearing membrane vesicles to the decondensing chromatin (Pyrpasopoulou et al 1996; Ye et al 1997).

10. Identification of interacting proteins: Immunoprecipitation of the p58 complex

To identify other factors that may modify or interact with p58 in vivo Simos and Georgatos (1992) subjected chicken erythrocyte nuclei to mild detergent extraction and immunoisolated p58 with an affinity purified antibody against a synthetic peptide comprised of the amino acid residues 61–80 (R1 peptide) of the chicken LBR NTD. Coomassie brilliant blue staining of the SDS-PAGE analysed immunoisolate showed up only the 58 kDa LBR band but silver staining revealed additional minor bands that were not present if the immunoprecipitation was done in the presence of excess R1 peptide. The appearance of apparently specific co-immunoprecipitating bands suggested that at least a fraction of p58 is solubilized as a multicomponent complex. The complex included the lamins A, B and C as shown by Western analysis and additional proteins that migrated at 18, 34 and 150 kDa.

The p34 could be eluted with 1 M NaCl from the turkey p58 complex purified on an anti-R1 immunoaffinity column. It was purified further on a Sephacryl S100 column followed by SDS-PAGE, and subjected to microsequencing following transfer to PVDF membrane and excision of the 34 kDa band. The sequence showed that p34 is the avian homologue of the human nuclear protein p32 which is associated with splicing factor 2 (SF2) (Simos and Georgatos 1994). This homology was confirmed by demonstrating that turkey p34 is recognized by affinity purified antibodies raised against the N- and C-terminal residues of HeLa p32. The p34/p32 may function as a linking component between the inner nuclear membrane and structures involved in RNA splicing. Interestingly, SF2 contains an RS-rich region which is similar to the RS-rich region of the LBR NTD. The RS-rich domains appear to provide the binding sites for p34/p32.

Incubation of the p58 complex with [g ³²P]ATP and then analysis by SDS-PAGE and autoradiography showed heavy phosphorylation of the p58 band and indicated the presence of a specific LBR kinase in the complex. This kinase was distinct from protein kinase A and cdc2 kinase because it was not inhibited by PKI, an inhibitor of protein kinase A or by L1, a specific peptide inhibitor of cdc2. These inhibitors could inhibit the phosphorylation of p58 by their cognate kinases. The LBR kinase migrated as a 110 kDa protein in SDS-PAGE. It was shown to phosphorylate specific Ser residues in the RS region of the NTD of GST-NTD fusion proteins and in synthetic peptides representing different sequences from LBR (Nikolakaki et al 1996, 1997). Phosphorylation by the LBR kinase reduced the binding of LBR to p34/p32.

p18 turned out to be an integral membrane protein that is primarily expressed in erythrocytes but not in liver cells and which is distributed equally between the inner and outer nuclear membranes (Simos et al 1996). It is similar to the mitochondrial isoquinoline-binding protein. p18 could also bind to B-lamins, but less strongly than to LBR. The p18 protein may be internally linked to the TM domain of LBR.

11. LBR as a probe for nuclear envelope
assembly and disassembly

Ellenberg et al (1997, 1998) expressed a fusion protein of LBR with green fluorescence protein (GFP) in living COS-7 cells. The fusion protein contained the first 238 amino acid residues of full length LBR followed by GFP. Therefore it included the LBR NTD and first transmembrane span which have been shown previously to be sufficient to target heterologous proteins to the inner nuclear membrane (Smith and Blobel 1993; Soullam and Worman 1993, 1995). This segment contains the binding domains for the three nucleoplasmic ligands of LBR; lamin B, HP1 and DNA. Confocal-time lapse imaging of the fusion protein in interphase cells showed that it was synthesized in the RER, from where it diffused to the inner nuclear membrane. About 8 h after injection of the plasmid construct there was a 5·3-fold higher level of LBR-GFP fluorescence in the nuclear envelope relative
to the ER. Fluorescence recovery after photobleaching (FRAP) analysis revealed that the LBR-GFP in the nuclear envelope is immobile but the subpopulation that remains in the ER is mobile and diffuses rapidly and freely. Presumably the tight binding of the NTD to the heterochromatin and/or lamina led to immobilization of the LBR-GFP in the nuclear envelope. Thus changes in LBR-GFP mobility within the ER/nuclear envelope membrane system appear to be responsible for its localization to the inner nuclear membrane during interphase.

During prometaphase the LBR-GFP redistributed from the nuclear envelope into the interconnected ER and exhibited the same high mobility and diffusion as observed in the interphase ER membranes. This indicated that during nuclear envelope disassembly the nuclear membrane loses its nucleoplasmic contacts and becomes ‘resorbed’ into the ER network. At the end of mitosis the ER membrane domains containing the LBR-GFP re-establish contacts with the chromosomes and wrap around the chromatin as a newly formed nuclear envelope. Antibodies to an LBR-like protein have been used to examine the ordered recruitment of functionally discrete vesicle types during nuclear envelope reassembly in Xenopus (Drummond et al 1999).

In summary, these experiments revealed that nuclear envelope disassembly occurs by dissolution of the nuclear membranes into the surrounding ER network rather than by fragmentation and vesiculation; and that nuclear envelope reassembly involves ER reorganization around decondensing chromatin.

12. Might LBR regulate genes and differentiation?

Overexpression of the LBR-GFP fusion protein in the COS-7 cells (see § 11) resulted in the formation of nuclear membrane invaginations into the nucleoplasm (Ellenberg et al 1997). In contrast, Smith and Blobel (1994) had reported that in yeast cells expressing chicken LBR, membrane stacks were formed in locations that are typical of yeast ER. The membrane stacks presumably represented the additional ER generated by the yeast cells to accommodate the heterologous LBR protein. The formation of membrane stacks in yeast but invaginations in COS-7 cells might reflect the "drawing in" of the extra membrane into the nucleoplasm by the interaction between the LBR NTD and HP1-type proteins in the COS-7 cells. Yeast nuclei might not show this interaction either because they do not contain as much heterochromatin as the COS-7 cells or because they lack proteins that interact with the heterologous LBR. Extending this idea further, it is conceivable that LBR levels might affect the extent to which the nuclear membrane and heterochromatin interact.

Treatment of HL-60 human myeloid leukemia cells with retinoic acid stimulates them to differentiate along the granulocyte pathway. Granulocyte differentiation induces major ultrastructural changes of the nucleus including the formation of lobules and production of envelope-limited chromatin sheets (ELCS) (Olins et al 1998). Concomitant with this increase in nuclear membrane and membrane-associated chromatin, there is a significant increase in the amounts of LBR and two antigenically related smaller peptides of 50 and 35 kDa. Might the key element underlying these changes simply be a retinoic acid induced increase in LBR expression? Might this happen even during normal granulocyte differentiation?

The hypothesis that the chromatin makes nonrandom contacts with the inner nuclear surface has acquired increasing experimental support (Mathog et al 1984; Nagele et al 1999). The binding of LBR to HP1-type chromatin proteins is consistent with this hypothesis. Blobel (1985) had suggested that the nuclear context in which genes operate is critical for their expression. If LBR levels can alter this context by changing the number of heterochromatin–nuclear membrane contacts it is conceivable that it might affect the expression of genes susceptible to position effect variegation. One way to test this idea might be to sort the COS-7 cells on the basis of their LBR-GFP fluorescence intensity and search for transcript differences by differential display PCR.

13. Conclusion

Studies with LBR have yielded new insights into the relationship between the nuclear envelope and the ER; they have provided a novel probe to study the dynamic interactions between the chromatin, splicing machinery, nucleoskeleton and intracellular membranes and hinted at a role for sterol alterations in nuclear envelope assembly and disassembly. The discovery of genes coding for protein paralogues of the LBR CTD but which lack sequences equivalent to the LBR NTD has provided new and exciting prospects for studying gene evolution. So it is not at all surprising that the LBR community has come to "expect the unexpected".

I thank J Gowrishankar for his careful reading of the manuscript and useful suggestions, Vincent Colot and Howard Worman for encouragment, and A Prakash for the figure. Work in my laboratory is supported in part by the Departments of Biotechnology and of Science and Technology, New Delhi, the Indo-French Centre for the Promotion of Advanced Research (Project 1403-2) and the Third World Academy of Sciences (96-230 RG/ BIO/AS).

References

MS received 20 April 1999; accepted 19 July 1999