The Translin Ring Specifically Recognizes DNA Ends at Recombination Hot Spots in the Human Genome*

(Received for publication, February 21, 1996, and in revised form, December 20, 1996)

Masataka Kasai Dagger §, Takao Matsuzaki par , Katsuo Katayanagi , Akira Omori par , Richard T. Maziarz **, Jack L. Strominger Dagger Dagger , Katsunori Aoki Dagger §§ and Kenji Suzuki ¶¶

From the Departments of Dagger  Immunology and ¶¶ Pathology, National Institute of Health, 1-23-1 Toyama, Shinjuku-ku, Tokyo 162, Japan,  Mitsubishi Chemical Corporation, 1000 Kamoshida, Aobaku, Yokohama 227, Japan, par  Mitsubishi-Kasei Institute of Life Sciences, Minami-ooya 11, Machida-shi, Tokyo 194, Japan, ** Division of Hematology and Oncology, Oregon Health Sciences University and the Portland Veterans Affairs Medical Center, Portland, Oregon 97207, Dagger Dagger  Department of Molecular and Cellular Biology, Harvard University, Cambridge, Massachusetts 02138, and §§ The First Department of Internal Medicine, School of Medicine, University of Tokyo, Bunkyo-ku, Tokyo 113, Japan

ABSTRACT
INTRODUCTION
EXPERIMENTAL PROCEDURES
RESULTS
DISCUSSION
FOOTNOTES
ACKNOWLEDGEMENTS
REFERENCES


ABSTRACT

We previously showed that consensus sequences exist at the chromosomal breakpoints in lymphoid malignancies and that these sequences are specifically recognized by a novel DNA binding protein, Translin. In the present study, the native form of Translin was established to be a ring-shaped structure by electron microscopy and crystallographic studies. It was also determined that this multimeric Translin formed by the subunits is responsible for its binding to target sequences situated only at single-stranded DNA ends. Furthermore, DNA-damaging reagents were found to initiate a signaling pathway for the active nuclear transport of Translin. The results support the hypothesis that staggered breaks occur at recombination hot spots and Translin has a pivotal function in recognition of the generated single-stranded DNA ends.


INTRODUCTION

The significance of chromosomal translocations in many hematological tumors has recently become very clear (1). A number of studies have shown that chromosomal translocations either result in the activation of proto-oncogenes by joining them to immunoglobulin (Ig) or T-cell receptor genes or lead to the creation of tumor-specific fusion proteins. In man, such translocations consistently occur at particular sites in the genome. Despite their significance to tumor etiology, the molecular basis for the existence of these "recombination hot spots" has been poorly understood.

Virtually identical sequences at the chromosomal breakpoint junctions in T-cell leukemias bearing the t(8;14)(q24;q11) and t(1;14)(p32;q11) rearrangements led us to suspect that a novel enzymatic mechanism might be operating on chromosomes 1p32 and 8q24 (2, 3). Similar sequences were also found within the clustered breakpoint region of the Bcl-2 oncogene in follicular lymphoma patients carrying the t(14;18)(q32;q21) translocations, which is one of the most common chromosomal abnormalities in human lymphoid neoplasms (4). Further extensive search of the nucleotide sequences at the breakpoint junctions of many chromosomal abnormalities in human lymphoid neoplasms involving 1p32, 3q27, 5q31, 8q24, 9q34, 9q34.3, 10q24, 11p13, 11q13, 14q11, 14q32, 14q32.1, 17q22, 18q21, 19p13, and 22q11 revealed the consensus sequence motifs, ATGCAG and GCCC(A/T)(G/C)(G/C)(A/T), with gaps or a few nucleotides intervening (5). In most cases, these sequences were found at the 5'-flanking site of the breakpoint junctions, with genes encoding physiologically important proteins located in their vicinity. Subsequent analysis identified a protein, Translin, that exhibits general binding activity to the above consensus sequences. Molecular gene cloning experiments revealed the protein to be a previously undescribed type with no significant homology to known proteins (5). In addition to our original observations indicating that Translin is expressed predominantly in the nuclei of lymphoid lineage cell lines in which the Ig or T-cell receptor loci are not in a germline DNA configuration, confocal laser-scanning microscopic analysis revealed that nuclear localization of Translin is limited to lymphoid lineage cells despite being present in the cytoplasm of cell lines of various lineage. These observations support the view that the nuclear localization of Translin is associated with its physiological role and that its nuclear importation is selectively controlled in lymphoid cells with an Ig or T-cell receptor gene rearrangement.

In the present investigation, to provide further insight into Translin function, the native form of multimeric Translin and its mode of binding to the target DNA sequences were determined by electron microscopic studies. It was further suggested that a signaling pathway for the active nuclear transport of Translin may be initiated by exposure to DNA-damaging reagents.


EXPERIMENTAL PROCEDURES

Recombinant Translin and Electron Microscopic Analysis

The Translin gene encoding 228 amino acids was amplified by PCR1 and cloned into a bacterial expression vector, pQE-9 (Qiagen Inc.). The resulting expression construct was transformed into the Escherichia coli host strain M15[pREP4], and production of recombinant Translin protein was induced by treatment with 2 mM isopropyl beta -D-thiogalactopyranoside for 4 h. The Translin was purified from the bacterial lysate using a Ni2+-agarose column, and a sample was prepared on a thin carbon film supported by a mesh copper grid and negatively stained with potassium phosphotungstate adjusted to pH 7.0. To analyze the mode of Translin/DNA interactions, heat-denatured ss DNA was incubated with the recombinant Translin in 50 mM phosphate buffer (pH 8.0) with 300 mM NaCl for 30 min at 20 °C. Samples were spread onto 5-µl droplets of redistilled water and then absorbed on carbon-coated grids. Micrographs of rotary-shadowed samples were taken in a Hitachi H-7000 electron microscope.

Crystallization of Translin and Data Collection

Crystals of the recombinant Translin were grown by the hanging drop method at 5 °C. The reservoir solution contained 15% (w/w) polyethylene glycol 4000 in 0.1 M MES buffer at pH 6.2. The protein solution contained 3.4 mg/ml protein in 50 mM phosphate buffer (pH 8.0) with 300 mM NaCl. Initial drops were formed with 4.5 µl of reservoir solution and 6.0 µl of protein solution. Bipyramidal crystals grew to 0.3 × 0.15 × 0.15 mm after several weeks. X-ray diffraction data were collected with a Weissenberg camera for macromolecules (13) at the photon factory (beam line 6A2) of the National Laboratory for High Energy Physics, Tsukuba, Japan. The crystals were found to belong to a tetragonal space group, either P41212 or P43212, with cell parameters of a = 97.2 and c = 283.6 Å. The ratio of the volume to the molecular mass was estimated to be 3.2 Å3/dalton assuming four molecules in an asymmetric unit. The intensity data were processed using the DENZO program (14). The Rmerge was 13% for 17,357 independent reflections with |I| > 1sigma (|I|) within the 3.5-Å resolution, and the completeness was 95%. The heavy atom derivative search for multiple isomorphous replacement is now under way.

PCR Amplification of Bcl-2 Genomic Fragments

DNA fragments in the mbr were synthesized by PCR using the Bcl-2 genomic fragment (4.3-kilobase HindIII) of chromosome 18q21 (8). PCR primers for each DNA fragment were as follows: Bcl-1048, B1960 (TTCTTAAGACATGTATCACT) and B3007M (TAAAGCAACTCTCTAAAGGT); Bcl-1103, B1960 and Bcl-24M (4); Bcl-142, B2920 (CCATGAGATTCATTCAGTTAA) and Bcl-24M. Parameters for PCR amplification were 1 min each at 94, 60, and 72 °C for 40 cycles, followed by a final extension period of 15 min at 72 °C. Single-stranded forms of each DNA were prepared by heating at 100 °C for 5 min followed by cooling on ice.

DNA-damaging Reagents and Detection of Translin

HeLa cells were grown on 9-cm Petri dishes treated with mitomycin C (1 µg ml-1) or etoposide (5 mM) for 6 or 18 h, and then the cells were lysed and nuclear extracts prepared as described previously (3, 4). Nuclear extracts (20 µg of total protein) were run on 12.5% acrylamide SDS-polyacrylamide gel electrophoresis, transferred to Hybond-polyvinylidene difluoride membrane (Amersham Life Science), and probed with rabbit anti-Translin antibody followed by horseradish peroxidase-conjugated goat anti-rabbit IgG. Antibody binding was detected by enhanced chemiluminescence according to the manufacturer's instructions (Amersham Corp.). Nuclear extracts of HeLa cells treated with mitomycin C were also assayed for binding to 32P-labeled Vdelta 3-23P (3) by gel shift assay.


RESULTS

The Native Form of Translin Is a Ring-shaped Octamer

The Translin cDNA encodes a protein of 228 amino acids with a predicted molecular size of 26 kDa. SDS-polyacrylamide gel electrophoresis analysis showed the recombinant Translin migrated as a single band of approximately 27 kDa under reducing conditions and 54 kDa under nonreducing conditions, indicating that Translin polypeptides form a dimer (5). In the native form, Translin dimers generate a multimeric structure whose molecular size is approximately 220 kDa, and this form is responsible for its DNA binding activity. Since Translin has the leucine zipper motif at the COOH terminus (Fig. 1), we hypothesized that polymerization of Translin dimers could be facilitated by the heptad repeat of hydrophobic amino acids (five leucines and one valine) to form a multimeric subunit structure, possibly an octamer. Therefore, the top view of native Translin would be expected to give a circular structure connected by disulfide bonds and leucine zipper motifs. Confirming this hypothesis, our electron microscopic study indicated that the complex formed by Translin subunits is ring-shaped (Fig. 2).


Fig. 1. Schematic representation of the Translin protein. The predicted amino acid sequence contains two relatively basic regions (amino acids 56-64 and 86-97), indicated by the hatched regions. The COOH-terminal region of the molecule (amino acids 177-212), indicated by the shaded region, contains the hypothetical structure referred to as the leucine zipper.
[View Larger Version of this Image (10K GIF file)]



Fig. 2. Visualization of native Translin under the electron microscope. The recombinant Translin was prepared on a thin carbon film supported by a mesh copper grid and negatively stained with potassium phosphotungstate.
[View Larger Version of this Image (161K GIF file)]


To determine the three-dimensional structure of Translin, a crystallographic investigation was started. For the preliminary study, the self-rotation function (6) was calculated with the program POLARRFN of the CCP4 package (7). The reflection data between 20- and 3.5-Å resolution were used, but the program limits the resolution to 13.7 Å for the integration radius of 80 Å. The highest peaks at omega  = 0 and kappa  = 90 and 180 correspond to the crystallographic 41 or 43 screw axis, and those at omega  = 90, kappa  = 180, and phi = 0, 45, 135, and 180 correspond to the crystallographic 2-fold rotation and 21 screw axes. There were no other peaks with significant height except in the region omega  = 0. Fig. 3 shows the self-rotation function versus kappa  at omega  = 0. The maximum background level at omega not equal  0 was 15. The fact that there is only one axis that gives a high score for the self-rotation function is quite consistent with the ring-shaped structure observed in the electron micrography. The figure shows that the axis normal to the ring lies parallel to the c axis of the crystal. Thus, there are four layers of the ring along the c axis in a unit cell, and the crystallographic 2-fold axis must be within the plane of the ring. The 2-fold axis implies the ring consists of even numbers of subunits, i.e. 6, 8, or 10. The low resolution rotation function peaks at kappa  = 45 and 135 together with the assembly molecular weight suggest that the Translin is an assembly of eight subunits.


Fig. 3. Histogram of self-rotation function at omega  = 0 along kappa . The integration radius is 80 Å. omega  is an inclination angle of an expected symmetry axis from the crystallographic c axis. kappa is the rotation angle around the expected symmetry axis. Peaks at kappa  = 90 and 180 are due to the crystallographic 41 or 43 symmetry.
[View Larger Version of this Image (25K GIF file)]


Translin Is a Single-stranded DNA End-binding Protein

The multimeric structure of Translin and its DNA binding activity raises the question of how DNA binding is achieved with its ring-shaped structure. It is conceivable that the DNA binding domain is created in the cylindrical structure of native Translin protein by combination with its basic regions. To confirm this, the mode of Translin/DNA interactions was analyzed using the target sequence of a recombination hot spot region on chromosome 18q21 (4). The chromosomal breakpoints of follicular lymphomas carrying the t(14;18)(q32;q21) are known to be clustered within a 150-base pair region in the mbr of the Bcl-2 oncogene on chromosome 18q21 (8-10). We synthesized an oligonucleotide, Bcl-CL1, that included the consensus sequence found within the 150-base pair region and confirmed by gel shift assay that Translin strongly binds to the end-labeled Bcl-CL1 (Fig. 4). Bcl-CL1·Translin complex formation was competitively blocked by increasing amounts of nonradioactive Bcl-CL1. The oligonucleotides Bcl-CL1a and Bcl-CL1c, with an additional 6 nucleotides ((ATT)2 at the 3' and 5' ends), demonstrated similar competitive activities. However, Bcl-CL1b and Bcl-CL1d, with 12 additional nucleotides ((ATT)4 at the 3' and 5' ends), competed less efficiently. Interestingly, Bcl-CL1e with 12 additional nucleotides at both ends failed to inhibit the complex formation. These results clearly indicate that the target sequence of Translin has to be situated at DNA ends for efficient binding to occur.


Fig. 4. Analysis of Translin binding to DNA ends by gel shift assay. Recombinant Translin (50 ng) was assayed for binding to 32P-labeled Bcl-CL1 by the gel shift assay as described previously (5). Increasing amounts of the unlabeled oligonucleotides (80, 160, and 320 pmol), Bcl-CL1, Bcl-CL1a, Bcl-CLb, Bcl-CLc, Bcl-CLd, and Bcl-CLe, were added for competitive studies during the incubation of Translin with 32P-labeled Bcl-CL1.
[View Larger Version of this Image (54K GIF file)]


To further confirm that Translin specifically binds to target sequences at DNA ends, two different double-stranded (ds) DNA fragments (Bcl-1103 and Bcl-1048) from the mbr of the Bcl-2 oncogene were prepared (Fig. 5A), and their binding to Translin was analyzed by electron microscopy. Although both ds DNA fragments failed to bind (data not shown), Translin bound to ss Bcl-1103 with the target sequence (Bcl-CL1) at the DNA end (Fig. 5, E-G) but not to ss Bcl-1048 without the target sequence (Fig. 5, C and D). To ascertain whether Translin is an ss DNA binding protein, a ds DNA fragment, Bcl-142 (142 base pairs) bearing the Bcl-CL1 at its end was prepared (Fig. 5A), and its binding to Translin was tested by gel shift assay. As shown in Fig. 5H, the end-labeled ds Bcl-142 failed to bind to Translin despite binding to ss Bcl-142.


Fig. 5. Specific binding of Translin to single-strand target sequences at DNA ends. A, schematic representation of the Bcl-2 gene on chromosome 18q21 and the DNA fragments, Bcl-1048, Bcl-1103, and Bcl-142, in the mbr. B-G, electron micrographs illustrating Translin binding to ss DNA with the target sequence (Bcl-CL1) at the DNA ends. B, recombinant Translin without DNA. C and D, no binding of Translin to the ss form of the Bcl-1048 without the target sequence at DNA ends. E-G, binding to the ss form of the Bcl-1103 with the target sequence. H, binding of Translin. The results of gel shift analysis indicate that Translin fails to bind to 32P-labeled double-stranded Bcl-142 with the target sequence (Bcl-CL1). The ss form of the Bcl-142, however, demonstrates concentration-dependent binding.
[View Larger Version of this Image (89K GIF file)]


Induction of Active Nuclear Transport of Translin

In the small number of examples tested so far, the native form of Translin was shown to be present in the cytoplasm of cell lines of various lineages, whereas a nuclear location was limited to the case of hematopoietic cell lines, especially lymphoid cell lines with rearranged Ig or T-cell receptor loci. To determine whether or not Translin may have any physiological significance in the cytoplasm of nonhematopoietic cells, we hypothesized that Translin may be involved in both DNA repair and apoptosis. Therefore, a nonhematopoietic cell line (HeLa) was treated with the DNA-damaging agents, mitomycin C and etoposide, and then homogenized for preparation of nuclear extracts. The presence or absence of Translin in the nucleus was tested by immunoblotting with anti-Translin antibody. As shown in Fig. 6A, nuclear transport of Translin was greatly induced by mitomycin C and etoposide within 6 h. Further incubation, however, resulted in a decrease in Translin protein levels. It was confirmed by gel shift assay that mitomycin C induced nuclear transport of Translin present in the native form (Fig. 6B). These experiments indicate that DNA damage initiates a signaling pathway, resulting in active nuclear transport of Translin.


Fig. 6. Induction of nuclear transport of Translin in response to DNA-damaging agents. A, detection of Translin by ECL in the nuclei of HeLa cells after treatment with mitomycin C or etoposide, DNA-damaging agents. B, results of a gel shift assay of Translin in the nuclei of HeLa cells treated with mitomycin C.
[View Larger Version of this Image (27K GIF file)]


Interaction of Translin with Clustered Chromosomal Break Sites

We have previously shown that Translin could also bind to a complementary minus strand of Bcl-CL1, Bcl-24M, containing a tandem repeat of the sequence that is similar to one of the consensus sequences, ATGCAG. To further narrow down the target sequence at the clustered break sites of 18q21 required for binding of Translin, the 3' end of the Bcl-CL1 and the 5' end of the Bcl-24M were deleted, and the resultant forms were tested for their binding to Translin by gel shift analysis. Fig. 7A summarizes the results, indicating that the minimal length of the target sequence is about 18 base pairs of both strands. The 3' end of the plus strand coincides with the most frequent break sites defined in the literature. The same result was also obtained for similar sequences present in that region of DXP genes (members of the Ig DH gene family) that is most frequently rearranged with the Bcl-2 oncogene in the t(14;18) translocation (data not shown). These results suggest that Translin could function by specifically interacting with the target sequence only when a staggered break occurs at chromosomal translocation hot spots, as shown in Fig. 7B.


Fig. 7. Interaction of Translin with the chromosomal break site at 18q21. A, minimal length of the target sequence required for the binding to Translin. Oligonucleotides with deletions at the 3' end of the Bcl-CL1 (Bcl-22P, Bcl-20P, Bcl-18P, and Bcl-16P) and the 5' end of the Bcl-24M (Bcl-22 M, Bcl-20M, Bcl-18M, and Bcl-16M) were 32P-labeled and tested for their binding to Translin by gel shift analysis. Binding was classified as high (+++), medium (++), and none (-) based on the radioactivity measured with a Fujix BAS2000 Bio-imaging Analyzer. B, schematic model for the interaction of the ring-shaped Translin (shaded region) with the generated single-stranded DNA ends at 18q21. The clustered breakpoints are indicated by stars.
[View Larger Version of this Image (37K GIF file)]



DISCUSSION

Despite the significance of recombination in the human genome, little is known about the molecular events occurring at recombination hot spot regions. Translin, which specifically binds to the consensus sequences ATGCAG and GCCC(A/T)(G/C)(G/C)(A/T) found at the breakpoint junctions in many cases of chromosomal translocations, is a unique DNA binding protein. We have shown by electron microscopic and crystallographic studies that the native form of Translin is a ring-shaped structure. Based on our findings for interaction of Translin with the target sequence in the major breakpoint region of the Bcl-2 oncogene on chromosome 18q21, Translin has to be located at DNA ends for efficient binding to occur. Our electron microscopic studies not only confirmed this but also proved that Translin is a ss DNA binding protein. Moreover, the same results support the view that the DNA binding domain is created in the cylindrical structure of Translin by combination with the relatively short basic regions of the Translin polypeptide (Fig. 1) because Translin could only bind to the end target sequence of the ss Bcl-1103 despite the existence of several identical target sequences in its interior (4). Theoretically, in this case, chromosomal breakage would have to precede recognition of target sequences by the ring-shaped Translin protein in vivo. In this respect, Translin may be a typical DNA end-binding protein, which is in contrast with one of the other DNA binding proteins, the Ku antigen, that initially binds to DNA ends and then moves to internal positions within the DNA molecule (11).

An important conclusion that we can draw from these observations is that, after chromosomal breakage at recombination hot spots, a single chain target sequence motif has to be exposed for efficient binding of Translin to occur in vivo. It is important to determine whether it is due to deletion of the 5' end by exonuclease or a staggered cut by unknown restriction endonucleases. Our results also support the view that the existence of a ss DNA region would increase the probability of a free DNA end being generated by a nuclease, facilitating recombination by interaction with DNA binding proteins (12). Further investigation is required to determine if Translin might protect single-stranded DNA ends from nuclease cleavage and promote the initiation of DNA strand exchange. The fact that nuclear transport of Translin is induced by DNA-damaging reagents also raise the question of whether or not an identical signal is involved in the active nuclear transport of Translin in lymphoid cell lines.

The primary amino acid sequences and ring-shaped structure of Translin are highly conserved through evolution, and the consensus target sequences appear to be similar.2 One attractive possibility is that the Translin targets universally exist throughout the eukaryotic genome and influence chromatin structure. The fact that chromosomal translocations recur very consistently in particular chromosomal regions that almost always contain physiologically important genes leads us to suspect that the underlying mechanisms are associated with some fundamental process in eukaryotic cells such as transcription or replication. Further characterization of the Translin actions including identification of interacting proteins should facilitate understanding of its role in the recombination hot spots and of the molecular basis of chromosomal translocations in many hematological diseases.


FOOTNOTES

*   This work was supported by the 2nd Term Comprehensive 10-year Strategy for Cancer Control Fund of the Ministry of Health and Welfare of Japan and a Human Science Research Fund awarded (to M. K.).The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§   To whom correspondence should be addressed. Tel.: 81-3-5285-1111 (ext. 2130); Fax: 81-3-5285-1150.
1   The abbreviations used are: PCR, polymerase chain reaction; MES, 2-(N-morpholino)ethanesulfonic acid; mbr, major breakpoint region; ds, double-stranded; ss, single-stranded.
2   K. Aoki, J. Inazawa, T. Takahashi, K. Nakahara, and M. Kasai, manuscript in preparation.

ACKNOWLEDGEMENTS

We express our appreciation to M. Seto for providing the Bcl-2 genomic clone and F. Hirose, H. Maekawa, and T. Higashinakagawa for valuable suggestions. We also thank T. Hasegawa, C. Sasaki, and G. Meng for technical assistance.


REFERENCES

  1. Rabbitts, T. H. (1994) Nature 372, 143-149 [CrossRef][Medline] [Order article via Infotrieve]
  2. Kasai, M., Maziarz, R. T., Aoki, K., Macintyre, E., and Strominger, J. L. (1992) Mol. Cell. Biol. 12, 4751-4757 [Abstract]
  3. Kasai, M., Aoki, K., Matsuo, Y., Minowada, J., Maziarz, R. T., and Strominger, J. L. (1994) Int. Immunol. 6, 1017-1025 [Abstract]
  4. Aoki, K., Nakahara, K., Ikegawa, C., Seto, M., Takahashi, T., Minowada, J., Strominger, J. L., Maziarz, R. T., and Kasai, M. (1994) Oncogene 9, 1109-1115 [Medline] [Order article via Infotrieve]
  5. Aoki, K., Suzuki, K., Sugano, T., Nakahara, K., Kuge, O., and Kasai, M. (1995) Nat. Genet. 10, 167-174 [Medline] [Order article via Infotrieve]
  6. Crowther, R. A., and Blow, D. M. (1967) Acta Crystallogr. 23, 544-548 [CrossRef]
  7. SERC Daresbury Laboratory (1979) Collaborative Computing Project 4. A Suite of Programs for Protein Crystallography, SERC Daresbury Laboratory, Warrington, United Kingdom
  8. Bakhshi, A., Jensen, J. P., Goldman, P., Wright, J. J., McBride, O. W., Epstein, A. L., and Korsmeyer, S. J. (1985) Cell 41, 899-906 [Medline] [Order article via Infotrieve]
  9. Cleary, M. L., and Sklar, J. (1985) Proc. Natl. Acad. Sci. U. S. A. 82, 7439-7443 [Abstract]
  10. Tsujimoto, Y., Gorham, J., Cossman, J., Jaffe, E., and Croce, C. M. (1985) Science 229, 1390-1393 [Medline] [Order article via Infotrieve]
  11. de Vries, E., van Driel, W., Bergsma, W. G., Arnberg, A. C., and van der Vliet, P. C. (1989) J. Mol. Biol. 208, 65-78 [Medline] [Order article via Infotrieve]
  12. Oberosler, P., Hloch, P., Ramsperger, U., and Stahl, H. (1993) EMBO J. 12, 2389-2396 [Abstract]
  13. Sakabe, N. (1991) Nucl. Instrum. Methods Physiol. Res. Sect. A 303, 448-463
  14. Otwinowski, Z. (1993) in Data Collection and Processing (Sawyer, L., Issacs, N. W., and Bailey, S., eds), pp. 55-62, Daresbury Laboratory, Warrington, United Kingdom

©1997 by The American Society for Biochemistry and Molecular Biology, Inc.