214–218
`
`Nucleic Acids Research, 2000, Vol. 28, No. 1
`
`© 2000 Oxford University Press
`
`Kabat Database and its applications: 30 years after the
`first variability plot
`George Johnson and Tai Te Wu*
`
`Departments of Biochemistry, Molecular Biology and Cell Biology, and of Biomedical Engineering,
`Northwestern University, Evanston, IL 60208, USA
`
`Received July 23, 1999; Revised and Accepted October 13, 1999
`
`ABSTRACT
`The Kabat Database was initially started in 1970 to
`determine the combining site of antibodies based on
`the available amino acid sequences at that time.
`Bence Jones proteins, mostly from human, were
`aligned, using the now-known Kabat numbering
`system, and a quantitative measure, variability, was
`calculated for every position. Three peaks, at positions
`24–34, 50–56 and 89–97, were identified and proposed
`to form the complementarity determining regions
`(CDR) of light chains. Subsequently, antibody heavy
`chain amino acid sequences were also aligned using
`a different numbering system, since the locations of
`their CDRs (31–35B, 50–65 and 95–102) are different
`from those of the light chains. CDRL1 starts right
`after the first invariant Cys 23 of light chains, while
`CDRH1 is eight amino acid residues away from the
`first invariant Cys 22 of heavy chains. During the past
`30 years, the Kabat database has grown to include
`nucleotide sequences, sequences of T cell receptors
`for antigens (TCR), major histocompatibility complex
`(MHC) class I and II molecules and other proteins of
`immunological interest. It has been used extensively
`by immunologists to derive useful structural and
`functional information from the primary sequences
`of these proteins. An overall view of the Kabat Database
`and its various applications are summarized here.
`The Kabat Database is freely available at http://immuno.
`bme.nwu.edu
`
`INTRODUCTION
`The purpose of maintaining the Kabat Database of aligned
`sequences of proteins of
`immunological
`interest,
`in our
`opinion, is to provide useful correlations between structure and
`function for this special group of proteins from their nucleotide
`and amino acid sequences to their tertiary structures (1). These
`sequences are thus aligned with the ultimate aim of under-
`standing how these proteins are folded and how they can
`perform their biological functions. We include only coding
`region sequences that have been published. In some cases, only the
`amino acid sequences were published, while the corresponding
`nucleotide sequences were deposited in GenBank. All stored
`
`sequences were then printed out and checked visually against
`available published sequences. We routinely survey for
`possible new sequences in journals in our libraries, Medline
`entries, cross-references from other papers, and author notification;
`however, we may still miss some sequences. GenBank, on the
`other hand, contains a substantial number of unpublished
`sequences. If there are doubts about these sequences or their
`annotations, please refer to the original papers. The Kabat
`numbering systems (see the Introduction of 2) for antibody
`light and heavy chains, for TCR alpha and beta chains, etc., go
`hand-in-hand with variability calculations. The locations of the
`CDRs are the theoretically derived positions which can be
`verified experimentally. Indeed, from the first antigen–antibody
`Fab complex (3) to the complexes of TCR, processed peptide
`and MHC class I molecule (4,5), it has been realized that alignment
`of amino acid sequences and variability calculations can be of
`utmost
`importance in understanding how these important
`macromolecules function biologically. Due to the rapid devel-
`opment of genetic and protein engineering methods, mouse
`and rat antibodies have been humanized to treat human
`cancers, viral infections, etc (6). CDRs of selected rodent anti-
`bodies are cut out and glued onto human antibody frameworks
`to minimize rejection by human patients.
`Our predicted CDRs are slightly different from Chothia’s. A
`careful comparison can be found from a hyperlink on our
`website to ‘Andrew’s Antibody Page’ (http://www.biochem.ucl.
`ac.uk/~martin/abs/index.html ).
`Massive amounts of sequence data are being continuously
`published in the scientific literature. It is imperative to collect
`and properly align the sequences so that they can be used by as
`many researchers in this field as possible. We have previously
`published five editions of these sequences (see the Introduction
`of 2). In 1991, the fifth edition (2) consisted of three volumes.
`Currently, the database is more than five times as large. As of
`September 29, 1999, the Kabat database contained 1 599 375
`and 2 517 756 nt for antibody light and heavy chain variable
`regions, respectively, as compared to 272 244 and 418 962 nt
`in 1991. Total numbers of entries, amino acids and bases of
`other categories of sequences can be obtained by using the
`‘Current Counts’ hyperlink on our website. The collection is
`available on our website (http://www.immuno.bme.nwu.edu )
`which is free due to the generous support by various research
`grants from NIH since 1970.
`Finally, numerous scientific papers have cited our database,
`quoting our fourth edition (7), fifth edition (2), or one of our
`more recent papers (8). On our part, we have been analyzing
`
`*To whom correspondence should be addressed. Tel: +1 847 491 7849; Fax: +1 847 491 4928; Email: t-wu@nwu.edu
`
`Ex. 1010 - Page 1 of 5
`
`AMGEN INC.
`Exhibit 1010
`
`

`

`the Kabat Database during the past few years with reference to
`the total numbers of antibody and TCR V-genes, possible
`evolutionary selection processes,
`importance of antibody
`CDRH3s as related to their fine specificities, etc.
`
`KABAT DATABASE
`The Kabat Database may be accessed for searching, sequence
`retrieval and analysis by a few different methods: electronic
`mail, WWW and ftp. The electronic mail interface has been
`available since 1993, the WWW interface since 1995 and
`various formats of the database in electronic format for nearly
`a decade (8). Our data formats, searching tools, output formats
`and database structures have gradually been adopted by other
`immunological databases and interfaces.
`Electronic mail interface
`An electronic mail interface (seqhunt2@immuno.bme.nwu.edu )
`provides a non-interactive method for searching and sequence
`retrieval (9). Sending mail to the server address with the single
`word ‘help’ (no quotes) in the message body returns instructions
`for using the server.
`All sequences classes are searchable and returnable. The
`query format allows making AND/OR/NOT constructed
`restrictions on the database and amino acid and nucleotide
`sequence pattern matching with allowable differences.
`Requests are processed as they are received and depending on
`the network traffic, take ~1–2 min to be searched and returned
`to the sender. The returned format is a fixed-line length record
`of 80 or fewer characters per line for ease in visual inspection
`and processing by user-written scripts or programs. The characters
`are plain text.
`The query format for the sent request consists of two parts.
`The first part contains directives for the server to follow while
`the second part contains specifications of the search. Specification
`of the extent of data returned, the number of documents to
`return, starting document and whether plain ASCII text or
`PostScript should be used in the return format may be entered.
`Further, one can direct the server to return a distribution, the
`variability or unaligned raw data for the search specified.
`The second part of the query contains the search restrictions
`on the database. Words separated by AND and OR may be
`used, as well as searching functions, like nucleotide/amino
`acid pattern matching and positional restriction matching.
`There are basically three steps in translating and performing
`a search on the Kabat Database: generate the question or query,
`translate it into a format the server can recognize and decide on
`the output options desired of the returned matches. For
`example, if matches of mouse kappa light chains of anti-
`phosphorylcholine antibodies are desired,
`the query and
`restriction on the database would be:
`Begin
`@mouse and kappa and phosphorylcholine
`The ‘@’ before mouse tells the server that matches of the
`species mouse are desired, rather than searching through the
`entire database record for instances of the word ‘mouse’. More
`complicated restrictions can be generated using parentheses for
`grouping and the minus sign ‘–’ for NOT. Finding all rat and
`rabbit sequences which are not kappa light chains, and
`returning them as amino acid sequences in PostScript format
`would be constructed as:
`
`Nucleic Acids Research, 2000, Vol. 28, No. 1
`
`215
`
`PSAA
`Begin
`(rat and rabbit) and –kappa
`Pattern matching is interpreted as the second part of an AND
`statement, such that finding all rat and rabbit sequences which
`are not kappa and contain the nucleotide pattern cagtacgtcag
`with three allowable mismatches, would be sent as:
`Begin
`(rat and rabbit) and –kappa [ implicit AND ]
`#NM 3
`cagtacgtcag
`More examples of searching and output options may be found
`in the ‘help’ file returned from the server.
`WWW interface
`The WWW interface (8) to the Kabat Database: http://immuno.
`bme.nwu.edu contains searching and analysis tools as well as
`links to database download sites and other interesting databases.
`Most of the features found in the electronic mail interface are
`found in the WWW interface, as well as other tools. The
`WWW interface is more interactive than the Email and returns
`results faster, depending on the network traffic.
`Searching and analysis tools
`SeqhuntII. This grouping of programs allows searches through
`the annotations and sequence pattern matching of the amino
`acid and nucleotide sequence data with allowable mismatches.
`Like the Email server, restrictions on the database may be
`formulated as AND/OR/NOT constructs. Output extent, output
`format, maximum documents and starting document may be
`specified. Browsing of the output results in HTML format
`allows the user to view the database entries in an easy-to-read
`format. ASCII text may be selected as output for use in user-
`generated scripts and programs. PostScript generation allows
`for printing on a PostScript supporting printer. Sequence
`matching is returned aligned with the target sequence with
`nucleotide or amino acid differences from the database
`sequence displayed in a case change. Since the database
`contains only coding regions of genes and proteins, the query
`sequence should be a portion of the coding region being sought.
`
`Variability. Variability and amino acid distributions of
`sequence groups may be generated for restrictions on the data-
`base. The variability plots are in PostScript format and may
`either be viewed on the screen with an appropriate PostScript
`viewer (e.g. GNU ghostscript or ghostview) or printed to a
`postscript-supporting printer. Plots for human and mouse TCR
`gamma and delta chain variable regions are shown in Figure 1.
`Scaling of the variability plots may be done allowing comparison
`of variability plots for different groupings of sequences.
`Distributions of the amino acids per position may be returned
`also, including the calculated variability for each position.
`
`Sequence alignment. Alignment of user-entered coding regions
`of
`immunoglobulin light chains according to the Kabat
`numbering system can be performed. Because of the relatively
`few alignment options available for
`light chains, most
`sequences can be aligned. One can start with around 10 amino
`acid residues or 30 nt. There is no lower limit on the length of
`sequence to be matched. In some cases though, visual inspection
`and alignment is necessary, as is for heavy chain alignment,
`
`Ex. 1010 - Page 2 of 5
`
`

`

`216
`
`Nucleic Acids Research, 2000, Vol. 28, No. 1
`
`Figure 1. Variability plots for human and mouse TCR gamma and delta chain variable regions, using 377 human gamma, 1260 human delta, 297 mouse gamma
`and 461 mouse delta partial and complete sequences.
`
`especially within the CDRH3 region, if additional codons or
`residues are inserted and denoted by ‘#’. If a suitable alignment
`counterpart from the database is not found for the target
`sequence, the user can contact us.
`
`FTP. Various formats of the database are available for down-
`load from NCBI’s repository under the directory ‘kabat’.
`Currently active formats include a FASTA-like raw sequence
`format and the database’s fixed length format of 80 or fewer
`
`characters per line and vertical alignment. Four main variations
`on the fixed length format exist to properly visually display
`single translations, pseudogene translations, J-minigenes and
`D-minigenes. Other immunological databases have adopted
`similar formats as exemplified by the three letter code amino
`acid translation followed by single letter code. A ‘dump’
`version of the database is periodically updated which contains
`unlimited line length records more suitable for mass
`processing on unix-based systems.
`
`Ex. 1010 - Page 3 of 5
`
`

`

`Nucleic Acids Research, 2000, Vol. 28, No. 1
`
`217
`
`Figure 2. Length distributions of CDR3s of human and mouse TCR gamma and delta chains, based on 135 human gamma, 546 human delta, 37 mouse gamma and
`66 mouse delta complete CDR3 sequences.
`
`OTHER APPLICATIONS
`As mentioned before,
`the Kabat Database was initially
`constructed for
`the purpose of
`identifying the antibody
`combining site (1). Starting from aligned amino acid sequences
`and using variability calculations, we have identified CDRs of
`antibody light and heavy chains, as well as those of TCRs.
`Such calculations can also provide useful predictions for MHC
`class I and II sequences (8), and to other aligned proteins
`sequences, e.g. HIV gp120, gp41, etc.
`The importance of CDRH3 to confer fine specificity to anti-
`bodies was realized a few years ago (10). Furthermore, the
`unique CDRH3 nucleotide sequences have recently been used
`as a sensitive diagnostic test to detect residue B cell malignancies
`in cancer patients. Thus, many of these sequences have been
`determined. But most of them have been excluded from
`GenBank due to their relative short lengths. We have been
`meticulously collecting them, and realized the importance of
`their length distributions in antibodies of various specificities
`(11), and possible differences between CDRH3s of human and
`mouse (12). In the case of rabbit, the CDRH3s have less length
`variation than human and mouse. This may be compensated by
`the length variations of the CDRL3s (13).
`
`The length variations of TCR alpha and beta chain CDR3s
`are very restricted (14). On the other hand, TCR gamma and
`delta chain CDR3s have more length variation, close to those
`of antibody heavy chains (Fig. 2). Whether they bind antigens
`directly is unclear.
`During recent years, various research groups have decided to
`sequence the entire coding region of different antibody and
`TCR V-genes, in order to have an idea of their total numbers. On
`the other hand, we have discovered that pair-wise comparisons of
`V-gene nucleotide sequences in the Kabat Database provide
`very accurate estimations of their total numbers (15,16). In
`addition, such comparisons seem to suggest that antibody and
`TCR V-genes have evolved under different selective pressures
`(17). In the case of MHC class I sequences, comparison of their
`aligned sequences has elucidated a new mechanism of generating
`novel MHC class I molecules by random assortment of their a1
`and a2 gene segments (18).
`
`DISCUSSION
`The Kabat Database has been around for 30 years. It has
`provided the immunology community a useful service, since it
`
`Ex. 1010 - Page 4 of 5
`
`

`

`218
`
`Nucleic Acids Research, 2000, Vol. 28, No. 1
`
`not only is a sequence database but also incorporates vital
`aspects of the biology of the immune system. Various analytical
`methods have been developed to study the structure and function
`relations of proteins of immunological interest.
`Electronic addresses:
`http://immuno.bme.nwu.edu
`seqhunt2@immuno.bme.nwu.edu
`Citing the Kabat Database:
`Authors using this database may cite this paper together with
`the electronic addresses.
`
`ACKNOWLEDGEMENT
`Supported in part by NIH Grant 5 R24 AI25616-10.
`
`REFERENCES
`1. Wu,T.T. and Kabat,E.A. (1970) J. Exp. Med., 132, 211–250.
`2. Kabat,E.A., Wu,T.T., Perry,H., Gottesman,K. and Foeller,C. (1991)
`Sequences of Proteins of Immunological Interest, Fifth Edition. NIH
`Publication No. 91-3242.
`3. Amit,A.G., Mariuzza,R.A., Phillips,S.E.V. and Poljak,R.J. (1986)
`Science, 233, 747–753.
`
`4. Garcia,K.C., Degano,M., Stanfield,R.L., Brunmark,A., Jackson,M.R.,
`Peterson,P.A., Teyton,L. and Wilson,I.A. (1996) Science, 274, 209–219.
`5. Garboczi,D.H., Ghosh,P., Utz,U., Fan,Q.R., Biddison,W.E. and
`Wiley,D.C. (1996) Nature, 384, 134–141.
`6. Jones,P.T., Dear,P.H., Foote,J., Neuberger,M.S. and Winter,G. (1986)
`Nature, 321, 522–525.
`7. Kabat,E.A., Wu,T.T., Reid-Miller,M., Perry,H. and Gottesman,K. (1987)
`Sequences of Proteins of Immunological Interest, Fourth Edition. US
`Govt. Printing Off. No. 165-492.
`8. Johnson,G., Kabat,E.A. and Wu,T.T. (1996) In Herzenberg,L.A.,
`Weir,W.M., Herzenberg,L.A. and Blackwell,C. (eds), Weir’s Handbook
`of Experimental Immunology I. Immunochemistry and Molecular
`Immunology, Fifth Edition. Blackwell Science Inc., Cambridge, MA,
`pp. 6.1–6.21.
`9. Johnson,G., Wu,T.T. and Kabat,E.A. (1995) In Paul,S. (ed.),
`Antibody Engineering Protocols. Humana Press, pp.1–15.
`10. Kabat,E.A. and Wu,T.T. (1991) J. Immunol., 147, 1709–1719.
`11. Johnson,G. and Wu,T.T. (1998) Int. Immunol., 10, 1801–1805.
`12. Wu,T.T., Johnson,G. and Kabat,E.A. (1993) Proteins, 16, 1–7.
`13. Sehgal,D., Johnson,G., Wu,T.T. and Mage,R.G. (1999) Immunogenetics,
`50, 31–42.
`14. Johnson,G. and Wu,T.T. (1999) Immunol. Cell Biol., 77, 391–394.
`15. Johnson,G. and Wu,T.T. (1997) Genetics, 145, 777–786.
`16. Johnson,G. and Wu,T.T. (1997) Immunol. Cell Biol.,75, 580–583.
`17. Johnson,G. and Wu,T.T. (1997) J. Mol. Evol., 44, 253–257.
`18. Johnson,G. and Wu,T.T. (1998) Genetics, 149, 1063–1067.
`
`Ex. 1010 - Page 5 of 5
`
`