- Email: [email protected]
Structural comparison of prostate-specific antigen and human glandular kallikrein using molecular modeling
ADULT
UROLOGY
STRUCTURAL COMPARISON OF PROSTATE-SPECIFIC ANTIGEN AND HUMAN GLANDULAR KALLIKREIN USING MOLECULAR MODELING DOMINIQUE P BRIDON, PH.D. BARRY L. DOWELL, PH.D. From the Department of Peptide Engineering and the Department of Cancer Product Development, Diagnostic Division, Abbott Laboratories, Abbott Park, Illinois
ABSTRACT-Objectives. Prostate-specific antigen (PSA), the most useful tumor marker for prostate cancer, is one of three members of the human kallikrein family of serine proteases. PSA and human glandular kallikrein (hK2, previously called hGK-1) share extensive homology and are both produced in the prostate under androgen control. Our goals were to use molecular modeling techniques to generate models of the tertiary structure of PSA and hK2 and to compare their molecular features and areas of homology using these models. Methods. Models of PSA and hK2 were generated by extrapolating from available crystallographic coordinates and amino acid sequences of homologous members of the serine protease family using standard comparative methods. Results. Porcine kallikrein (57% homology) and rat tonin (53% homology) were used as templates for PSA. Porcine kallikrein (67% homology) was used as a template for hK2. The models were superimposed to define regions of nonhomology between PSA and hK2. Conclusions. Three-dimensional protein models of PSA and hK2 were generated. These models have potential uses in analyzing antigen-antibody interactions, modeling of inhibitor complexes of both PSA and hK2, and furthering our understanding of the molecular interactions involved in the clinical detection of PSA and hK2.
Prostate-specific antigen (PSA), a 33 kD tissuespecific glycoprotein, is a very useful marker for the diagnosis and management of prostate cancer.lm3 PSA is a member of the kallikrein family.4-8 The kallikreins are a multigene family of serine proteases. Three kallikrein genes have been defined in humans, designated KLKl, KLKZ, and KLK3.g These genes have been localized to chromosome 19. The KLKl gene product, pancreatic/renal glandular kallikrein (hKl), is produced in the kidney, pancreas, and submandibular gland.lO The gene products of KLKZ, human glandular kallikrein (hK2, formerly hGK-l), and KLK3 (PSA or hK3) share many common features. Both are produced by the prostate epithelium and are under androgen regulation. 11 Their amino acid sequences are highly homologous (77%). Although hK1 and hK2 have trypsin-like activity, PSA has a chymotrypsin-like Submitted accepted
(with
(Rapid Communication): revisions): December
December 29, 1994
U ROLOGYQ / MAY 1995 I VOLUME 45. NUMBER 5
5, 1994,
activity. Significant progress has been made in understanding the biology and biochemistry of PSA. In contrast, although mRNA for hK2 has been identified in the prostate, the gene product has not been isolated from seminal fluid or serum.4-gv11 Definition of the biology and potential diagnostic utility of hK2 requires the development of a sensitive and specific immunoassay. Because the amino acid sequences of PSA and hK2 are highly homologous, numerous reports have recognized the potential cross-reactivity of antibodies to PSA with hK2.12-15 A structural comparison of PSA and hK2 can identify the nonhomologous regions that are critical for selectivity between PSA and hK2 and allow better characterization of current PSA immunoassays and the rational design of peptide-based immunoassays for both PSA and hK2. However, the x-ray structures of PSA and hK2 have not been described in the literature. In recent years, standard comparative methods have been extensively used to construct molecular 801
models of proteins with unknown x-ray structures.i6 The actual model structure is extrapolated from the available crystallographic coordinates and amino acid sequences of homologous members of the protein’s family PSA and hK2 belong to the serine protease family.4-7,17 This family of proteins has been subject to several structural studies based on molecular modeling. l8 Furthermore, the threedimensional structures of several members of this family have been reported in the literature and deposited in the Brookhaven Protein Structure Database.ig These include porcine kallikrein, porcine elastase, bovine trypsin, bovine chymotrypsin, Streptomyces griseus tonin, rat tonin, and rat mast cell protease. Among the database entries, PSA and hK2 are homologous with porcine kallikrein and rat tonin. Our results with modeling PSA and hK2 by comparative methods are described in this report. MATERIAL
AND METHODS
PROTEIN SEQUENCES
Porcine kallikreinZo and rat toninZ1 were selected as being the closest to PSA in sequence homology, as determined by Profile 3D graphic package.22 Porcine kallikreinZo was selected for the building of hK2 model using the same methodology SEQUENCE ALIGNMENT
Pileup from Genetic Computer Group,23 which relies on the identity or homology of amino acids, was used to perform the original alignments. These alignments were then further edited according to the more accurate method of comparative modeling as described by Greer,” which uses threedimensional overlap as the alignment criterion. The final allocation of structurally conserved (SCR) and variable regions (VR) was guided by mutational divergence using homology.22 The residue numbering for PSA and hK2 follows that of the chymotrypsinogen molecule throughout this article, unless otherwise indicated.
evaluated for steric and polarity inconsistencies, and individual side chains were modified when necessary using the Rotor command. A similar protocol was followed to build most of the VR, which present high homology with existing structures. In the case of large deletion or loop extension, the loop library from homology was used. The selected loop was then inserted and the adjacent residues were spliced and minimized to decrease the steric constraints. The models were refined by energy minimization with the Discover program. Initial calculations were performed using a steepest descent algorithm until a global energy derivative less than 100 kcal mol.i Ae2 was achieved. A secondary minimization through conjugate gradient iterative procedures was performed until an acceptable root mean square (RMS) energy gradient was obtained. RESULTS PSA MODEL High homology of PSA with porcine kallikrein (57% identity) and with rat tonin (53%), was obtained using the Bestfit software package.23 The modeling of the SCRs was straightforward. Two VRs, 19-23 and 84-95, using the PSA numbering (Fig. l), presented ambiguous results. We used the porcine kallikrein structure to solve VR 19-23 because PSA and kallikrein have the same number of residues in the loop. Tonin shows two deletions in that loop. We used the loop library from homology to generate a model for the VR 84-95, although it is clear that no single structure will ever be able to represent this flexible loop. The selected loop represents a beta-bend connecting two antiparallel beta-strands (Fig. 2A). After the final energy minimization was performed, no nonbond distance of less than 2.0 A was contained in the model. The RMS derivative for the alpha carbon trace of the model was 1.3 A when compared with porcine kallikrein. This RMS value excludes seven residues from the nondefined VR 84-95 loop.
STRUCTURAL ALIGNMENT
The porcine kallikreinZo and rat tonin2’ atomic coordinate files from the Brookhaven Protein Structure Databanklg were read into the Insight II graphics package. Models of PSA and hK2 were built independently using the homology software package. 22 Assignment of SCR and VR was performed using the Greer18 comparative method applied to serine proteases. Replacement of residues within the SCR were automatically performed by homology using the Phi and Psi angles of the alpha carbon backbone. Each mutation was then
802
hK2 MODEL
The model of hK2 was accomplished by comparative modeling methods using porcine kallikrein as a template (Fig. 3). High homology of hK2 with porcine kallikrein (63% identity) was obtained using the Bestfit software package.23 The modeling of the SCRs presented no major challenges. Again, two VRs (84-95 and 138-142, using the chymotrypsinogen numbering) presented ambiguous results. VR 84-95 was modeled with the same method of interpretation used for PSA. UROLOGY@! MAY~~~~/VOLUME~~,NUMBER~
FIGURE 1. Sequence alignment for various serine proteinases. The alignment is performed based on the overlap of the threedimensional structures as described by Creer. 78 The upper residue numbering given for all the serine proteinases follows that of the chymotrypsinogen molecule. The lower residue numbering follows that of the prostatespecific antigen molecule. CHT = bovine chymotrypsinogen; TON = rat tonin; KAL = porcine kallikrein; HK2 = human glandular kallikrein1; PSA = human prostate-specific antigen. Residues critical for enzymatic activity are shown in bold. Selected loops for specific detection of hK2 are underlined. Clycosylation sites for PSA and hK2 are shaded.
CHT TON KAL HK2 PSA
17 IVNGEEAVPG IVGGYKCEKN IIGGRECEKN IVGGWECEKH IVGGWECEKH 1
SWPWQVSLQD SQPWQVAVIN SHPWQVAIYH SQPWQVAVYS SQPWQVLVAS
CHT TON KAL HK2 PSA
67 WAGEFDQGS VLLGRNNLFK VGWLRHNLFE V~LGRHNLFE JLhGmFH 50 loop
SSE.KIQKLK .DEPFAQRRL .NENTAQFFG .PEDTGQRVP .PEDTGQVFQ 3
IAKVFKNSKY NSLTI..... VRQSFRHPDY IPLIVTNDTE VTADFPHPGF NLSADGK... VSHSFPHPLY ~SLLKHQSL VSHSFPH-DRPGDDSSHDL loop 4
CHT TON KAL HK2 PSA
106 TLLKLSTAAS MLLHLSEPAD MLLRLQSPAK MLLRLSEPAK MLLRLSEPAE 99
FSQTVSAVCL ITGGVKVIDL ITDAVKVLEL ITDVVKVLGL LTDAVKVMDL
PSASDDFAAG PT..KEPKVG PT..QEPELG PT..QEPALG PT..eEPALG
TTCVTTGWGL STCLASGWGS STCEASGWGS TTCYASGWGS TTCYASGWGS
151 TRYTN..ANT TNPSE..MW IEPGPDDFEF IEPEE..FLR 1EPEEm.m 138 loop 5
CHT TON KAL HK2 PSA
154 PDRLQQASLP SHDLQCVNIH PDEIQCVQLT PRSL&VSLH =QCVDLH 146
LLSNTNCKKY LLSNEKCIET LLQNTFCAHA LLikDMCARA VISW loop
WGTKIKDAMI YKDNVTDVML BPBKVTESML YSEKVTEFML
CAGA..SGVS CAGEMEGGKD CAGYLPGGKD CAGLWTGGKD CAGRWTGGKS
199 SCMGDSGGPL TCAGDSGGPL TCMGDSGGPL TCGGDSGGPL TCSGDSGGPL 193
YARVTALVNW YAKLIKFTSW YTKLIFYLDW YTKWHYRKW YTKVVHYRKW
245 VQQTLAAN IKKVMKENP IBBTITENP IKDTIAANP IKDTIVANP 237
CHT TON KAL HK2 PSA
202 VCKKNGAWTL VGIVSWGSS. ICD.. ..GVL QGITSGGATP ICN.. ..GMW QGITSWGHTP VCN.. ..GVL QGITSWGPEP VCN.. ..GVL QGITSWGSEP 196
FIGURE 2. (A) Overlay of the trace atoms of the (white), and human prostate-specific antigen (red). of porcine kallikrein (white) and human glandular teins for selected variable regions (VRs) are shown
UROLOGY@ I MAY 1995 I VOLUME45,
NUMBER
5
KTGFHFCGGS . . .EYLCGGV Y.SSFQCGGV H.GWtiCGGV R.GRAVC GGV loop 1
LINENWVVTA LIDPSWVITA LVNPKWVLTA LVHPQWVLTA LVHPQWVLTA
6 TCSTS.TPGV CAKP.KTPAI CGSA.NKPSI CALP.EKPAV CALP.ERPSL
65 AHCGVTTSDV AHCYSN.NYQ AXXN.DNYE AHCLK.KNSQ AHCIR. %PKSV loop 2 48 103 . . . . ..NNDI QPVHDHSNDL . . ..DYSHDL RPDEDSSHDL 97
alpha carbon backbone of rat tonin (blue), porcine kallikrein [B) Overlay of the trace atoms of the alpha carbon backbone kallikrein, hK2 (green). Divergences among the different proon the figures.
803
FIGURE 3. Space-filling model of prostate-specific antigen (red) and human glandular kallikrein etoms of the catalytic triad are represented in green and yellow. Clycosylation sites are represented A subset showing the active site is represented in white. Partially buried residue 189 responsible specificity is represented in pale blue.
As already described for PSA, a seven-residue insertion was added to the protein, resulting in a beta-bend connecting two antiparallel betastrands (Fig. 2B). Like PSA, the size of that VR and the lack of precise structural information strongly decrease the value of the modeled protein in that region. VR 138142 was modeled using the loop library from homology.22 The selected loop was fitted between the known residues. After the insertion, the four neighboring residues were minimized to release the steric constraints. After the final energy minimization was performed, no nonbond distance of less than 2.0 A was contained in the model (Fig. 2B). The RMS derivative for the alpha carbon trace of the model was 1.4 A when compared with porcine kallikrein. This RMS value excluded seven residues from the nondefined VR 84-95 loop. COMMENT The three-dimensional structures of PSA and hK2 were modeled using comparative modeling methods. A view of the final model with spacefilling shells demonstrates the global topology and active site cleft of the two models and is shown in Figure 3. Both structures carry the catalytic triad, His 41, Asp 96, and Ser 189 (His 45, Asp 102, and Ser 195 in chymotrypsinogen numbering), which is critical for the serine protease-like activity The
804
hK2 (purple). in blue. Eight for substrate
aspartic acid residue 183 of hK2 (Asp 189 in chymotrypsinogen numbering), which is critical for substrate specificity, gives the protein a restricted trypsin-like activity Replacement of the aspartic acid 189 of hK2 into a serine gives PSA a chymotrypsin-like activity These residues are represented in Figure 3, which shows with more detail the cleft of the active site. Despite previous reports, only one glycosylation site has been identified for PSA at position 45. hK2 is glycosylated at residue 78. Both amino-terminus and carboxyterminus are represented on the model. The 16 amino acid carboxy-terminus of both proteins has been described as highly antigenic.r3 Superposition of the two models was performed using Insight II. 22 Regions corn mon to the two proteins were highlighted in blue (Fig. 4). Areas of nonhomology between PSA and hK2 that are critical for the development of selective immunoassays have been highlighted in white. Six linear peptides were identified as potential peptide immunogens to produce antibodies for the specific detection of hK2 in the presence of PSA. These peptides were selected because the residues are on the surface of the protein, are located in sharp turns, and are located in areas of nonhomology between PSA and hK2 (as shown on Fig. 4). These six peptides (called loop l-6) are described in Figures 1 and 5.
UROLOGY@ / MAY 1995 I VOLUME 45, NUMBER 5
FIGURE 4. Structural comparison of prostate-specific antigen [PSA) and hK2. Figure represents the alpha carbon trace of PSA showing structurally conserved regions (blue) and variable regions (white) when compared with hK2. Cysteine residues and sulfur atoms (space filling) are represented in yellow. Catalytic triad is represented in red [His 4 1, Asp 96, and Ser 189). N-terminus is represented in green and C-terminus in brown. Clycosylation site for PSA is in blue and for hK2 in purple. Residue Ser 183 critical for chymotrypsin activity of PSA is represented in dark yellow.
Recently, successful production of sheep antipeptide antisera to the 41-56 peptide region has been reported. 24 This antisera was used in a competitive immunoassay and for affinity chromatography. This peptide is represented in loops 2 and 3 of our model (Fig. 5). Assessment of each loop is currently in progress for the synthesis of specific peptide epitopes and the improvement of existing peptide-based immunoassays. These models can be used for functional analysis of antibody-antigen interactions, for the generation of a model representing the PSA-antichymotrypsin complex, and more generally for a better understanding of protein interactions involved in the detection of serum PSA. Subsequent to the submission of this manuscript, two related articles have been published.25,26 Dominique P. Bridon, Ph.D. Department of Peptide Engineering Diagnostics Division Abbott Laboratories Abbott Park, IL
U ROLOCY@
/ MAY 1995
i VOLUME
45,
NUMBER
5
FIGURE 5. Selected loops for the specific detection of hK2 in presence of prostate-specific antigen (PSA).Structurally conserved regionsbetweeen PSA and hK2 are indicated in blue. Nonhomologous regions are indicated in white. Selected loops are indicated in red. Complete amino acids description of each loop can be found in Figure 1. Catalytic triad is represented in green. Ser 183 responsible for substrate specificity is in pink. Clycosylation site Asn 45 is colored in dark blue.
REFERENCES 1. Oesterling JE: Prostate specific antigen: a critical assessment of the most useful tumor marker for adenocarcinoma of the prostate. J Urol 145: 907-923, 1991. 2. Stamey TA, Yang N, Hay AR, McNeal JE, Freiha FS, and Redwine E: Prostate-specific antigen as a serum marker for adenocarcinoma of the prostate. N Engl J Med 317: 909-916, 1987. 3. Brawer MK, and Lange PH: Prostate specific antigen: its role in early detection, staging, and monitoring of prostatic carcinoma. J Endourol 3: 227-236, 1989. 4. Henttu P, and Vihko P: cDNA coding for the entire human prostate specific antigen shows high homologies to the human tissue kallikrein genes. Biochem Biophys Res Commun 160: 903-910, 1989. 5. Henttu P, and Vihko P: Prostate-specific antigen and human glandular kallikrein: two kallikreins of the human prostate. Ann Med 26: 157-164, 1994. 6. Schedlich LJ, Bennetts BH, and Morris B: Primary structure of a human glandular kallikrein gene. DNA 6: 429-437, 1987. 7. Lundwall A, and Lilja H: Molecular cloning of human prostate specific antigen cDNA. FEBS Lett 214: 317-322,1987. 8. Digby M, Zhang XY, and Richards RI: Human prostate specific antigen (PSA) gene: structure and linkage to the kallikrein-like gene, hK2. Nucleic Acids Res 17: 2137, 1989. 9. Berg T, Bradshaw RA, Carretero OA, Chao J, Chao L, Clements JA, Fahnestock M, Fritz H, Gauthier F, MacDonald RJ, et al: A common nomenclature for members of the tissue (glandular) kallikrein gene families. Agents Actions 38 (suppl): 19-25, 1992. 10. Murtha P, Tindall DJ, and Young CY: Androgen induction of a human prostate-specific kallikrein, hKLK2:
805
characterization of an androgen response element in the 5’ promoter region of the gene. Biochemistry 32: 6459-6464, 1993. 11. Clements J, and Mukhtar A: Glandular kallikreins and prostate-specific antigen are expressed in the human endometrium J Clin Endocrinol Metab 78: 1536-1539, 1994. 12. Vessella RL, and Lange PH: Issues in the assessment of PSA immunoassays. Urol Clin North Am 20: 607-619, 1993. 13. Ambruster DA: Prostate-specific antigen: biochemistry, analytical methods, and clinical application. Clin Chem 39: 181-195, 1993. 14. Lilja H: Structure, function, and regulation of the enzyme activity of prostate-specific antigen. World J Urol 11: 188-191, 1993. 15. Lilja H, Cockett AT, and Abrahamsson PA: Prostate specific antigen predominantly forms a complex with alpharantichymotrypsin in blood. Cancer 70: 230-234, 1992. 16. Johnson MS, Srinivasan N, Sowdhamini R, and Blundell TL: Knowledge-based protein modeling. Crit Rev Biochem Mol Bio129: l-68, 1994. 17. Watt KW, Lee PJ, M’Timkulu T, Chan WP, and Loor R: Human prostate specific antigen: structural and functional similarity with serine proteases. Proc Nat1 Acad Sci USA 83: 3166-3170, 1986. 18. Greer J: Comparative modeling methods: application to the family of the mammalian serine proteases. Proteins 7: 317-334, 1990.
806
19. Bernstein FC, Koetzle TF, Williams GJ, Meyer EE Jr, Brice MD, Rodgers JR, Kennard 0, Shimanouchi T, and Tasumi M: The Protein Data Bank: a computer-based archival file for macromolecular structures. J Mol Biol 112: 535-542, 1977. 20. Bode W, Chen Z, Bartels K, Kutzbach C, SchmidtKastner G, and Bartunik H: Refined 2 A X-ray crystal structure of porcine pancreatic kallikrein A, a specific trypsin-like serine proteinase. Crystallization, structure determination and its comparison with bovine trypsin. J Mol Biol 164: 237-282, 1983. 21. Fujinaga M, and James MN: Rat submaxillary gland serine proteinase, tonin. Structure solution and refinement at 1.8 A resolution. J Mol Biol 195: 373-396, 1987. 22. Profile 3D, Insight II ~3.0, Discover ~3.0 and Homology ~2.1.0, Biosym Technology, San Diego, Calif. 23. Genetic Computer Group, University Research Park, 575 Science Drive, Suite B, Madison, Wis. 24. Finlay JA, Klee GG, Hogen K, Egli PA, Mikolajczk S, Young CY, Tindall DJ, Rittenhouse HG, and Wolfert RL. Development of an immunoassay for human glandular kallikrein (hK2) (abstract). Presented at 1994 Fall Meeting of Society for Basic Urological Research, Stanford, October 20-23,1994. 25. Vihinen M: Modeling of prostate specific antigen and human glandular kallikrein structures. Biochem Biophy Commun Acta 204: 1251-1256, 1994. 26. Villoutreix BO, Getzoff ED, and Griffin JH: A structural model for the prostate disease marker prostate-specific antigen. Protein Sci 3: 2033-2044, 1994.
UROLOGY@ /MAY
1995 I VOLUME 45, NUMBER
5
UROLOGY
STRUCTURAL COMPARISON OF PROSTATE-SPECIFIC ANTIGEN AND HUMAN GLANDULAR KALLIKREIN USING MOLECULAR MODELING DOMINIQUE P BRIDON, PH.D. BARRY L. DOWELL, PH.D. From the Department of Peptide Engineering and the Department of Cancer Product Development, Diagnostic Division, Abbott Laboratories, Abbott Park, Illinois
ABSTRACT-Objectives. Prostate-specific antigen (PSA), the most useful tumor marker for prostate cancer, is one of three members of the human kallikrein family of serine proteases. PSA and human glandular kallikrein (hK2, previously called hGK-1) share extensive homology and are both produced in the prostate under androgen control. Our goals were to use molecular modeling techniques to generate models of the tertiary structure of PSA and hK2 and to compare their molecular features and areas of homology using these models. Methods. Models of PSA and hK2 were generated by extrapolating from available crystallographic coordinates and amino acid sequences of homologous members of the serine protease family using standard comparative methods. Results. Porcine kallikrein (57% homology) and rat tonin (53% homology) were used as templates for PSA. Porcine kallikrein (67% homology) was used as a template for hK2. The models were superimposed to define regions of nonhomology between PSA and hK2. Conclusions. Three-dimensional protein models of PSA and hK2 were generated. These models have potential uses in analyzing antigen-antibody interactions, modeling of inhibitor complexes of both PSA and hK2, and furthering our understanding of the molecular interactions involved in the clinical detection of PSA and hK2.
Prostate-specific antigen (PSA), a 33 kD tissuespecific glycoprotein, is a very useful marker for the diagnosis and management of prostate cancer.lm3 PSA is a member of the kallikrein family.4-8 The kallikreins are a multigene family of serine proteases. Three kallikrein genes have been defined in humans, designated KLKl, KLKZ, and KLK3.g These genes have been localized to chromosome 19. The KLKl gene product, pancreatic/renal glandular kallikrein (hKl), is produced in the kidney, pancreas, and submandibular gland.lO The gene products of KLKZ, human glandular kallikrein (hK2, formerly hGK-l), and KLK3 (PSA or hK3) share many common features. Both are produced by the prostate epithelium and are under androgen regulation. 11 Their amino acid sequences are highly homologous (77%). Although hK1 and hK2 have trypsin-like activity, PSA has a chymotrypsin-like Submitted accepted
(with
(Rapid Communication): revisions): December
December 29, 1994
U ROLOGYQ / MAY 1995 I VOLUME 45. NUMBER 5
5, 1994,
activity. Significant progress has been made in understanding the biology and biochemistry of PSA. In contrast, although mRNA for hK2 has been identified in the prostate, the gene product has not been isolated from seminal fluid or serum.4-gv11 Definition of the biology and potential diagnostic utility of hK2 requires the development of a sensitive and specific immunoassay. Because the amino acid sequences of PSA and hK2 are highly homologous, numerous reports have recognized the potential cross-reactivity of antibodies to PSA with hK2.12-15 A structural comparison of PSA and hK2 can identify the nonhomologous regions that are critical for selectivity between PSA and hK2 and allow better characterization of current PSA immunoassays and the rational design of peptide-based immunoassays for both PSA and hK2. However, the x-ray structures of PSA and hK2 have not been described in the literature. In recent years, standard comparative methods have been extensively used to construct molecular 801
models of proteins with unknown x-ray structures.i6 The actual model structure is extrapolated from the available crystallographic coordinates and amino acid sequences of homologous members of the protein’s family PSA and hK2 belong to the serine protease family.4-7,17 This family of proteins has been subject to several structural studies based on molecular modeling. l8 Furthermore, the threedimensional structures of several members of this family have been reported in the literature and deposited in the Brookhaven Protein Structure Database.ig These include porcine kallikrein, porcine elastase, bovine trypsin, bovine chymotrypsin, Streptomyces griseus tonin, rat tonin, and rat mast cell protease. Among the database entries, PSA and hK2 are homologous with porcine kallikrein and rat tonin. Our results with modeling PSA and hK2 by comparative methods are described in this report. MATERIAL
AND METHODS
PROTEIN SEQUENCES
Porcine kallikreinZo and rat toninZ1 were selected as being the closest to PSA in sequence homology, as determined by Profile 3D graphic package.22 Porcine kallikreinZo was selected for the building of hK2 model using the same methodology SEQUENCE ALIGNMENT
Pileup from Genetic Computer Group,23 which relies on the identity or homology of amino acids, was used to perform the original alignments. These alignments were then further edited according to the more accurate method of comparative modeling as described by Greer,” which uses threedimensional overlap as the alignment criterion. The final allocation of structurally conserved (SCR) and variable regions (VR) was guided by mutational divergence using homology.22 The residue numbering for PSA and hK2 follows that of the chymotrypsinogen molecule throughout this article, unless otherwise indicated.
evaluated for steric and polarity inconsistencies, and individual side chains were modified when necessary using the Rotor command. A similar protocol was followed to build most of the VR, which present high homology with existing structures. In the case of large deletion or loop extension, the loop library from homology was used. The selected loop was then inserted and the adjacent residues were spliced and minimized to decrease the steric constraints. The models were refined by energy minimization with the Discover program. Initial calculations were performed using a steepest descent algorithm until a global energy derivative less than 100 kcal mol.i Ae2 was achieved. A secondary minimization through conjugate gradient iterative procedures was performed until an acceptable root mean square (RMS) energy gradient was obtained. RESULTS PSA MODEL High homology of PSA with porcine kallikrein (57% identity) and with rat tonin (53%), was obtained using the Bestfit software package.23 The modeling of the SCRs was straightforward. Two VRs, 19-23 and 84-95, using the PSA numbering (Fig. l), presented ambiguous results. We used the porcine kallikrein structure to solve VR 19-23 because PSA and kallikrein have the same number of residues in the loop. Tonin shows two deletions in that loop. We used the loop library from homology to generate a model for the VR 84-95, although it is clear that no single structure will ever be able to represent this flexible loop. The selected loop represents a beta-bend connecting two antiparallel beta-strands (Fig. 2A). After the final energy minimization was performed, no nonbond distance of less than 2.0 A was contained in the model. The RMS derivative for the alpha carbon trace of the model was 1.3 A when compared with porcine kallikrein. This RMS value excludes seven residues from the nondefined VR 84-95 loop.
STRUCTURAL ALIGNMENT
The porcine kallikreinZo and rat tonin2’ atomic coordinate files from the Brookhaven Protein Structure Databanklg were read into the Insight II graphics package. Models of PSA and hK2 were built independently using the homology software package. 22 Assignment of SCR and VR was performed using the Greer18 comparative method applied to serine proteases. Replacement of residues within the SCR were automatically performed by homology using the Phi and Psi angles of the alpha carbon backbone. Each mutation was then
802
hK2 MODEL
The model of hK2 was accomplished by comparative modeling methods using porcine kallikrein as a template (Fig. 3). High homology of hK2 with porcine kallikrein (63% identity) was obtained using the Bestfit software package.23 The modeling of the SCRs presented no major challenges. Again, two VRs (84-95 and 138-142, using the chymotrypsinogen numbering) presented ambiguous results. VR 84-95 was modeled with the same method of interpretation used for PSA. UROLOGY@! MAY~~~~/VOLUME~~,NUMBER~
FIGURE 1. Sequence alignment for various serine proteinases. The alignment is performed based on the overlap of the threedimensional structures as described by Creer. 78 The upper residue numbering given for all the serine proteinases follows that of the chymotrypsinogen molecule. The lower residue numbering follows that of the prostatespecific antigen molecule. CHT = bovine chymotrypsinogen; TON = rat tonin; KAL = porcine kallikrein; HK2 = human glandular kallikrein1; PSA = human prostate-specific antigen. Residues critical for enzymatic activity are shown in bold. Selected loops for specific detection of hK2 are underlined. Clycosylation sites for PSA and hK2 are shaded.
CHT TON KAL HK2 PSA
17 IVNGEEAVPG IVGGYKCEKN IIGGRECEKN IVGGWECEKH IVGGWECEKH 1
SWPWQVSLQD SQPWQVAVIN SHPWQVAIYH SQPWQVAVYS SQPWQVLVAS
CHT TON KAL HK2 PSA
67 WAGEFDQGS VLLGRNNLFK VGWLRHNLFE V~LGRHNLFE JLhGmFH 50 loop
SSE.KIQKLK .DEPFAQRRL .NENTAQFFG .PEDTGQRVP .PEDTGQVFQ 3
IAKVFKNSKY NSLTI..... VRQSFRHPDY IPLIVTNDTE VTADFPHPGF NLSADGK... VSHSFPHPLY ~SLLKHQSL VSHSFPH-DRPGDDSSHDL loop 4
CHT TON KAL HK2 PSA
106 TLLKLSTAAS MLLHLSEPAD MLLRLQSPAK MLLRLSEPAK MLLRLSEPAE 99
FSQTVSAVCL ITGGVKVIDL ITDAVKVLEL ITDVVKVLGL LTDAVKVMDL
PSASDDFAAG PT..KEPKVG PT..QEPELG PT..QEPALG PT..eEPALG
TTCVTTGWGL STCLASGWGS STCEASGWGS TTCYASGWGS TTCYASGWGS
151 TRYTN..ANT TNPSE..MW IEPGPDDFEF IEPEE..FLR 1EPEEm.m 138 loop 5
CHT TON KAL HK2 PSA
154 PDRLQQASLP SHDLQCVNIH PDEIQCVQLT PRSL&VSLH =QCVDLH 146
LLSNTNCKKY LLSNEKCIET LLQNTFCAHA LLikDMCARA VISW loop
WGTKIKDAMI YKDNVTDVML BPBKVTESML YSEKVTEFML
CAGA..SGVS CAGEMEGGKD CAGYLPGGKD CAGLWTGGKD CAGRWTGGKS
199 SCMGDSGGPL TCAGDSGGPL TCMGDSGGPL TCGGDSGGPL TCSGDSGGPL 193
YARVTALVNW YAKLIKFTSW YTKLIFYLDW YTKWHYRKW YTKVVHYRKW
245 VQQTLAAN IKKVMKENP IBBTITENP IKDTIAANP IKDTIVANP 237
CHT TON KAL HK2 PSA
202 VCKKNGAWTL VGIVSWGSS. ICD.. ..GVL QGITSGGATP ICN.. ..GMW QGITSWGHTP VCN.. ..GVL QGITSWGPEP VCN.. ..GVL QGITSWGSEP 196
FIGURE 2. (A) Overlay of the trace atoms of the (white), and human prostate-specific antigen (red). of porcine kallikrein (white) and human glandular teins for selected variable regions (VRs) are shown
UROLOGY@ I MAY 1995 I VOLUME45,
NUMBER
5
KTGFHFCGGS . . .EYLCGGV Y.SSFQCGGV H.GWtiCGGV R.GRAVC GGV loop 1
LINENWVVTA LIDPSWVITA LVNPKWVLTA LVHPQWVLTA LVHPQWVLTA
6 TCSTS.TPGV CAKP.KTPAI CGSA.NKPSI CALP.EKPAV CALP.ERPSL
65 AHCGVTTSDV AHCYSN.NYQ AXXN.DNYE AHCLK.KNSQ AHCIR. %PKSV loop 2 48 103 . . . . ..NNDI QPVHDHSNDL . . ..DYSHDL RPDEDSSHDL 97
alpha carbon backbone of rat tonin (blue), porcine kallikrein [B) Overlay of the trace atoms of the alpha carbon backbone kallikrein, hK2 (green). Divergences among the different proon the figures.
803
FIGURE 3. Space-filling model of prostate-specific antigen (red) and human glandular kallikrein etoms of the catalytic triad are represented in green and yellow. Clycosylation sites are represented A subset showing the active site is represented in white. Partially buried residue 189 responsible specificity is represented in pale blue.
As already described for PSA, a seven-residue insertion was added to the protein, resulting in a beta-bend connecting two antiparallel betastrands (Fig. 2B). Like PSA, the size of that VR and the lack of precise structural information strongly decrease the value of the modeled protein in that region. VR 138142 was modeled using the loop library from homology.22 The selected loop was fitted between the known residues. After the insertion, the four neighboring residues were minimized to release the steric constraints. After the final energy minimization was performed, no nonbond distance of less than 2.0 A was contained in the model (Fig. 2B). The RMS derivative for the alpha carbon trace of the model was 1.4 A when compared with porcine kallikrein. This RMS value excluded seven residues from the nondefined VR 84-95 loop. COMMENT The three-dimensional structures of PSA and hK2 were modeled using comparative modeling methods. A view of the final model with spacefilling shells demonstrates the global topology and active site cleft of the two models and is shown in Figure 3. Both structures carry the catalytic triad, His 41, Asp 96, and Ser 189 (His 45, Asp 102, and Ser 195 in chymotrypsinogen numbering), which is critical for the serine protease-like activity The
804
hK2 (purple). in blue. Eight for substrate
aspartic acid residue 183 of hK2 (Asp 189 in chymotrypsinogen numbering), which is critical for substrate specificity, gives the protein a restricted trypsin-like activity Replacement of the aspartic acid 189 of hK2 into a serine gives PSA a chymotrypsin-like activity These residues are represented in Figure 3, which shows with more detail the cleft of the active site. Despite previous reports, only one glycosylation site has been identified for PSA at position 45. hK2 is glycosylated at residue 78. Both amino-terminus and carboxyterminus are represented on the model. The 16 amino acid carboxy-terminus of both proteins has been described as highly antigenic.r3 Superposition of the two models was performed using Insight II. 22 Regions corn mon to the two proteins were highlighted in blue (Fig. 4). Areas of nonhomology between PSA and hK2 that are critical for the development of selective immunoassays have been highlighted in white. Six linear peptides were identified as potential peptide immunogens to produce antibodies for the specific detection of hK2 in the presence of PSA. These peptides were selected because the residues are on the surface of the protein, are located in sharp turns, and are located in areas of nonhomology between PSA and hK2 (as shown on Fig. 4). These six peptides (called loop l-6) are described in Figures 1 and 5.
UROLOGY@ / MAY 1995 I VOLUME 45, NUMBER 5
FIGURE 4. Structural comparison of prostate-specific antigen [PSA) and hK2. Figure represents the alpha carbon trace of PSA showing structurally conserved regions (blue) and variable regions (white) when compared with hK2. Cysteine residues and sulfur atoms (space filling) are represented in yellow. Catalytic triad is represented in red [His 4 1, Asp 96, and Ser 189). N-terminus is represented in green and C-terminus in brown. Clycosylation site for PSA is in blue and for hK2 in purple. Residue Ser 183 critical for chymotrypsin activity of PSA is represented in dark yellow.
Recently, successful production of sheep antipeptide antisera to the 41-56 peptide region has been reported. 24 This antisera was used in a competitive immunoassay and for affinity chromatography. This peptide is represented in loops 2 and 3 of our model (Fig. 5). Assessment of each loop is currently in progress for the synthesis of specific peptide epitopes and the improvement of existing peptide-based immunoassays. These models can be used for functional analysis of antibody-antigen interactions, for the generation of a model representing the PSA-antichymotrypsin complex, and more generally for a better understanding of protein interactions involved in the detection of serum PSA. Subsequent to the submission of this manuscript, two related articles have been published.25,26 Dominique P. Bridon, Ph.D. Department of Peptide Engineering Diagnostics Division Abbott Laboratories Abbott Park, IL
U ROLOCY@
/ MAY 1995
i VOLUME
45,
NUMBER
5
FIGURE 5. Selected loops for the specific detection of hK2 in presence of prostate-specific antigen (PSA).Structurally conserved regionsbetweeen PSA and hK2 are indicated in blue. Nonhomologous regions are indicated in white. Selected loops are indicated in red. Complete amino acids description of each loop can be found in Figure 1. Catalytic triad is represented in green. Ser 183 responsible for substrate specificity is in pink. Clycosylation site Asn 45 is colored in dark blue.
REFERENCES 1. Oesterling JE: Prostate specific antigen: a critical assessment of the most useful tumor marker for adenocarcinoma of the prostate. J Urol 145: 907-923, 1991. 2. Stamey TA, Yang N, Hay AR, McNeal JE, Freiha FS, and Redwine E: Prostate-specific antigen as a serum marker for adenocarcinoma of the prostate. N Engl J Med 317: 909-916, 1987. 3. Brawer MK, and Lange PH: Prostate specific antigen: its role in early detection, staging, and monitoring of prostatic carcinoma. J Endourol 3: 227-236, 1989. 4. Henttu P, and Vihko P: cDNA coding for the entire human prostate specific antigen shows high homologies to the human tissue kallikrein genes. Biochem Biophys Res Commun 160: 903-910, 1989. 5. Henttu P, and Vihko P: Prostate-specific antigen and human glandular kallikrein: two kallikreins of the human prostate. Ann Med 26: 157-164, 1994. 6. Schedlich LJ, Bennetts BH, and Morris B: Primary structure of a human glandular kallikrein gene. DNA 6: 429-437, 1987. 7. Lundwall A, and Lilja H: Molecular cloning of human prostate specific antigen cDNA. FEBS Lett 214: 317-322,1987. 8. Digby M, Zhang XY, and Richards RI: Human prostate specific antigen (PSA) gene: structure and linkage to the kallikrein-like gene, hK2. Nucleic Acids Res 17: 2137, 1989. 9. Berg T, Bradshaw RA, Carretero OA, Chao J, Chao L, Clements JA, Fahnestock M, Fritz H, Gauthier F, MacDonald RJ, et al: A common nomenclature for members of the tissue (glandular) kallikrein gene families. Agents Actions 38 (suppl): 19-25, 1992. 10. Murtha P, Tindall DJ, and Young CY: Androgen induction of a human prostate-specific kallikrein, hKLK2:
805
characterization of an androgen response element in the 5’ promoter region of the gene. Biochemistry 32: 6459-6464, 1993. 11. Clements J, and Mukhtar A: Glandular kallikreins and prostate-specific antigen are expressed in the human endometrium J Clin Endocrinol Metab 78: 1536-1539, 1994. 12. Vessella RL, and Lange PH: Issues in the assessment of PSA immunoassays. Urol Clin North Am 20: 607-619, 1993. 13. Ambruster DA: Prostate-specific antigen: biochemistry, analytical methods, and clinical application. Clin Chem 39: 181-195, 1993. 14. Lilja H: Structure, function, and regulation of the enzyme activity of prostate-specific antigen. World J Urol 11: 188-191, 1993. 15. Lilja H, Cockett AT, and Abrahamsson PA: Prostate specific antigen predominantly forms a complex with alpharantichymotrypsin in blood. Cancer 70: 230-234, 1992. 16. Johnson MS, Srinivasan N, Sowdhamini R, and Blundell TL: Knowledge-based protein modeling. Crit Rev Biochem Mol Bio129: l-68, 1994. 17. Watt KW, Lee PJ, M’Timkulu T, Chan WP, and Loor R: Human prostate specific antigen: structural and functional similarity with serine proteases. Proc Nat1 Acad Sci USA 83: 3166-3170, 1986. 18. Greer J: Comparative modeling methods: application to the family of the mammalian serine proteases. Proteins 7: 317-334, 1990.
806
19. Bernstein FC, Koetzle TF, Williams GJ, Meyer EE Jr, Brice MD, Rodgers JR, Kennard 0, Shimanouchi T, and Tasumi M: The Protein Data Bank: a computer-based archival file for macromolecular structures. J Mol Biol 112: 535-542, 1977. 20. Bode W, Chen Z, Bartels K, Kutzbach C, SchmidtKastner G, and Bartunik H: Refined 2 A X-ray crystal structure of porcine pancreatic kallikrein A, a specific trypsin-like serine proteinase. Crystallization, structure determination and its comparison with bovine trypsin. J Mol Biol 164: 237-282, 1983. 21. Fujinaga M, and James MN: Rat submaxillary gland serine proteinase, tonin. Structure solution and refinement at 1.8 A resolution. J Mol Biol 195: 373-396, 1987. 22. Profile 3D, Insight II ~3.0, Discover ~3.0 and Homology ~2.1.0, Biosym Technology, San Diego, Calif. 23. Genetic Computer Group, University Research Park, 575 Science Drive, Suite B, Madison, Wis. 24. Finlay JA, Klee GG, Hogen K, Egli PA, Mikolajczk S, Young CY, Tindall DJ, Rittenhouse HG, and Wolfert RL. Development of an immunoassay for human glandular kallikrein (hK2) (abstract). Presented at 1994 Fall Meeting of Society for Basic Urological Research, Stanford, October 20-23,1994. 25. Vihinen M: Modeling of prostate specific antigen and human glandular kallikrein structures. Biochem Biophy Commun Acta 204: 1251-1256, 1994. 26. Villoutreix BO, Getzoff ED, and Griffin JH: A structural model for the prostate disease marker prostate-specific antigen. Protein Sci 3: 2033-2044, 1994.
UROLOGY@ /MAY
1995 I VOLUME 45, NUMBER
5