Pergamon

Progress in Growth Factor Research, Vol. 6. Nos. 2~l ., pp. 255-263, 1995 Copyright © 1996 Elsevier Science Ltd. All rights reserved

Printed in Great Britain. 0955-2235/95 $29.00 + .00

0955-2235(95)00017-8

MATRIX METALLOPROTEINASES AS INSULIN- LIKE GROWTH FACTOR BINDING PROTEIN-

DEGRADING PROTEINASES

John L. Fowlkes, *t Kathryn M. Thrailkill, t Delila M. Serra, t Ko Suzuki~ and Hideaki Nagase:l:

tDepartments of Pediatrics, Duke University Medical Center, Durham, NC 27710, U.S.A. ~Department of Biochemistry and Molecular Biology, University of Kansas Medical Center,

Kansas City, KS 66160, U.S.A.

Insulin-like growth factor (IGF) bioavailability is modulated by specific IGFBPs, six o f which are known (IGFBPs 1 4 ) . Since IGFBPs have equal or higher affinity for IGFs than do IGF receptors, it is believed that degradation of IGFBPs by IGFBP- degrading proteinases is an important step in regulating IGF bioactivity. Recent studies from our laboratory have demonstrated that at least two IGFBPs, i.e. IGFBP- 3 and -5, are degraded under physiologic conditions by matrix metalloproteinases ( MMPs). In vitro, we have demonstrated that IGFBP-3 is degraded in human dermal

fibroblast cultures by MMPs using a variety o f techniques, including proteinase inhi- bition with a specific inhibitor o f MMPs, i.e. tissue inhibitor o f metalloproteinases (TIMP-1), immunoabsorption with specific antisera to human MMPs and a unique method developed in our laboratory, IGFBP-3 substrate zymography. Using similar methods, we have also demonstrated that MMPs, along with an unidentified 97-kDa proteinase, degrade IGFBP-5 in murine osteoblast cultures. In rat pregnancy serum, we have shown that degradation of lGFBP-3 is associated with M M P activity present in the serum, which likely arises from the placental compartment. Analysis o f the cleavage products o f IGFBP-3 produced by MMPs 1, 2 and 3 reveals that MMPs cleave exclusively within the non-homologous, mid-region o f the molecule. Together, these studies suggest that MMPs, beyond their previously described roles as extra- cellular matrix degrading enzymes, may also exert effects on cellular growth and differentiation via degradation of lGFBPs.

Keywords: Insulin-like growth factor (IGF), insulin-like growth factor binding protein (IGFBP), matrix metalloproteinase (MMP), stromelysin, interstitial collagnease, gelatinase.

*Correspondence to: J.L. Fowlkes at: Department of Pediatric Endocrinology, Division of Endocrinology, Box 3080, Duke University Medical Center, Durham, NC 27710, U.S.A.

Acknowledgements--We wish to thank Dr Christophere Maack (Celtrix Pharmaceuticals, Santa Clara, CA, U.S.A.) for providing rhlGFBP-3 and IGFBP-3 antisera and Dr Michael Kiefer (Chiron, Corp., Emeryville, CA, U.S.A.) for providing rhlGFBP-5. This work was supported by National Institutes of Health Grants AR39189 (to H.N.) and DK02276 (to J.L.F.), and a March of Dimes Basil O'Connor Starter Scholar Research Award (#5-FY93-0953) to J.L.F. and a Duke Children's Telethon Grant (to K.M.T.).

255

256 J. L. Fowlkes et al.

INTRODUCTION

Insulin-like growth factors (IGF-I and IGF-II) are ubiquitous growth-promoting peptides which exert both mitogenic and/or differentiation-inducing effects on numerous cells and tissues [1, 2]. For IGFs to exert their effects at the cell surface, they first must bind specific, high affinity cell-surface receptors. However, the inter- action of IGFs with cell-surface receptors is modulated in a complex manner by a group of soluble IGF-binding proteins (IGFBPs), six of which have been purified and cloned [1, 2]. To date, several studies have demonstrated that all six IGFBPs have similar or higher affinities for IGF-I and IGF-II than does the type-I IGF receptor [2], suggesting that disruption of the IGF-IGFBP complex is a probable prerequisite to IGF-IGF-receptor interaction. We and others have provided evidence that this can be accomplished by limited proteolysis of the carrier IGFBP.

Limited proteolysis has been described for IGFBPs 2-5, and various proteinases have been postulated to function as IGFBP-degrading proteinases [1, 2]. Our labo- ratories have recently reported that matrix metalloproteinases (MMPs) function as IGFBP-degrading proteinases, both in vitro and in vivo [3-5]. Although this family of zinc-dependent proteinases is generally believed to be important in the degrada- tion and turnover of extracellular matrix molecules, both in physiologic and in pathologic conditions [6-8], our studies suggest a new role for these enzymes as regulators of IGF bioavailability and bioactivity.

MMPS AS IGFBP-DEGRADING PROTEINASES I N VITRO

IGFBP-3

In human dermal fibroblasts, we have demonstrated that both endogenous and exogenous IGFBP-3 is degraded into several immunoreactive fragments by cell-free conditioned media [9]. These studies revealed that the IGFBP-3-degrading proteinases produced by dermal fibroblasts were zinc-dependent [9]. Based on these initial findings, studies were undertaken to identify the metalloproteinases which degrade IGFBP-3.

To characterize the IGFBP-3-degrading metalloproteinases produced by human fibroblasts, we developed and employed IGFBP-3-substrate zymography [4]. This method allows for the determination of the M r of IGFBP-degrading proteinases; used in combination with proteinase inhibitors, IGFBP-3-substrate zymography can also be used to classify proteinases. Figure 1 demonstrates that several distinct IGFBP-3-degrading proteinases (M r 52,000-72,000) were detected in human fibro- blast conditioned media (lane 1), suggesting that more than one IGFBP-3-degrad- ing proteinase was produced by human fibroblasts. The addition of EDTA (10 mM) to the incubation inhibited all of the IGFBP-3-degrading proteinases (lane 2), demonstrating that all detected IGFBP-3-degrading proteinases were bivalent, cation-dependent enzymes. Since the identified proteinases were cation-dependent and their molecular masses were similar to those of previously described gelatin- degrading proteinases, i.e. gelatinases [5, 10], fibroblast conditioned media was also analyzed by gelatin-substrate zymography (lane 3). This method demonstrated that human fibroblast conditioned media contained several gelatinases with identical

MMPs as IGFBP-degrading Proteinases

M lxlO ~1

257

97

69

46

30

- - + - - E D T A

1 2 3 FIGURE 1. Characterization of lGFBP-3-degrading and gelatin-degrading proteinase activity in human fibroblast conditioned media. Samples of human fibroblast conditioned media (50 #1) were analyzed by IGFBP-3-snbstrate zymography in the agsence (lane 1) or the presence (lane 2) of 10 mM EDTA as described in Ref. 141. Dark bands represent the IGFBP-3-degrading proteinase activity in the sample. Human fibroblast conditioned media (50 #1) was also analyzed by gelatin-substrate zymography. Areas of lysis (clear zones) represent gelatin-degrading proteinase activity in the sample (lane 3). Molecular weight markers are indicated on the left.

molecular masses to those that degraded rhlGFBP-3 (compare lanes 3 and 1, respectively). These were the first data suggesting that gelatinases might function as IGFBP-degrading proteinases. Further studies demonstrated that incubation of human fibroblast conditioned medium with the organomercurial 4- aminophenylmercuric acetate (APMA) resulted in a shift in M r of the IGFBP-3- degrading proteinases to several lower molecular weight forms (~ 18-46 kDa). This is consistent with previous studies showing that organomercurials such as APMA convert zymogen forms of MMPs (proMMPs) to lower molecular weight, enzy- matically active species [8].

To confirm the presence of MMPs in human fibroblast conditioned media, the media were immunoblotted with specific antisera generated against human MMP-1 (interstitial collagenase; Fig. 2, lane 1), human MMP-2 (gelatinase A; lane 2), MMP-3 (stromelysin 1; lane 3) or a combination of all three antisera (lane 4). All three proMMPs were detected at their expected molecular masses (proMMP-1, 52/55 kDa; proMMP-2, 72 kDa; and proMMP-3, 57/60 kDa [7, 8]), which were identical in size to IGFBP-3-degrading proteinases detected by IGFBP-3-substrate zymography (lane 5). Moreover, when tissue inhibitor of metalloproteinases (TIMP-1), a specific glycoprotein inhibitor of all MMPs [7, 8], was added to cell- free fibroblast conditioned media containing exogenous rhlGFBP-3, degradation of intact rhlGFBP-3 was inhibited by >80%. These data strongly suggested that

258

Mr(x 1 0"31

J. L. Fowlkes et al.

97.

69 . i~ I . . . . .

46.

30.q

1 2 3 4 5 FIGURE 2. Demonstration that human fibroblast conditioned media contains MMPs. Human fibroblast conditioned medium was immunobintted with specific antisera generated against MMP-1 (lane 1), MMP-2 (lane 2), MMP-3 (lane 3) or a combination of all three antisera (lane 4) as described in REf. [41. For comparison, human fibrohlast conditioned medium was also analyzed by IGFBP-3-substrate zymography (lane 5). Molecular weight markers are indicated on the left.

MMPs were involved in IGFBP-3 degradation in human fibroblast cultures. Further support for this hypothesis was provided when MMP-1, MMP-2 and MMP-3 were sequentially removed from fibroblast conditioned media by immunoabsorption and gelatin-Sepharose chromatography. Removal of MMPs from fibroblast conditioned media eliminated IGFBP-3-degrading proteinase activity in conditioned media as detected by IGFBP-3-substrate zymography, while re-addition of MMP-1, -2 and -3 to MMP-depleted conditioned media restored IGFBP-3-degrading activity. Together, these data demonstrate that the IGFBP-3-degrading metalloproteinases present in fibroblast conditioned media are MMPs, and that MMPs are the princi- ple IGFBP-3-degrading proteinases produced by human fibroblasts.

IGFBP-5

Studies from our laboratory have shown that IGFBP-5 is uniquely regulated in MC3T3-E1 osteoblasts throughout a 30-day period of proliferation and differen- tiation: IGFBP-5 is first detectable in conditioned media of replicating pre-osteoblasts (day 5); IGFBP-5 concentrations then peak between culture days 8-12, and decline to almost undetectable levels thereafter, despite persistence of IGFBP-5 mRNA as late as day 28 [11]. These observations suggested that post- translational mechanisms might be involved in the regulation of IGFBP-5 levels. To determine if IGFBP-5 proteolysis contributed to the disappearance of IGFBP-5 in conditioned media of mature osteoblast cultures, MC3T3-E1 cells were cultured for 30 days and conditioned media from serial time points were examined for their

M M P s as IGFBP-degrading Proteinases 259

ability to degrade [125I]IGFBP-5 [5]. We found that [1:5I]rhlGFBP-5 was degraded by proteinase activity present in conditioned media, and that activity increased in a time-dependent fashion as cultures matured. IGFBP-5-degrading proteinase activ- ity was inhibited by chelating agents, and it was variably inhibited by certain serine proteinase inhibitors. Because bone cells in culture produce MMPs, we hypothe- sized that MMPs produced by MC3T3-E1 osteoblasts accounted for the cation- dependent, IGFBP-5-degrading proteinase activity. Indeed, the addition of TIMP-1 to MC3T3-E1 conditioned media inhibited >85% of IGFBP-5 proteolytic activity, suggesting that MMPs were primary proteinases involved in the degradation of IGFBP-5. Furthermore, using [125I]rhlGFBP-5-substrate zymography, IGFBP-5 degrading proteinases were identified in a broad band with M r 52,000-72,000 and in a more narrow band at M r 97,000 (Fig. 3, lane 1). The activity of all proteinases detected by [125I]rhlGFBP-5-substrate zymography increased throughout differenti- ation, consistent with our previous findings. The 52-72 kDa proteinases were inhib- ited by EDTA (lane 2), demonstrating that they were cation-depending proteinases, and were inhibited by TIMP-1 (lane 4), confirming their identities as members of the MMP family. In contrast, the 97-kDa proteinase was partially inhibited by PMSF (lane 3), but not by EDTA or TIMP-1, suggesting that the 97-kDa proteinase may be a serine proteinase, but not an MMP.

To clarify the identities of the MMPs involved in IGFBP-5 degradation, MC3T3- E1 conditioned media were immunoprecipitated with antisera to human MMP-1, MMP-2, MMP-3 or MMP-9. Only antisera to human MMP-1 and human MMP- 2 immunoprecipitated IGFBP-5 degrading proteinases with M r of 52 kDa and 69/72 kDa, respectively. Consistent with these data, conditioned media from MC3T3-E1 cells contained proteins which were detected by immunoblotting with antisera to human MMP-1 and human MMP-2, and MMP-2 immunoreactivity, in particular, increased significantly throughout differentiation. Together, these studies suggest that murine MMPs antigenically related to human MMP-1 and MMP-2 are involved in IGFBP-5 degradation in MC3T3-EI cultures.

Mr(xl0 -3)

97.4--,

69.0-*

46.0-* . . . . . Control -+ TA +PMSF +TIMP-I

FIGURE 3. Detection and characterization of IGFBP-5-degrading proteinase activity by IJ2SI]rhiGFBP-5 - substrate zymography. MC3T3-E1 conditioned media (50/A) were analyzed by 112Sl]rhlGFBP-5-substrate zymography (see Ref. [10]) in the absence (control) or in the presence of the proteinase inhibitors 10 mM EDTA (lane 2), 10 mM P M S F (lane 3), or 10/zg/ml TIMP-1 (lane 4). Lytic areas represent IGFBP-5- degrading proteinases.

260 J. L. Fowlkes et al.

MMPS AS IGFBP-DEGRADING PROTEINASES IN VIVO

During the later stages of pregnancy in the rodent and human, IGFBP-3 under- goes limited proteolysis [12-16]; this process yields IGFBP-3 fragments which bind IGFs with significantly lower affinity than does intact IGFBP-3 [16]. Through proteolysis of IGFBP-3, free IGF-I concentrations increase in human pregnancy serum [16]. In the pregnant rat, proteolysis of IGFBP-3 is associated with increased intravascular IGF-I clearance [12] and enhanced redistribution of serum IGFs to maternal tissues and the conceptus [17].

Previous studies have demonstrated that the IGFBP-3-degrading proteinase activity present in rat and human pregnancy serum is inhibited by chelating agents. Because the serum concentrations of MMP-1, MMP-2 and MMP-9 increase in maternal serum throughout pregnancy [18-20] and because their naturally occur- ring inhibitor, TIMP-1, decreases concurrently [21], we speculated that MMPs might be involved in the limited proteolysis of IGFBP-3 during pregnancy [3]. As demonstrated in Fig. 4, TIMP-1, when added to rat pregnancy serum (lane 3), inhibited [125I]rhlGFBP-3 proteolysis by >95%. In addition, purified human MMPs, principally MMP-1 and MMP-3, degraded IGFBP-3 in solution, and MMP-3, in particular, produced a pattern of IGFBP-3 degradation products identical in size to fragments produced by pregnancy serum. Furthermore, the combined addition of anti-human MMP-1 IgG and anti-human MMP-3 IgG blocked >95% of the degra- dation of [~25I]rhlGFBP-3 by rat pregnancy serum. Taken together, these data demonstrate that MMPs are involved in IGFBP-3 proteolysis in the serum of late gestation rats, and suggest that MMP-1 and MMP-3 are the principal MMPs

A B

In .4.J e-

l in 4.~

¢0 .,.4 f_ e-

q--

¢v-j :b . I

a . -,.-0

I t . e - GO 4) t.--* " O e -

¢0 ¢0 U

, -4 .4J t-i

GO O

~D

100

5O

o l l 1 2 3 1 2 3

F I G U R E 4. Inhibi t ion o f IGFBP-3-degrading protease activity in rat pregnancy scram by T I M P - 1 . (A)

[ I ~ I ] r h l G F B P - 3 was incubated with buffer (lane 1) or rat pregnancy serum without (lane 2) or with (lane 3) T I M P - I as described in Ref. Pl - (B) Three separate experiments were analyzed for the effect of T I M P -

1 to inhibit the degradation [t25IIrhlGFBP-3. Bars correlate with lanes 1, 2 and 3 as outlined above. * P < 0.0001.

MMPs as IGFBP-degrading Proteinases 261

involved. Some investigators have characterized IGFBP-3 proteinase activity as a cation-dependent, serine proteinase; our results are consistent with these findings, since serine proteinases such as chymotrypsin, trypsin, plasmin, plasma kallikrein and neutrophil elastase have all been shown to activate one or more proMMPs [8].

CONCLUSIONS

These studies demonstrate that MMPs function as IGFBP-degrading proteinases both in vivo and in vitro under physiologic conditions. The demonstration that MMPs degrade IGFBPs broadens the range of substrates described for these extra- cellular proteinases [6-8]. While the physiologic ramifications of IGFBP-degrada- tion by MMPs is not entirely clear, MMPs have been implicated in a variety of processes such as tumor cell invasion, morphogenesis, trophoblast invasion, carti- lage and bone repair and turnover, wound healing and angiogenesis [16]. Since IGFs also participate in many of these same processes, changes in MMP produc- tion and/or secretion may alter IGF bioavailability due to their regulation of IGFBP degradation. In vivo, such alterations are present in pregnancy (as described above); however, increased MMP production has been reported in several other disorders in which IGFBP-3 proteolysis has also been reported, including brain tumors [22, 23], breast cancer [2, 19] and prostate cancer [24-26].

It is presently unclear how MMPs effect IGF bioavailability at the cellular level; however, previous studies have demonstrated that cytokines, such as bFGF and TGF-fl~, bind to extracellular matrix, where they can be released via proteolysis to exert their effects at the cell surface. Since IGFBP-3 and -5 have been shown to associate with extracellular matrix [1], IGFs bound to matrix-associated IGFBPs may be released by MMPs to interact with cell surface receptors during periods of cellular proliferation. Our data also demonstrate that MMPs-1, -2 and -3 cleave IGFBP-3 exclusively within the mid-region of the molecule [4]; thus, another conse- quence of MMP-mediated IGFBP proteolysis may be to cleave IGFBPs into frag- ments which possess unique functions. In particular, N-terminal fragments of IGFBP-3 and IGFBP-5 have been shown to enhance IGF-action in bone cells [1]. Together, these studies suggest that an important interplay exists between MMPs and the IGF-IGFBP axis and that under the appropriate conditions, MMPs may serve to release stores of IGFs, allowing for a rapid cellular response to IGFs, which requires no new synthesis of these peptides.

REFERENCES

1. Jones JI, Clemmons DR. Insulin-like growth factors and their binding proteins: biological actions. Endocr Rev. 1995; 16: 3-34.

2. Oh Y, Muller HL, Pham H, Lamson G, Rosenfeld RG. Insulin-like growth factor binding protein (IGFBP)-3 levels in conditioned media of Hs578T human breast cancer cells are post-transcrip- tionally regulated. Growth Regul. 1993; 3: 84--87.

3. Fowlkes JL, Suzuki K, Nagase H, Thrailkill KM. Proteolysis of insulin-like growth factor binding protein-3 during rat pregnancy: a role for matrix metalloproteinases. Endocrinology 1994; 135: 2810-2813.

4. Fowlkes JL, Enghild J J, Suzuki K, Nagase H. Matrix metalloproteinases degrade insulin-like growth factor-binding protein-3 in dermal fibroblast cultures. J Biol Chem. 1994; 269: 25,742-25,746.

262 J. L. Fowlkes et al.

5. Thrailkill KM, Quarles LD, Nagase H, Suzuki K, Serra DM, Fowlkes JL. Characterization of insulin-like growth factor-binding protein 5-degrading proteases produced throughout murine osteoblast differentiation. Endocrinology 1995; 136: 3527-3533.

6. Alexander CM, Werb Z. Extracellular matrix degradation. In: Hay ED, ed. Cell biology of extra- cellular matrix, 2nd edn. New York: Plenum Press; 1991: 255-302.

7. Murphy G, Docherty AJP. The matrix metalloproteinases and their inhibitors. Am J Respir Cell Mol Biol. 1992; 7: 120-125.

8. Nagase H, Ogata Y, Suzuki K, Enghild J J, Salvesen G. Substrate specificities and activation of matrix metalloproteinases. Biochem Soc Trans. 1991; 19: 715-718.

9. Fowlkes JL. Degradation of insulin-like growth factor (IGF)-binding protein-3 (IGFBP-3) by a metal-dependent protease produced by human fibroblasts: effects of IGFs on protease activity. Endocrine J. 1994; 2: 63-68.

10. Rechler MM. Insulin-like growth factor binding proteins. Vitamin Horm. 1993; 47:1-114. 11. Thrailkill KM, Siddhanti SR, Fowlkes JL, Quarles LD. Differentiation of MC3T3-EI osteoblasts

is associated with temporal changes in the expression of IGF-I and IGFBPs. Bone 1995; 17: 307-313.

12. Davenport ML, Clemmons DR, Miles MV, Camacho-Hubner C, D'Ercole AJ, Underwood LE. Regulation of serum insulin-like growth factor-I (IGF-I) and IGF-binding proteins during rat preg- nancy. Endocrinology 1990; 127: 1278-1286.

13. Davenport ML, Pucilowska JB, Clemmons DR, Lundblad R, Spencer JA, Underwood LE. Tissue- specific expression of insulin-like growth factor binding protein-3 protease activity during rat preg- nancy. Endocrinology 1992; 130: 2505-2512.

14. Fielder PJ, Thordarson G, Talamantes F, Rosenfeld RG. Characterization of insulin-like growth factor binding proteins (IGFBPs) during gestation in mice: effects of hypophysectomy and an IGFBP-specific protease activity. Endocrinology 1990; 127: 2270-2280.

15. Giudice LC, Farrell EM, Pham H, Lamson G, Rosenfeld RG. Insulin-like growth factor binding proteins in maternal serum throughout gestation and in the puerperium: effects of a pregnancy- associated serum protease activity. J Clin Endocrinol Metab. 1990; 71: 806-816.

16. Lassarre C, Binoux M. Insulin-like growth factor binding protein-3 is functionally altered in preg- nancy plasma. Endocrinology 1994; 134: 1254-1262.

17. Bastian SEP, Walton PE, Wallace JC, Ballard FJ. Plasma clearance and tissue distribution of labelled insulin-like growth factor-I (IGF-I) and an analogue LR31GF-I in pregnant rats. J Endocrinol. 1993; 138: 327-336.

18. Morrison JM, Clark IM, Powell EK, Cawston TE, Hackett GA, Smith SK. Tissue collagenase-serum levels during pregnancy and parturition. Eur J Obstet Gynecol Reprod Biol. 1994; 54: 71-75.

19. Zucker S, Lysik RM, Zarrabi MH, Moll U. M r 92,000 type IV collagenase is increased in plasma of patients with colon cancer and breast cancer. Cancer Res. 1993; 53: 140-146.

20. Zucker S, Lysik RM, Gurfinkel M, Zarrabi MH, Stetler-Stevenson W, Liotta LA, Birkadal-Hansen H, Mann W. Immunoassay of type IV collagenase/gelatinase (MMP-2) in human plasma. J Immunol Methods 1992; 148: 189-198.

21. Clark IM, Morrison JJ, Hackett GA, Powell EK, Cawston TE, Smith SK. Tissue inhibitor of metalloproteinases - - serum levels during pregnancy and labor, term and preterm. Obstet Gynecol. 1994; 83: 532-537.

22. Apodaca G, Rutka JT, Bouhana K, Berens ME, Giblin JR, Rosenblum ML, McKerrow JH, Banda MJ. Expression of metalloproteinase inhibitors by fetal astrocytes and glioma cells. Cancer Res. 1990; 50: 2322-2329.

23. MUller HL, Oh Y, Gargosky SE, Lehrnbechter T, Hintz RL, Rosenfeld RG. Concentrations of insulin-like growth factor binding protein-3 (IGFBP-3), IGF, and IGFBP-3 protease activity in cerebral spinal fluid of children with leukemia, central nervous system tumor, or meningitis. J Clin Endocrinol Metab. 1993; 77:1113-1119.

24. Cohen P, Graves CBG, Peehl DM, Kamarei M, Guidice LC, Rosenfeld RG. Prostate specific antigen is an IGF binding protein-3 (IGFBP-3) protease found in seminal plasma. J Clin Endocrinol Metab. 1992; 75: 1046-1053.

25. Hamdy FC, Fadlon E J, Cottam D, Lawrey J, Thurrell W, Silcocks PB, Anderson JB, Williams JL, Rees RC. Matrix metalloproteinase 9 expression in primary human prostatic adenocarcinoma and benign prostatic hyperplasia. Br J Cancer 1994; 69: 177-182.

M M P s as IGFBP-degrading Proteinases 263

26. Kanety H, Madjar Y, Dagan Y, Levi J, Papa MZ, Pariente C, Goldwasser B, Karasik A. Serum insulin-like growth factor binding protein-2 (IGFBP-2) is increased and IGFBP-3 is decreased in patients with prostate cancer: correlation with serum prostate-specific antigen. J Clin Endocrinol Metab. 1993; 77: 229-233.