Since it was first claimed in 1995 that human chorionic gonadotropin (hCG) inhibits HIV-associated Kaposi’s sarcoma (KS), focus has been lost on establishing a credible hypothesis to explain a novel action ascribed to hCG, which is, after all, a pregnancy hormone. After 7 years, no real progress has been made, and there remains no consensus as to which component of hCG preparations is responsible for inducing apoptosis of HIV-associated KS. It would certainly seem apparent that many basic experiments on hCG species have been overlooked in the studies that have been published. Furthermore, the much wider literature concerning the association of hCG with promoting oncogenesis has been ignored. This review puts into context the numerous studies on hCG, HIV-associated KS, and oncogenesis.
hCG2 has recently undergone a roller coaster ride in research terms: it has gone from being presented as an active anti-HIV KS (1) agent to being a mere contaminant of the active compound. The manner in which these hCG/KS studies have been conducted was referred to as a “stab in the dark” approach in an article published in the Journal of the National Cancer Institute (2) . The author, Darzynkiewicz, likened the research to “blaming the butler” without really investigating a true hCG-related culprit thoroughly (2) . Although speculation on the actual function of hCG in induction of apoptosis in KS has only recently been suggested in the high impact factor journals, research into the role of hCG in the progression of common epithelial cancers has been appearing in the literature for more than 30 years. However, the reported actions are diametrically opposed: hCG causes cell death in KS, but growth in carcinomas. Recent research has given rise to a hypothesis by which these opposing effects can be united under a single biological action. The central pillar of this is the recently discovered structural homology of the hCG subunits to the cystine knot family of growth factors. This article reviews the structural forms of hCG produced by tumors (as opposed to the pregnancy hormone), the action of these molecules on HIV-associated KS, and their role in epithelial oncogenesis. We propose a unifying action by which the free β-subunit (hCGβ) may simultaneously stimulate apoptosis in KS and decrease apoptosis in epithelial tumors such as bladder carcinoma.
The placental product hCG is one of the four members of the glycoprotein hormone family. The other members of the group are thyroid-stimulating hormone, follicle stimulating hormone, and LH, all of which are pituitary products. Each hormone shares a common glycoprotein hormone α-subunit noncovalently linked to a functionally distinct β-subunit, thus forming four different intact heterodimeric hormones (3) . Individual genes at four separate loci code the subunits. Although each hormone performs a discrete function, the β-subunits are homologous; in particular, the amino acid sequences of the LHβ and hCGβ are highly conserved (82% in the first 115 amino acids; Ref. 4 ). The similarity between the structures of hCGβ and LHβ allows each hormone to act on the same receptor. Intact α-β heterodimers are required to bind and stimulate this receptor; although individual subunits have been shown to bind with low affinity, they have no biological activity (3) .
Although the intact hormone hCG is produced by the placenta and by germ-cell tumors, it is the free β-subunit (hCGβ) that is produced by epithelial tumors, independent of glycoprotein hormone α-subunit gene expression (5 , 6) . In both cases, the ultimate fate of the molecule is renal degradation to a urinary breakdown product called hCGβcf. The intact hormone undergoes a complex series of degradations leading to the dissociation of the α- and β-subunits. The α-subunit is predominantly excreted unaltered, whereas the β-subunit is further degraded in the circulation to “nicked” hCGβ and in the kidney to hCGβcf (7, 8, 9) .
HIV, KS, and hCG
KS usually presents as different-sized lesions consisting of aggregates of spindle-shaped cells interspersed with endothelium-lined channels, which remain confined to the s.c. layer of the skin. KS is the most common tumor in patients infected with HIV-1 (10) , and the risk of developing the sarcoma is increased 20,000-fold in AIDS sufferers (11) . The first mention of hCG anti-KS activity was noted as part of an investigation into the significance of the male gender bias of HIV-associated KS (1) . In fact, Kaposi observed a bias toward male presentation for this condition of 15:1 in his initial study of this sarcoma (12) . The ratio has subsequently dropped to 4:1 (13) . A research group at the National Institute of Health (Bethesda, MD; headed by Robert Gallo) noticed that mice with induced KS that became pregnant went into tumor remission. They went on to describe how preparations of hCG and hCGβ “killed KS cells in vitro and in vivo, apparently by apoptosis,” and attributed the effect to LH/hCG receptors that were identified on the cell surface (1) . This novel finding surprised oncologists and reproductive physiologists because it was contrary to commonly held views in the field of carcinoma research, which had established links between expression of hCG and its subunits and poor patient prognosis. At the time, adjuvant tumor therapy involving hCG vaccines was already under investigation (14 , 15) , and it was argued that the murine models used were unsuitable given that mice do not possess a CGβ gene and were therefore unlikely to respond to hCG (16) . Rabkin et al. (17) responded with a suggestion that a factor other than hCG must explain the tumor regression in pregnant mice. They noted in their study that nearly half (43%) of women newly diagnosed with KS had been exposed to elevated hCG within the past 2 years.
Phase 1 trials investigating hCG as an anti-KS agent soon followed, despite the initial counterarguments and concerns that were raised. In many cases these trials confirmed the initial findings described by Lunardi-Iskander et al. (1) . In the first study, 150,000–700,000 IU of hCG were administered i.m. three times a week and resulted in complete remission of all KS lesions (18 , 19) . A subsequent study reported on the administration of 250, 500, 1000, and 2000 IU of hCG by intralesional injection, resulting in a dose-dependent KS regression by induction of apoptosis (20) . In response to this, Krown (21) commented that the quantity of hCG required to bring about a response was not financially viable as a treatment option. He also noted the variation in the quality of hCG preparations: the free subunits of hCG and the urinary metabolite hCGβcf are common contaminants. In fact, he alluded to the presence of a contaminant being the active anti-KS component and went on to suggest that this might be hCGβcf. Credence to this hypothesis was given when others noted the structural similarity of hCGβcf to the CKGF, PDGF (22) .
Is It hCG or Its Free Subunits That Kill KS Cells?
The possible variations in hCG preparations was investigated by Lang et al. (23) in light of the concerns regarding purity. An in vitro model was studied in which cultured KS spindle cells were exposed to eight different hCG and hCG-subunit preparations. The surprising outcome was that only two of the commercially available hCG preparations induced KS cell death: preparation CG10 from Sigma and Steris Profasi from Serono, an infertility preparation. Of greater significance was that recombinant intact hCG did not induce cell death but individual recombinant preparations of the free subunits, hCGα and hCGβ, did bring about death of the cultured KS spindle cells. A more detailed examination of the mode of action for the CG10 hCG preparations indicated that KS spindle cell death was a result of induced apoptosis. The authors of this study went on to suggest that the observed action of the hCG subunits was probably mediated by a putative orphan receptor (23) .
Clinical Trials Continued Regardless.
Clinical trials continued into the effectiveness of hCG against KS, and a Phase II study was carried out on 18 human males, each with evident KS lesions. Six patients were treated at each hCG dosage level: 5,000 IU/day, 10,000 IU three times/week, and 10,000 IU/day. Complete remission of lesions occurred in 4 of the 12 in the lower dosage groups and all in 6 of the higher dosage group. No dose-dependent toxic effects were seen (24) .
The Hunt for the True Anti-KS Agent Continues.
The questions concerning purity and the actual active component of the hCG preparations still remained unanswered. Albini et al. (25) again commented on the great variability in purity among preparations and insisted that the active compound was a known contaminant, hCGβcf. Using preparations of hCG, its subunits, and its breakdown products, they showed that it was the hCGβcf that induced apoptosis in KS spindle cells and not the intact hormone hCG. They also demonstrated that it was KS spindle cells (and not associated endothelial cells) that were affected by the hCG-related molecule (25) . In further studies an anti-KS hCG preparation called APL (Wyeth-Ayerst) was fractionated on a Sephadex G-100 size-exclusion column. Each fraction was tested on KS cells in culture and for hCG steroidogenic activity. Intact hCG elutes early, close to the void volume (V0), but the study showed that the later fractions (which had no steroidogenic activity) reduced KS cell proliferation considerably (26) . However, it is clear from the data that these later fractions are at the total volume (Vt) of the chromatography pool and would contain concentrated quantities of low-molecular-weight molecules not resolved by the column matrix; possibly preservative and isolation chemicals used in the hCG extraction procedure. Significantly (and what the authors neglected to comment on), the fractions following the main steriodogenic peak of intact hCG (which had minimal steroidogenic activity) also brought about a decrease in KS cell proliferation. These fractions would have contained the subunits hCGβ and hCGα, along with breakdown products such as hCGβcf. Unsurprisingly, a HAF was next proposed as the functional component of the impure hCG preparations (27) . Size-exclusion chromatography was again used in the separation of an anti-KS hCG, but this time with Superdex G-200. The resulting fractionation had allegedly eliminated, based on size alone, the possibility of hCG, its subunits, and the major breakdown product hCGβcf being accountable for any anti-KS activity. Instead, two proteins, designated HAF-C and HAF-U, with molecular masses of 15–30 and 2–4 kDa, respectively, were identified as the active components (27) . The fact that Superdex G-200 is incapable of resolving proteins of less than 10 kDa and that 15–30 kDa would be the resolvable size of hCGβcf and/or hCGα remained unaddressed by the authors. Interestingly, a larger protein of ∼44 kDa also had anti-KS activity and was only commented on briefly in the discussion. On such a column matrix, hCGβ would resolve at an approximate mass of 44 kDa (28) . Despite such arguments, the HAF concept is possibly not completely without foundation. A protein that associates with hCGβcf has been isolated and identified as an anti-KS RNase 18 kDa in size (29) . Whether this protein is solely responsible for the activity described or whether it is acting in conjunction with hCG subunits is still unknown. Numerous urinary proteins associated with urinary hCGβcf have since been identified with antiviral activities (30 , 31) .
HCG and Anti-HIV Activity.
A direct antiviral activity of hCG has also been reported. In an earlier publication it was suggested that the high levels of circulating hCG prevented babies at risk from HIV from being infected in utero by specifically inhibiting viral replication in maternal blood (32) . Furthermore, the authors went on to suggest that hCG (0.01–1 IU) prevents HIV-1 transmission from lymphocytes to trophoblast cells during gestation (33) . Interestingly, when first-trimester placental cells were infected with HIV-1 in vitro, a 90% drop in hCG synthesis was observed (34) . Subsequently HIV-producing lymphocytes were exposed to 10–100 ng/ml pure hCGβ (no detectable hCGα), with a resulting inhibition of p24 gag protein synthesis (35) . In the same report it was suggested that it was the hCGβ in hCG preparations that was responsible for the anti-KS activity reported by others. Most recently, Robert Gallo’s group also reported a considerable decrease in HIV-1 expression in a murine KS-Y1 model (along with an inhibition of HIV infection rate of macrophages) in response to impure hCG and isolated HAF protein preparations (Table 1⇓ ; Refs. 27 , 30 , 31 , 37 ).
hCGβ and Non-Germ Cell Carcinomas
Ectopic production of free hCGβ been shown in cervical and endometrial carcinoma as well as many other non-germ cell tumors of the ovary, vulva, breast, prostate, lung, liver, kidney, colon, pancreas, and kidney (Table 2)⇓ . Ectopic expression by bladder carcinoma is well described and occurs in ∼35% of cases (64) . It is not uncommon for ectopic hCGβ production to be explained by dedifferentiation (trophoblastic differentiation), where it is assumed that the tissue has reverted back to pluripotence, i.e., taking on the characteristics of the syncytiotrophoblast and thus expressing hCGβ. However, in many cases the sole criterion for the claim of trophoblastic differentiation is the detection of hCGβ. Because common epithelial tumors will express hCGβ, most of these claims are the result of a false dogma. A clear distinction exists whereby germ cell tumors will express both hCGα and hCGβ, resulting in the production of the gonadotropic intact hormone hCG, but ectopic expression by common epithelial tumors is almost exclusively of the free β-subunit. Only rarely is the intact hormone found in advanced-stage carcinomatosis (65) . Such dedifferentiation is thus a much rarer phenomenon than is often claimed.
The ectopic expression of hCGβ by bladder cancers has been reported extensively and serves as a model for such ectopic expression by common epithelial cancers. Significantly, there is a correlation of hCGβ expression by such tumors with grade, stage, and most importantly, prognosis.
Correlation with Tumor Grade and Stage of Disease.
Even at the most optimistic, the overall incidence of detectable hCGβ expression by bladder tumors is <50% and is thus too low for diagnostic screening purposes. However, there is a clear association between the incidence of hCGβ expression and the stage of disease. When the literature was examined closely, it was clear that in all cases where elevated serum levels of hCGβ were found in association with bladder cancer, the patients had metastatic disease (66) . Detailed examination of immunohistochemical studies also revealed that nearly all hCGβ-positive tumors were grade 3 and that most nonmetastatic tumors were invasive (67) .
Correlation with Prognosis.
Many authors have commented on the association between hCGβ and the aggressive nature of tumors. Martin et al. (68) compared the response rates to radiotherapy between hCGβ-positive (n = 29) and hCGβ-negative tumors (n = 71). A statistically significant lower response rate was seen in the hCGβ-positive tumor group (24% versus 59%; P < 0.0005). Moutzouris et al. (69) also found a statistically significant association between failure to respond to radiotherapy and hCGβ expression. Furthermore, the authors of a survival analysis noted that patients with tumors that do not express hCGβ survived longer; Marcillac et al. (70) reported that serial measurement of hCGβ levels in bladder cancer patients predicted recurrence and relapse before clinical changes. Dobrowolski et al. (71) advocated serial hCGβ measurement to predict the superficial or invasive nature of the disease.
In a prospective survival study, we reported a significant association with subsequent development of metastases (P < 0.01) for urinary hCGβ-positive patients with invasive T2–T4 disease. Furthermore, survival analyses showed a strong association between hCGβ expression and early death (P < 0.001; Ref. 42 ).
Biological Action of hCGβ on Epithelial Tumors
The reason that hCGβ expression by bladder cancers confers a tendency for the tumor to resist radiotherapy and to metastasize is unknown. Expression of fetal proteins in cancers is a well-recognized phenomenon, and an established understanding now exists that cancer is a form of cellular regression. Genes that promote the function of growth factors prenatally are again switched on as oncogenes, and the fetal protein is translated in adult human cells. The hCGβ/hLHβ gene cluster responsible for the production of hCGβ has been shown not to be amplified or rearranged in bladder tumor cells, indicating that this ectopic expression is more likely to be the result of gene regulation being altered in some way (72) . It was therefore suggested that hCGβ must have some form of biological role, acting on the cells from which it is secreted. Furthermore, given that the LH/hCG receptor is not expressed by these tissues, any activity observed must be occurring via an as yet unidentified pathway (73) . On the basis of this supposition, bladder tumor cell lines known to be secretors of hCGβ (65) were subjected to stimulation by hCGβ in vitro. It was shown that cell numbers increased after incubation with hCGβ in a dose-dependent manner. No effect whatsoever could be seen after treatment with intact hCG, hCGα, or hCGβcf, and a dose-dependent inhibition by anti-hCGβ antiserum was also observed. Furthermore, addition of these antibodies to the culture medium of bladder cancer cell lines inhibited the growth of the cell lines, which produced endogenous free hCGβ. These same antibodies did not affect the growth of bladder cell lines that did not produce free hCGβ (74) . It was also noted that the cell lines least affected by the β-subunit were those that secreted the higher concentrations of the same molecule, suggesting autocrine stimulation. These findings strongly suggested that free hCGβ was acting as a growth-stimulating factor. However, recent studies have now shown that the increase in tumor cell population in response to hCGβ was not as the result of an increase in cell replication, but rather was attributable to a reduction in cell death/apoptosis within the culture (75) . This growth factor-like behavior was more pertinent given the structural similarities between hCGβ, NGF, TGFβ, and PDGFB (22) , where the common structures of the family suggest a common function, i.e. growth regulation.
Structural Homology of hCGβ with the CKGFs
In 1994, Lapthorn et al. (22) successfully desialated and crystallized hCG and, from subsequent electron density maps, determined its three-dimensional structure. The most striking feature was the arrangement of three disulphide bridges in the center of each subunit. The positions of the three cystines is almost identical in both subunits, where two disulphides (linking residues 34–88 and 38–90 on the β-subunit and 28–82 and 32–84 on the α-subunit) bridge the antiparallel strands of the peptide chain, forming a central loop, through which the third disulphide passes (linking residues 9–57 on the β-subunit and 10–60 on the α-subunit). This structure has been identified before in a group of growth factors, which are designated by its name, the cystine knot. The CKGF family includes TGFβ, PDGFB, and NGF, but despite sharing structure similarities, they carry out quite distinct functions. The TGFβ family is a ubiquitous family of proteins that includes the inhibins and the activins. TGFβ-1 and -2 have been the most described and are multifunctional growth factors with both stimulatory and inhibitory cellular activity. These opposing actions are largely dependent on the embryonic origin of the target tissue. PDGFB is an autocrine and paracrine mitogenic stimulator of mesenchymal and glial cells, whereas NGF is a potent apoptotic inhibitor of both central and peripheral nervous system neuronal cells [reviewed by Sun and Davies (76)] . When Lapthorn et al. (22) grouped the glycoprotein hormones into this family, they suggested that their similar structure might indicate similar biological functions. Their attention was particularly drawn to the striking similarity between PDGFB and hCGβcf. The inclusion of hCG (and indeed all of the glycoprotein hormones) in the CKGF family has recently been reinforced by reports demonstrating that individual subunits of hCG and LH formed homodimers in a similar manner to TGFβ, PDGFB, and NGF (28 , 77 , 78) .
A Unifying Action: hCGβ and Cross-Talk with TGFβ
In both HIV-associated KS and epithelial bladder cancer, it appears that hCGβ—and possibly its urinary metabolite hCGβcf—alter the growth of the tumors cells by reducing cell numbers in KS but by increasing the cell population in bladder carcinomas. In the absence of a receptor for hCGβ/hCGβcf, structural homology with the CKGFs suggests that cross-stimulation of the CKGF receptors is occurring, but via what receptor? In the case of hCGβcf, the topological homology with PDGF has long been known. However, it is interesting to note that the growth modulation action occurs via apoptosis: induction of apoptosis in the case of KS, and inhibition of apoptosis in the case of bladder carcinomas. Only TGFβ has been shown to have bifunctional growth stimulatory and inhibitory actions on target tissues, presumably by modulating apoptosis. We have previously highlighted the topological homology of hCGβ with TGFβ (22 , 74) . Furthermore, the fact that TGFβ is coded for by a locus adjacent to that of the hCGβ-hLHβ gene cluster (79) and that high levels of TGF-β and its receptor are expressed by urothelial tumors despite their growth-inhibitory effects (80, 81, 82) appear to be far from coincidental.
It was suggested that the effects observed on Kaposi’s lesions were brought about by hCGβ presumably binding to a specific receptor and inducing apoptosis (1 , 37) . A similar system is probably present in bladder carcinoma, but the metabolic switch in the urothelium may inhibit (and not promote) apoptosis. All of this would appear to fit if hCGβ were to act in an opposing manner to its structural counterpart TGFβ, a well-established bifunctional growth factor responsible for both mesothelial tissue growth and induction of epithelial apoptosis. We have recently shown that, like TGFβ, the free β-subunit of hCG forms homodimers; this molecule may antagonistically interact with the TGFβ receptor complex, preventing apoptosis in carcinomas but conversely preventing TGFβ-induced growth stimulation in tumors of mesothelial origin such as KS If this proves to be the case, it would explain the phenotypic links between hCGβ expression, bladder cancer resistance to radiotherapy, metastasis, and patient mortality as well as KS apoptosis in response to hCGβ treatment.
As mentioned in the “Introduction,” many basic experiments have yet to be conducted in this field. For example, an important question to address is whether the cell lines established from KS lesions secrete hCGβ themselves. This is critical because the KS lesions consist of spindle cells of mesothelial origin interspersed with endothelial and epithelial cell structures. Frequently the epithelial components overgrow the spindle cells, as in the case of the cell line KS-SLK (83) . It is therefore necessary to closely scrutinize which cell lines were used in KS in vitro model experiments. In a recent study on the levels of hCGβ expressed by >70 cell lines, we found that KS-SLK cells express 14 ng/ml hCGβ per 107 cells, a level comparable with the highest level of expression by bladder carcinoma cell lines (Table 3⇓ ; Ref. 84 ). Many of the cell death studies have been conducted on KS-Y1 cells (Kaposi spindle cells), and although KS-SLK are mentioned, to date no results have been reported for this cell line. What cell type actually represents the true sarcoma in vitro (and its embryonic origin) is of critical importance in light of the literature reviewed here.
It is yet to be determined whether these two antithetical responses of hCGβ are functioning via the same pathways and receptors. From reviewing the research literature on hCGβ and epithelial oncogenesis, it would appear likely that hCG, hCGβ, and hCGβcf form complex interactions with other CKGFs and/or their receptors to bring about changes in cellular responses, as seen in the partnership between PDGF and TGFβ. Rather than a “stab in the dark,” this reasoned approach takes into account the diverse research literature concerning the biology of the entire CKGF family and its relationship to cellular control. Although yet to be proven, it certainly forms the basis of an elegant working hypothesis whereby hCGβ and hCGβcf are mimicking other CKGFs, and we may be observing a cell morphology-/embryology-dependent, bifunctional blockade of the TGFβ receptor.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
↵1 To whom requests for reprints should be addressed, at Williamson Laboratory for Molecular Oncology, Department of Obstetrics and Gynecology, St. Bartholomew’s and the Royal London School of Medicine and Dentistry, St. Bartholomew’s Hospital, West Smithfield, London EC1A 7BE, United Kingdom. Phone: 44-20-7601-8951; Fax: 44-20-7601-7050; E-mail:
↵2 The abbreviations used are: hCG, human chorionic gonadotropin; KS, Kaposi’s sarcoma; hCGβ and -α, human chorionic gonadotropin β- and α-subunits, respectively; LH, luteinizing hormone; LHβ, luteinizing hormone β-subunit; hCGβcf, hCGβ core fragment; CKGF, cystine knot growth factor; PDGF, platelet-derived growth factor; HAF, human chorionic gonadotropin-associated factor; NGF, nerve growth factor; TGFβ, transforming growth factor β.
- Received July 16, 2002.
- Revision received May 7, 2003.
- Accepted May 7, 2003.