Vasoactive Peptides with Angiogenesis-Regulating Activity Predict Cancer Risk in Males

It is well established that the progression from malignant transformation into manifest tumor disease is dependent on nonmalignant cells of the host, perhaps most importantly through the recruitment of blood vessels, that is, the angiogenic switch (1, 2). The importance of factors derived from the intruding cancer cells in these events has been extensively studied; however, how the cardiovascular phenotype- and host-derived factors may regulate tumor development and long-term cancer risk remains ill-defined.

There is experimental evidence to support the concept that the vasoactive peptides adrenomedullin (ADM; refs. 3–9), vasopressin (10–14), and atrial natriuretic peptide (ANP; refs. 15–21) in addition to their effects on vascular tone and salt and water balance, may have direct effects on angiogenesis. In fact, there is a suggested link between angiogenesis and hypertension, that is, the loss of terminal arterioles and capillaries has been implicated in the pathogenesis of hypertension through increasing peripheral resistance (22). Hypertension is one of the most common side effects of antiangiogenic treatment of cancer patients (23), and hypoxia-induced angiogenesis was shown to attenuate vascular resistance and hypertension (24). Furthermore, patients with treatment refractory hypertension displayed elevated levels of VEGF (25). Angiogenesis thus emerges as an important regulatory mechanism in the control of blood pressure levels.

Here, we hypothesize that host levels of vasoactive peptides predict long-term cancer risk by regulating the transition of early stages of tumor development into clinically manifested cancer disease. To address this issue, we measured stable fragments of the precursors of these hormones [midregional pro-ADM (MR-proADM), midregional pro-ANP (MR-proANP), and C-terminal pre-provasopressin (copeptin)] in fasting plasma in a large population-based Swedish prospective cohort study and related baseline levels of these 3 biomarkers to cancer incidence during a median follow-up time of 15 years.

The Malmö Diet and Cancer (MDC) study is a population based, prospective epidemiologic cohort of 28,449 men (born 1923–1945) and women (born 1923–1950) from the city of Malmö in southern Sweden who underwent baseline examinations between 1991 and 1996 (26). From this cohort, 6,103 persons were randomly selected (1991–1994) to participate in the MDC Cardiovascular Cohort (MDC-CC), which was designed to investigate the epidemiology of carotid artery disease (27). Fasting plasma samples were available in a total of 5,543 subjects in the MDC-CC and 336 subjects had cancer prior to the baseline examination (84 males and 252 females). Subjects in the MDC-CC from whom fasting plasma was available, that were free from prior cancer and had data on the complete set of covariates included in model 2 (see Statistical analyses) were included in the data set analyzed in this study (4,061 individuals; 1,768 males and 2,293 females).

Blood pressure was measured with a mercury column sphygmomanometer after 10 minutes of rest in the supine position. Data on smoking, cancer heredity, and use of antihypertensive and antidiabetic medications were ascertained from a questionnaire. Heredity was defined as having at least 1 first-degree relative diagnosed with cancer. Current smoking was defined as any cigarette smoking within the past year. Diabetes mellitus was defined as having fasting whole blood glucose of more than 6.0 mmol/L, self-reported physician diagnosis of diabetes, or use of antidiabetic medications. Body mass index (BMI) was defined as the weight in kilograms divided by the square of the height in meters. Myocardial infarction prior to the baseline exam was defined and retrieved as described previously (28).

In fasted EDTA plasma specimens frozen immediately after collection at the MDC-CC baseline exam, we measured MR-proANP, MR-proADM, and copeptin with immunoluminometric sandwich assays as described previously (BRAHMS, AG, Germany) (29–32). N-terminal pro-B-type natriuretic peptide (N-BNP) was determined by the Dimension RxL automated N-BNP method (Siemens Diagnostics; ref. 33), and cystatin C was measured with a particle-enhanced immunonephelometric assay (N Latex Cystatin C; Dade Behring) (34). We measured fasting high-density lipoprotein (HDL) cholesterol, insulin, and triglycerides according to standard procedures at the Department of Clinical Chemistry, University Hospital Malmö, and low-density lipoprotein (LDL) cholesterol was calculated according to Friedewald's formula. All participants gave written informed consent, and the study was approved by the Ethical Committee at Lund University, Lund, Sweden.

Cancer events were defined as malignant tumors according to the Swedish Cancer Registry (SCR). We included cancer in situ of the breast in our definition, but not cancer in situ of the cervix. The rationale for this is that in situ cancer of the breast is clearly defined as a malignancy, which in many cases requires adjuvant therapy including radiation and/or systemic treatment, whereas in situ cancer of the cervix is more benign and most commonly only is treated by local excision without adjuvant therapy. Information on cancer events (both prevalent and incident events) was retrieved up until January 1, 2008 (end of follow-up) by record linkage with the SCR using a unique 10-digit civil registration number. Approximately, 99% of all tumors diagnosed at Swedish Hospitals are registered in the SCR and 98% are morphologically verified (35, 36).

Tumor site was registered according to ICD-7 and the ICD version used at diagnosis. Histopathologic type was coded according to the C24 classification (37).

We analyzed time to first event in relation to baseline biomarker levels using Cox proportional hazards models with age as time scale variable. We analyzed the 2 genders separately as the major cancer types are gender specific. The 3 biomarkers alone and in combination were always entered into the model together with current smoking, cancer heredity, cystatin C (as marker of glomerular filtration rate), and N-BNP (as marker of subclinical heart failure; model 1 covariates) unless otherwise specified in the text. The motive for adjustment for cystatin C was that all the 3 biomarkers are mainly cleared from plasma by glomerular filtration. The rationale for entering N-BNP into the model, as a sensitive marker for subclinical heart failure and left ventricular dysfunction, was to test and adjust for any potential relationship between heart failure and cancer. We, in addition adjusted for physical activity and alcohol consumption measured as described previously (38) as well as for cigarette pack-years (multiples of 20 cigarettes smoked per day × smoking duration) in a subset of the population who had complete data for these additional covariates (1,528 males and 758 young males).

Biomarkers with skewed distributions (MR-proANP, copeptin, and N-BNP) were logarithmically transformed before analysis, and the relationship between biomarker levels and incident cancer is expressed as HR per 1 SD increase in the respective biomarker and in quartile analyses as HR for each quartile with the lowest quartile defined as the referent (HR = 1.0) and as HR per quartile increase to obtain the P_trend over quartiles.

To get a summed effect estimate of the relationship between the 3 biomarkers and incident cancer, we summed the Z-scores for the 3 biomarkers weighted for their respective β-coefficient from the corresponding Cox proportional hazards model (MR-proADM, MR-proANP, and copeptin entered simultaneously together with model 1 covariates), and the weighted sum of the Z-scores was referred to as “biomarker score.”

All Cox proportional hazards models, which were significant after model 1 adjustment, were further adjusted for in model 2 covariates (apart from in analyses of subtypes of cancer) which included all model 1 covariates together with BMI, systolic and diastolic blood pressure, antihypertensive treatment, myocardial infarction prior to baseline, diabetes mellitus, LDL, HDL, and fasting insulin.

In the Cox proportional hazards models for analyses of subtypes of cancer, the numbers of events for each subtype differs from the overall distribution of first events (Table 2) as a first cancer subtype event was allowed to be preceded by another subtype of cancer without being censored. We allowed other subtypes of incident cancer than the one specifically analyzed to remain in the “control group.”

In all Cox proportional hazards models, subjects were censored at the time of event, death, emigration from Sweden, or at the end of follow-up. The proportionality of hazards assumption was confirmed by Schoenfeld global test.

Crude Kaplan–Meier curves of cumulative incidence (beginning at baseline) were created for comparison of the biomarker score quartiles in analysis of incident cancer.

All analyses were conducted with Stata software (version 8.0; Stata Corp), and throughout a 2-sided P < 0.05 was considered statistically significant.

The characteristics of the study population without cancer prior to the baseline exam are shown in Table 1. During the follow-up period [median (interquartile range) 14.6 (13.6–15.2) years in males and 14.8 (14.1–15.6) years in females], in males 366 first-cancer events occurred, 259 subjects were censored due to death, and 15 due to emigration from Sweden. In females, 368 first cancer events occurred, 193 subjects were censored due to death, and 14 due to emigration from Sweden. The complete spectrum of various subtypes of incident cancer events is shown in Table 2.

In all analyses, the proportionality of hazards assumption was met. In males, there was an independent relationship between MR-proANP and copeptin, respectively, and incident cancer and a borderline significant relationship between MR-proADM and incident cancer (Table 3). As shown in analyses of quartiles, the relationship with incident cancer seemed to be graded over the distribution of MR-proANP, copeptin, and MR-ADM (Table 3). In contrast, N-BNP had no relationship with incident cancer with an HR (95% CI) per SD increase in N-BNP of 1.01 (0.90–1.13; P = 0.931). When MR-proANP, copeptin, and MR-ADM were entered simultaneously in the model together with model 1 covariates and backward elimination with a retention P < 0.10 was applied, the 3 biomarkers were all retained and significantly related to future cancer development with per SD increase in biomarker level HR of 0.83 (0.75–0.93; P = 0.001) for MR-proANP, 1.14 (1.01–1.29; P = 0.028) for copeptin, and 1.12 (1.00–1.24; P = 0.042) for MR-proADM. To get a summed effect estimate of the relationship between the 3 biomarkers and incident cancer, the biomarker score was entered into the Cox proportional hazards model together with model 1 covariates. The per SD increase of the biomarker score HR for incident cancer was significant and the top versus bottom quartile of the biomarker score identified a near 2-fold difference in risk of cancer (Table 3 and Fig. 1). Additional adjustment in model 2 did not change these results (data not shown). In contrast, we found no evidence of an association between any of the 3 biomarkers or the biomarker score with incident cancer among females (Table 4).

None of the individual biomarkers were significantly related to any of the main cancer subtypes in males, suggesting that the biomarkers are related to general cancer susceptibility (Table 5). In line with this, the point estimates of the per SD HRs for the cancer subtype specific biomarker scores were all above 1.0 (between 1.14 and 1.19) and that of the most common form of male cancer, that is, prostate cancer, was borderline significant (P = 0.061). Excluding subjects who got other subtypes of cancer than the one specifically analyzed from the “control group” gave similar results (data not shown). Importantly, as a result of lower power in the analyses of subtypes of incident cancer, we cannot exclude the possibility that one or a few subtypes of cancer, rather than all subtypes together, may explain the biomarker association with general cancer susceptibility. We found no evidence of biomarker association with the main subtypes of cancer in females except for a borderline significant inverse relationship between MR-proADM and pulmonary cancer (P = 0.05; Supplementary Table S1).

Screening and any potential prevention of cancer is likely to be most meaningful in subjects with relatively long remaining lifetime expectancy. In addition, cardiovascular conditions affecting the levels of the studied biomarkers, such as heart failure and hypertension, as well as use of antihypertensive medications, increase steeply with age and may obscure any relationship between the biomarkers and cancer. We therefore studied the 3 biomarkers in relation to incident cancer separately in subjects below (“young”) and above (“old”) the median of age (57.9 and 57.7 years for males and females, respectively).

In young males (n = 884 subjects who developed n = 133 first incident cancer events during follow-up; 66 subjects were censored due to death and 9 due to emigration from Sweden), the HRs per SD biomarker increase were significant for all 3 biomarkers and were markedly higher than in the male population as a whole (Table 3). The HR of the biomarker score was near double as high as among all males. In quartile analyses, the risk of cancer was shown to increase gradually with the biomarker score and the difference in cancer risk between the top versus bottom quartile was more than 3-fold. Additional adjustment in model 2 did not change these results (data not shown). In older men (n = 884 subjects who developed n = 233 first incident cancer events during follow-up; 193 subjects were censored due to death and 6 due to emigration from Sweden), we found no significant associations between the biomarkers and incident cancer (Table 3). Thus, each of the 3 studied biomarkers, and in particular the combination of the 3, strongly and independently predicts cancer in the younger half of our male study population, a relationship that seems to drive the association seen in the entire group of males. Similar to the results of analyses of biomarker associations with the main subtypes of cancer in all males, the corresponding analyses in men below the median age did not reveal any specific subtype of cancer driving the biomarker association observed with general cancer susceptibility (Table 5), although it cannot be excluded that a few cancer subtypes, rather than all subtypes together, explains the overall association. In females, we did not detect any biomarker relationship with cancer whether the younger or older females were studied (Table 4).

To test if the relationship between the 3 biomarkers and incident cancer in males is linked to mechanisms preceding cancer development or whether it is actually driven by established cancer which is subclinical and not yet diagnosed, we excluded all cancer events diagnosed within the first 4 years of follow-up in analyses of all males (67 of 366 cases excluded) and young males (20 of 133 cases excluded). After these exclusions, the results were similar to the results obtained without excluding early events with HR per SD increase of 0.82 (0.71–0.95; P = 0.009) for MR-proANP, 1.16 (1.02–1.33; P = 0.023) for copeptin, 1.13 (0.99–1.29; P = 0.081) for MR-proADM and 1.28 (1.13–1.44; P < 0.001) for the biomarker score in all males. In young males, the corresponding HRs per SD were 0.71 (0.58–0.86; P = 0.001) for MR-proANP, 1.24 (1.01–1.54; P = 0.037) for copeptin, 1.19 (0.99–1.44; P = 0.069) for MR-proADM and 1.49 (1.24–1.79; P < 0.001) for the biomarker score. Thus, our results indicate that decreased levels of MR-proANP and increased levels of copeptin and MR-proADM precede cancer development rather than being markers of already existing but nondiagnosed cancer.

In a subset of the male population with complete data on smoking quantity, smoking duration, physical activity, and alcohol consumption (1,528 males and 758 young males), we substituted adjustment for current smoking with pack-years and additionally adjusted for physical activity and alcohol consumption on top of model 1. The biomarker score remained significant also after these adjustments with HR per SD increase of 1.22 (1.08–1.37; P = 0.001) in all males and 1.39 (1.15–1.69; P = 0.001) in young males.

Finally, we tested whether the biomarkers are related to risk of cancer mortality (Table 6). Among the 1768 males, 107 subjects died from cancer during follow-up. Whereas MR-proANP and copeptin were not significantly related to cancer mortality, 1 SD increase of MR-proADM was associated with a 1.25-fold increased risk of cancer mortality, and males in the top quartile versus the bottom quartile of MR-proADM had an almost doubled risk of cancer mortality. The biomarker score was also related to risk of cancer mortality in males, although that association was mainly accounted for by MR-proADM (Table 6). After model 2 adjustment, the relationship between MR-proADM and cancer mortality in males was borderline significant (P = 0.085), whereas the biomarker score remained significant (P = 0.031).

In the younger half of the male population, 33 subjects died in cancer during follow-up. One SD increase of MR-proADM was associated with a 1.55-fold increased risk of cancer mortality and the risk of dying in cancer was 2.5-fold higher in the top as compared with bottom quartile of MR-proADM, and 4-fold higher in the top as compared with bottom quartile of the biomarker score (Table 6), whereas MR-proANP and copeptin were not individually significantly related to cancer mortality (Table 6). After model 2 adjustment, MR-proADM (P = 0.003) and the biomarker score (P = 0.002) remained significantly related to cancer mortality.

In the older half of the male population and among females (all females, young females, and old females), there was no relationship between biomarker levels and cancer mortality (data not shown).

A growing body of experimental data indicates that vasoactive peptides may be directly involved in the pathogenesis of tumor disease through regulation of angiogenesis and malignant cell behavior. We have studied the relationship between circulating levels of vasoactive peptides classically involved in blood pressure regulation and the long-term risk of cancer. We present evidence that low MR-proANP and high MR-proADM and copeptin in plasma predict future cancer disease. The relationship between the 3 biomarkers and cancer incidence appeared dependent on gender as well as age, as it was specifically found in younger males of the study population. Given the age-related decline in androgens, it is interesting to note that multiple lines of evidence from recent in vivo studies suggest that androgens have potent proangiogenic activities exclusively in males via, for example, VEGF-related signaling and the mobilization of bone-marrow–derived endothelial progenitor cells (39, 40). Although the cause of the gender-specific associations seen in our study clearly needs to be elucidated in future studies, it may be speculated that angiogenesis-regulating activities of vasoactive peptides synergize with androgen-dependent pathways during tumor development. Another possible explanation for the age dependency is that the studied biomarkers are affected by various medications and cardiovascular conditions which are much more prevalent among older subjects, such as antihypertensive medication and subclinical cardiovascular diseases, and thus could obscure relationships with incident cancer.

ADM is known as a potent, proangiogenic factor that acts either directly on vascular cells or through secondary induction of, for example, VEGF (3–9). Accordingly, knockout of ADM or the ADM high-affinity receptor resulted in early embryonic lethality due to severe cardiovascular abnormalities (41, 42). Heterozygotic ADM knockout animals were viable and displayed reduced angiogenic potential as well as reduced tumor growth, providing genetic evidence that host levels of ADM regulate angiogenesis-dependent tumor development (43). In addition to its proangiogenic effects, ADM has been described as a hypoxia-induced factor with pleiotropic effects in malignant cells, indicating that ADM may promote tumor development and progression through dual effects on malignant cells and angiogenesis (6, 8, 44–46). Interestingly, pharmacologic inhibition of ADM as well as its receptors has been shown to attenuate tumor angiogenesis and tumor growth in experimental models of cancer, implying ADM vasoactive peptide as a potential target in cancer therapy (47, 48). Vasopressin has both direct and indirect effects on angiogenesis, for example, it was shown to stimulate protein synthesis in endothelial cells (12). In other studies, vasopressin was found to induce endothelin-1, that is, a vasoactive peptide known to stimulate tumor angiogenesis in endothelial cells (10, 11), to induce VEGF in vascular smooth muscle cells (13), and to increase capillary density in vivo (14). Several studies have documented an antiproliferative effect of ANP in endothelial cells as well as in different types of cancer cells (15–21). Indeed, it has been proposed that ANP counteracts tumorigenesis through feedback inhibition of its receptor, natriuretic peptide receptor A, that seems to drive tumor angiogenesis through VEGF induction (49). Moreover, endothelin-1- and VEGF-dependent angiogenesis were inhibited by ANP (15, 18, 19), suggesting that ANP has opposing effects on some of the angiogenic signaling pathways triggered by ADM and vasopressin.

Although there are examples of measurable risk factors that motivate primary preventive interventions, such as mastectomy for carriers of BRCA1 and BRCA2 mutations, and coloscopy in individuals at increased risk of colorectal cancer, biomarkers easily measurable in blood or urine that signal increased risk of cancer in general are lacking to date. Thus, one of the most important clinical implications of our findings is that they point at blood pressure regulatory elements as potential pharmacologic targets in primary prevention of cancer and that candidates for any such future therapy could be identified by measurement of MR-proANP, copeptin, and MR-proADM.

Although experimental studies in conjunction with our clinical data point at a link between vasoactive peptides and cancer, the potential relationship between hypertension and cancer remains obscure; we did not find any significant correlation between baseline blood pressure and incident cancer. It has been speculated that compensatory, proangiogenic mechanisms counteract predisposition for increased blood pressure, and a continuous angiogenic drive was recently implicated in treatment refractory hypertension (25). Variance of the ANP gene (NPPA) is associated with reduced ANP production and increased risk of hypertension (50). It may be speculated that such primary events in hypertension development elicit compensatory induction of, for example, ADM and vasopressin resulting in a proangiogenic phenotype and increased risk of cancer development. Notably, the cancer-associated biomarker pattern (low MR-proANP and high MR-proADM and copeptin) is typically seen in obesity (51–53), one of the major prehypertensive conditions. However, obesity alone cannot explain the relationship between the biomarkers and cancer as it was unchanged after adjustment for both BMI and obesity.

Our study has limitations. For example, we used the rather rough measure of “current smoking” as measure of smoking exposure. Substituting current smoking with a refined smoking exposure variable (never smoked/stopped smoking/smoke sometimes/smoke regularly) gave almost identical results as those reported in Table 3. When adjusting for pack-years as well as for physical activity and alcohol consumption on top of model 1, the biomarker score remained strongly and significantly associated with incident cancer, despite a reduction of the sample size due to missing data on some individuals for these covariates. However, there is still a possibility that our results are confounded by other exposure factors influencing cancer susceptibility, which we did not measure in the study. Furthermore, although our findings are promising, replication studies are warranted, not least of the subgroup analyses (below and above the median of age). Just as there is a need of replication of our positive findings in all men and young men, the negative findings in women as well as the findings in relation to subtypes of cancer may be a result of low power and thus do not exclude relationships with the studied biomarkers.

In conclusion, we here show that the vasoactive biomarkers MR-proANP, MR-proADM, and copeptin predict later development of cancer in males, particularly in younger males. Our findings may have implications for cancer risk prediction and present novel, potentially drug modifiable, mechanisms underlying cancer development.

Dr. Struck is employed by BRAHMS Biomarkers, Thermo Fisher Scientific, who holds patent rights on the use of MR-proANP, MR-proADM, and copeptin assays and their use in the prediction of cancer. No potential conflicts of interest were disclosed by the other authors.

This work was supported by The Swedish Cancer Society (to M. Belting), the Swedish Research Council (O. Melander and M. Belting); the Lund University Hospital donation funds (M. Belting); the Medical Faculty, Lund University (O. Melander and M. Belting); the Governmental funding of clinical research within the national health services (ALF; O. Melander and M. Belting), the Swedish Heart and Lung Foundation (O. Melander), Malmö University Hospital (O. Melander), the Albert Påhlsson Research Foundation (O. Melander), the Crafoord Foundation (O. Melander and M. Belting), the Gunnar Nilsson Cancer Foundation (M. Belting), Region Skåne (O. Melander), the King Gustaf V and Queen Victoria Foundation (O. Melander), the Marianne and Marcus Wallenberg Foundation (O. Melander), and the Knut and Alice Wallenberg Foundation (O. Melander).

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.