High throughput mRNA profiling highlights associations between myocardial infarction and aberrant expression of inflammatory molecules in blood cells

Inflammation plays a key role in the pathophysiology of atherosclerosis and in the development of acute coronary events.¹  Activated leucocytes, cytokines, and chemokines are prominent features of an atherosclerotic plaque. Moreover, plasma levels of markers of inflammation such as cell adhesion molecules, cytokines, proatherogenic enzymes, and C-reactive protein (CRP) were found to predict cardiovascular events in a variety of clinical settings.²  However surrogate markers like CRP are far removed from the actual disease process since they reflect how the liver reacts to disease in the vasculature. Therefore, the nature of a chronic systemic inflammatory state and the role of circulating leucocytes in maintaining such a state remain unclear.

The inflammatory state of leucocytes may be the byproduct of the local inflammation in the vessel wall, or it may reflect an active or latent infection that is in part responsible for the atherosclerotic process. This is supported by observations of increased neopterin and procalcitonin levels in patients with cardiovascular disease.^3-6  Whatever the mechanism, some inflammatory mediators may directly influence the atherosclerotic process in several ways. Interleukin 6, for example, lowers high density lipoprotein and alters lipoprotein metabolism,⁷  and CRP may facilitate low-density lipoprotein uptake by macrophages.⁸ 

Plasma protein levels do not fully reflect the inflammatory signature of leucocytes in whole blood. Tissue leucocytes and endothelial cells may also contribute to the plasma levels of inflammatory markers, and many inflammatory mediators are also produced by other cell-types. To overcome these obstacles in assessing the inflammatory status of circulating leucocytes, we have developed a sensitive quantitative assay that is capable of measuring a panel of mRNA levels in large series of whole blood samples in a single reaction. The panel was composed of target genes encoding cytokines, chemokines, their receptors (as representatives of the soluble mediators of the inflammatory response), genes encoding nuclear factor κB (NFκB) pathway components (as representatives of the main intracellular signal transduction route of inflammation), tissue factor (as the inducible component of the clotting system), and genes encoding several intracellular components involved in the link between NFκB and apoptosis, a link that is considered important for the survival of immune cells. We have applied this novel high throughput technology in a large population-based case-control study on myocardial infarction to assess whether inflammatory mRNA in circulating cells is increased in patients with myocardial infarction compared to control subjects.

Patients were men consecutively diagnosed with a first myocardial infarction before the age of 70 years between January 1990 and January 1996. Two of the following 3 characteristics had to be identifiable in the discharge record or hospital charts to confirm acute myocardial infarction: typical chest pain, electrocardiographical changes indicative of evolving myocardial infarction, or a transient rise in cardiac enzymes to more than twice the normal upper limit. Control subjects were men without a history of myocardial infarction who had a minor orthopaedic intervention between January 1990 and May 1996 and who had received prophylactic anticoagulation treatment after this intervention. Control subjects were identified via the Leiden Anticoagulation Clinic, which serves the same region as the hospitals where the patients were recruited and which was responsible for monitoring prophylactic anticoagulation for several weeks or months after the surgery. They had not used anticoagulants for at least 6 months prior to inclusion and control subjects were frequency matched to the patients on 10-year age groups. To ensure that inflammatory reactions surrounding the cardiac event or orthopaedic intervention had subsided, subjects were included in this study at least 6 months after the date of the event, or index date. Median time between index date and blood collection was 2.8 years (range, 0.6 years to 6.3 years) for cases and control subjects alike. Details of the population-based case-control Study of Myocardial Infarctions Leiden (SMILE), which included 1206 men, are described elsewhere.⁹ 

Medication use and history of diabetes prior to the index date were ascertained by interview with control subjects and retrieved from discharge letters for patients. At time of the blood draw they were assessed by using a structured interview. A person was classified as hypertensive or hypercholesterolemic when he was prescribed specific medications for these conditions. The study was approved by the review committee of the Leiden University Medical Center and the subjects gave written informed consent in accordance with institutional guidelines.

Morning fasting–citrated blood samples were drawn from the antecubital vein. Immediately thereafter, aliquots of 100 μL were added to 900 μL lysis buffer.¹⁰  The samples were stored at –70°C until further use. RNA was isolated using a silica-based method¹⁰  and analyzed by multiplex ligation–dependent probe amplification (MLPA) using a kit developed in collaboration with MRC-Holland (Amsterdam, The Netherlands) for the simultaneous detection of 38 messenger RNA molecules.¹¹  This MLPA profiling method is insensitive to the total amount of mRNA that is included in the reaction; therefore, the profile is independent of the total white blood cell (WBC) count. All samples were tested with the same batch of reagents, and a negative and lipopolysaccharide-stimulated control sample were included on each plate. The final polymerase chain reaction (PCR) fragments amplified with carboxyfluorescein-labeled primers were separated by capillary electrophoresis on a 16-capillary ABI-Prism 3100 Genetic Analyzer (Applied Biosystems, Nieuwerkerk aan de IJssel, The Netherlands). Peak area and height were processed using GeneScan analysis software (Applied Biosystems). The levels of mRNA for each gene were expressed as a normalized ratio of the peak area divided by the peak area of a control gene, resulting in the relative abundance of mRNAs of the genes of interest. Our probe set is listed in Table 1 and contains probes for mRNAs of 35 inflammation- and apoptosis-related proteins and 3 control genes, B2M, CDKN1A, and PARN. Expression levels of β2 microglobulin (B2M) were above the upper limit of detection in all samples, whereas CDKN1A and PARN mRNA levels were detectable in all 1152 samples and in 1135 samples, respectively. Therefore, areas were normalized to PARN and to CDKN1A resulting in similar expression profiles. We have opted to present the results relative to CDKN1A because of its insensitivity to in vitro lipopolysaccharide stimulation of an inflammatory response¹¹  (C.A.S., unpublished data, November, 2002).

Because this mRNA profiling technique was not available when the SMILE study was initiated and the samples were collected, this study was not preceded by a sample size calculation.

The relative mRNA levels of cases and control subjects were compared using a Mann-Whitney test. Persons with values above the upper detection limit were assumed to have the highest levels, whereas persons with values below the detection limit were assumed to have the lowest mRNA level for that marker. Odds ratios (ORs) were calculated as an estimate of the relative risk for myocardial infarction. Quartiles were defined on the basis of the mRNA distribution among control subjects. The lowest quartile was used as a reference category for calculating odds ratios. 95% confidence levels (CI95) were calculated according to the method by Woolf¹²  or were derived from the standard errors calculated by the logistic model. For variables with more than a quarter of the readings below the lower limit of detection, odds ratios were calculated for persons with detectable levels compared with those with nondetectable levels. About three-fourths of the samples had readings for PTP4A2 above the upper detection limit; therefore, the relative risk of myocardial infarction was estimated by calculating the odds ratio for persons with levels above the upper detection limit compared with those within the detection limits. Odds ratios were adjusted for traditional cardiovascular risk factors such as age, smoking, hypertension, hypercholesterolemia, diabetes, body mass index (BMI), alcohol use, and quartile of CRP by using unconditional logistic regression. The odds ratios were further adjusted for time between the index date and blood sampling. Adjusted odds ratios were also calculated limited to samples drawn at least 2 years after the index date. Because 185 patients stopped smoking in the interval between the time of their myocardial infarction and blood withdrawal, and smoking may influence some of the inflammatory markers, odds ratios were also calculated restricted to men who were nonsmokers both before and after the index date. Odds ratios were also adjusted for use of medication at time of blood sampling. Separate restriction analyses to include only men who were not on lipid-lowering therapy, antihypertensives, and aspirin at time of myocardial infarction and at time of blood sampling were also performed.

It was possible to isolate RNA and perform MLPA analysis on 524 cases and 628 control subjects that fit the criteria required for this study. Mean age was 56.4 years (range, 32.3-70.0 years) and 57.4 years (range, 27.2-74.8 years) for patients and controls, respectively. Out of the patients, 62% were smokers compared with 33% of control subjects (Table 2).

Men with a history of myocardial infarction had higher median leukocyte levels (normalized to CDKN1A) of most inflammatory mRNAs compared with control subjects (Table 1). Analysis based on mRNA levels after stratification in quartiles resulted in elevated odds ratios (Table 3) that were not influenced much by adjustment for traditional risk factors and plasma levels of CRP. The adjusted odds ratios increased 2- to 3-fold for the highest versus lowest quartiles of MIF, PI9, CCL3, PTPN1, and BMI1. PDE4B, GSTP1, NFKB1, IL12A, PDGFB, MYC, IL1B, IL8, NFKB1A, and IL15(1) had odds ratios between 1.5 and 1.9 after adjustment, while LTA, IFNG, IL1RN, TNFRSF1A, IL18, and CCL4 had adjusted odds ratios of 1.3 or 1.4. THBS1 and PARN were not associated with myocardial infarction (adjusted OR [OR_adj] 0.9 and 1.0, respectively). The full data in quartiles are shown in Table 4 for those markers that show increasing odds ratios with increasing levels of mRNA.

Odds ratios for markers with more than a quarter of the readings below detection levels are shown in Table 5. In this comparison of detectable versus undetectable levels, IL15(2), MCP-2, and IL2 were associated with an approximately 2-fold increase in odds ratio. IL12B had an adjusted odds ratio of 1.6 and IL6, TNF, IL4(2), and IL10 had a somewhat lower adjusted odds ratio of 1.4. Too few samples had detectable levels of IL13 and TF mRNA (9 and 15 samples, respectively). MCP-1 and NFKB2 had an odds ratio of 1.2 (CI95, 0.9-1.6 for both) and IL4 was not associated with myocardial infarction. IL1A was peculiar because it gave an odds ratio below 1, indicating a protective effect (OR, 0.6; CI95, 0.5-0.8). Most of the readings for PTP4A2 were above the upper detection limit. The OR_adj for persons with levels above the upper detection limit compared with those with levels within the detection limits was 1.7 (CI95, 1.2-2.3).

Few of the crude odds ratios were changed by adjustment for cardiovascular risk factors, and only 7 mRNA markers (LTA, MIF, IL1B, IL1RN, MYC, NFKB1, and CCL4) showed differences between the crude and adjusted odds ratios greater than 0.2. Multivariate analysis indicated that smoking accounted for most of these differences (Tables 3 and 5). A stratified analysis showed that the inflammatory markers have elevated odds ratios even in nonsmokers (Tables 3 and 5), and, in most cases, the odds ratios were higher in men who did not smoke on and after the index date even after adjustment for traditional risk factors. This was particularly striking for the highest compared with the lowest quartile of mRNA levels of MIF with an age-adjusted odds ratio of 5.2 (95CI, 2.9-9.5) in nonsmokers.

Analysis excluding persons with levels of CRP above 10 mg/L did not materially change the odds ratios (data not shown). Adjustment for use of medication at time of blood sampling in the logistic regression model had little effect on the odds ratios. Odds ratios remained elevated even after restriction to subjects who never used aspirin or medications for hypercholesterolemia or hypertension. Adjustment for traditional risk factors had little effect on these odds ratios. Time since the index date did not change the adjusted odds ratios for the highest quartiles, except for those of MIF, PI9, and PTPN1, which increased further to 3.5 (95CI, 2.3-5.2), 2.9 (95CI, 2.0-4.2), and 2.5 (95CI, 1.7-3.7), respectively (Tables 3 and 5). The odds ratios of these markers were elevated even when cases who had a myocardial infarction less than 2 years before sample collection were excluded from the analysis (MIF, 2.0 [1.2-3.1]; PI9, 2.0 [1.3-3.0]; and PTPN1, 2.4 [1.5-3.8]). Besides the odds ratios of MIF and PI9, in this analysis, only the odds ratios of MYC, IL8, and IL2 decrease somewhat (Tables 3 and 5).

Patients with a history of myocardial infarction have a different mRNA signature for inflammatory markers than healthy control subjects, with higher expression of circulating inflammatory RNA. More specifically, increased mRNA levels of MIF, PI9, CCL3, PTPN1, BMI1, and IL15(2) were found to be associated with myocardial infarction, with odds ratios in the range of 2 to 3 for patients compared with control subjects. Elevated mRNA levels of PDGFB, IL12A, IL12B, MYC, NFKB1, GSTP1, IL18, IL15(2), IL1B, IL1RN, IL8, NFKBIA, PDE4B, MCP-2, and IL2 all yielded odds ratios for myocardial infarction varying between 1.5 and 2.

The highest odds ratio was observed for patients with high mRNA levels of macrophage migration inhibitory factor (MIF). Recently, this protein has been associated with an array of autoimmune and inflammatory diseases including severe sepsis,¹³  arthritis,¹⁴  bronchial asthma,¹⁵  and acute respiratory distress syndrome.^16,17 MIF counteracts the immunosuppressive effects of glucocorticoids,¹⁸  and prolongs the inflammatory response by inhibiting apoptosis of macrophages.¹⁹  Its plasma levels increase up to 5-fold in the acute phase of myocardial infarction, but decrease to levels similar to those in healthy persons within 3 weeks. The rapid increase is thought to be due to the production or release of MIF by necrotic tissue in the heart, cellular infiltrate at local sites, and peripheral blood mononuclear cells.^20,21  These studies were performed on a small number of patients and controls, and aimed at the inflammatory response during and shortly after the myocardial event. The major source of MIF may differ in the acute and subacute stages of myocardial infarction.²¹  Our study shows that expression of MIF in circulating cells is higher in men with myocardial infarction even in the stable state long after the event.

PI9 was also prominently associated with myocardial infarction. This protein protects cells from apoptosis due to granzyme B released by cytotoxic lymphocytes to kill abnormal cells.²²  It also inhibits the conversion of the inactive precursors of IL1B and IL18 into the active forms. Its involvement in atherosclerosis has been suggested by its altered expression in atherosclerotic lesions.²³  The present finding of increased expression in the circulation of men with a history of myocardial infarction adds evidence to this and suggests that it may exert an influence outside the plaque itself.

The absence of a noticeable effect of adjustment or of restrictions to people on particular groups of medication shows that these associations are not brought about by traditional risk factors or by medication use. The only cardiovascular risk factor that had an effect on some of the odds ratios (LTA, MIF, IL1B, IL1RN, MYC, NFKB1, and CCL4) was smoking. This finding is not surprising since it is generally accepted that smoking enhances vascular inflammation, and some systemic inflammatory markers, including CRP, are higher in smokers than former or never smokers.^24,25  The results of the stratified analysis on nonsmokers show that the inflammatory markers have an effect themselves, and that smoking masked some of this effect.

Adjustment for levels of CRP did not change the odds ratios. This underlines that in cardiovascular disease inflammation plays a role in 3 compartments. The first is in the inflamed vascular wall where many inflammatory cells are known to accumulate. The second is in the liver where acute phase reactants such as CRP are synthesized, most likely in response to cytokines that are produced elsewhere in the body. The results from our study indicate that there is a third inflammatory compartment in circulating leucocytes. Apparently, the relationship between events in leucocytes on the one hand and acute-phase protein production in the liver on the other, if there is any, is not a simple one.

In quantitative or semiquantitative mRNA profiling methods accuracy is generally poor. As described in our original methods paper reproducibility of the MLPA using independent duplicate samples was satisfactory: interassay correlation between 3 representative data sets of independent samples was 0.96 and intra-assay variation between 4 independent samples was 0.97.¹¹  With respect to the reproducibility of the presented profiling results we would like to add the following. Based on the repeat measurements of the lipopolysaccharide samples in the present study we find coefficients of variation for individual mRNA species of around 0.25. Second, we recently measured a second mRNA profile, centered on the expression of members of the Toll-like receptor family of proteins, in the SMILE samples (S.B.W., P.R.H., unpublished results, April 2004). PI9, MIF, and PARN were also included in the second study, and we were able to replicate the findings reported here.

Our study evaluated mRNA-based profiles in a case-control design. The inflammatory profile that we have observed may reflect a chronic inflammatory state in circulating blood cells that is predictive for myocardial infarction. Alternatively, it may be a result of the myocardial infarction itself either directly, since several of the mRNA markers that we assessed may also increase during the acute phase of a myocardial infarction (TNFA,²⁶ IL6²⁷ ; IL1RN and IL10²⁸ ; MIF^20,21 ), or indirectly through decreased left ventricular function in myocardial infarction patients,²⁹  a variable that was not assessed in the SMILE study. Support for the hypothesis that we may indeed be dealing with causative risk factors comes from the observation that the increased odds ratios were observed in blood samples drawn at least 6 months (median, 2.8 years) after the index date, and from the observation that time since the event did not decrease the odds ratios. In any case, an inflammatory state of cells in the circulation may well have implications on the outcome and progress of inflammation-related diseases.

After the human genome effort, gene expression profiling is rapidly developing into a powerful tool. It has shown promising results in identifying patterns of aberrant expression in cancer patients, and in determining subtypes of leukemias.³⁰  These studies have been based on microarray techniques that, due to their cost, have limited analysis to typically tens of samples. To study more complex diseases a substantially larger number of samples need to be analyzed to get reproducible and meaningful results. The present RNA profiling study utilizes more than 1000 samples to study myocardial infarction, the largest number of samples for such a study to date. This greatly reduces noise and effects of factors not directly related to the disease and it also decreases the chances of false positives, which is a major problem in microarray studies.

In conclusion, this study presents the direct measurement of molecular signatures from more than 30 inflammatory genes in almost 1200 individuals. These large-scale mRNA measurements give direct insight into the inflammatory status of circulating leucocytes without the limitations of the more common measurements of CRP and cytokine/chemokine levels in blood or other smaller RNA studies. Leucocytes in the blood of patients with myocardial infarction are more active in transcription of inflammatory genes. Our results therefore represent a further step in demonstrating that inflammation in activated leukocytes is a hallmark in the etiology of cardiovascular disease.

S.B.W. and C.A.S. were responsible for setting up the RNA assays. S.B.W. performed all laboratory analyses. F.R.R. and C.J.M.D. were responsible for setting up the SMILE study and for collecting all patient data and blood samples. S.B.W. performed all statistical data analyses, and F.R.R. and C.J.M.D. oversaw these. P.H.R. and F.R.R. initiated the study and carry overall responsibility for the laboratory work and for the writing of the paper. P.H.R. and F.R.R. also take responsibility for the integrity of the work as a whole, from inception to published article.

We would like to thank Hella Aberson for her help in developing and establishing the MLPA technique, Ank Ververs-Schreijer for data management, and Thea Visser-Oppelaar for technical help in the SMILE study. A special thanks goes to all the men who consented to participate in this study.