Article Text

Download PDFPDF

Extended report
HDL protein composition alters from proatherogenic into less atherogenic and proinflammatory in rheumatoid arthritis patients responding to rituximab
  1. Hennie G Raterman1,
  2. Han Levels2,
  3. Alexandre E Voskuyl1,
  4. Willem F Lems1,
  5. Ben A Dijkmans3,
  6. Michael T Nurmohamed3
  1. 1Department of Rheumatology, VU University Medical Center, Amsterdam, The Netherlands
  2. 2Department of Experimental Vascular Medicine, Academical Medical Center, Amsterdam, The Netherlands
  3. 3Department of Rheumatology and Internal Medicine, VUMC, Amsterdam, The Netherlands
  1. Correspondence to Michael T Nurmohamed, Department of Rheumatology and Internal Medicine, VU University Medical Centre, PO Box 7057, 1007 MB Amsterdam, The Netherlands; mt.nurmohamed{at}


  • Competing interests HR, HL and AV have no potential conflict of interest or financial support to disclose. MN received research and speaking fees from Roche. BD and MN received research grants from Wyeth, Abbott and Schering-Plough. WFL is an advisory member of Roche, Abbott and MSD.

  • Ethics approval Approval provided by the Ethical Medical Committee University Medical Center.

  • Provenance and peer review Not commissioned; externally peer reviewed.

View Full Text

Statistics from


Rheumatoid arthritis (RA) is a chronic inflammatory joint disease affecting approximately 1% of the general population.1 RA patients suffer from an elevated cardiovascular (CV) morbidity and mortality as compared with the general population.2,,4 The excess of CV disease (CVD) in RA patients is predominantly due to accelerated atherosclerosis.5 Dyslipidaemia is a well recognised risk factor for atherosclerosis and has increasing been acknowledged in RA, mostly in untreated RA patients.6,,10 Recent studies suggest an initial, transient improvement of the lipid profile by pivotal anti-inflammatory drugs like tumour necrosis factor (TNF) blocking agents,11,,14 whereas a recent meta-analysis showed a modest effect on total cholesterol and high density lipoprotein (HDL) cholesterol with no overall effect on the atherogenic index (ie, total cholesterol to HDL cholesterol ratio) after 26 weeks.15 More intriguing was the recently described favourable alteration in HDL composition in ankylosing spondylitis patients upon TNF blockade. This study observed the acute phase reactant serum amyloid A (SAA) to disappear from the HDL particle, thereby increasing the atheroprotective ability of the HDL particle.16 Information about the effects of newer biological drugs as B cell depletion therapy (rituximab; RTX) on cholesterol metabolism is sparse.17 ,18

In randomised clinical trials, RTX decreases disease activity in approximately 60%–65% of RA patients failing to respond to TNF blocking agents.19 By targeting CD20 positive B lymphocytes, RTX results in B cell depletion by inducing apoptosis. This process is dependent on the presence and conformation of lipid rafts, containing CD20, within the cell membrane of CD20 positive cells. As statins change the conformation of these CD20 containing lipid rafts, anti-CD20 therapy with RTX might be impaired in patients using statins.20 Recently, the inhibiting effect of statins on the antirheumatic effects of RTX has been described in RA patients participating in a Dutch prospective registry study of 187 RA patients, probably as a result of a conformation change.21 From this viewpoint statins might, hypothetically, not be the first drug of choice for CV risk management in RTX treated patients as it is unknown whether the atheroprotective capacities of statins will be outweighed by the negative effects on disease activity status with subsequent inflammation mediated dyslipidaemia. In this context, it is important to study the effects of RTX on the lipid profile to exclude detrimental effects of RTX on lipid metabolism in RA patients.

Therefore, we studied changes in lipids and HDL protein composition upon 6 months RTX treatment in RA patients.

Materials and methods

Patient recruitment

Consecutive RA patients, according to the 1988 revised American College of Rheumatology criteria for the diagnosis of RA, attending the outpatient clinics of the VU University Medical Center and Jan van Breemen Research Institute/Reade scheduled for RTX were followed prospectively.22

Inclusion criteria for this study are according to the guidelines of the Dutch Society for Rheumatology for treatment with RTX, that is, active disease status (defined as 28 joint disease activity score (DAS28) >3.2 at treatment initiation) despite previous treatment with TNF-blocking agents, unless contraindicated in the opinion of the treating physician, and previous treatment with methotrexate and one other disease modifying antirheumatic drug. All patients provided written informed consent and both participating clinics received approval by the local medical ethics committee.

Treatment and clinical evaluation

Patients received RTX 1000 mg intravenously on days 1 and 15, in combination with clemastine (2 mg intravenously), methylprednisolone (100 mg intravenously) and paracetamol 1000 mg orally, as premedication. Four weeks after the first infusion and from 12 weeks onwards every 3 months patients the DAS28.23 After 6 months, the change in DAS28 was assessed and expressed as clinical response according to European League Against Rheumatism response criteria.24 At all visits, blood sampling for collection of serum and plasma was performed. The use of concomitant disease modifying antirheumatic drugs, prednisolone or non-steroidal anti-inflammatory drugs during the study duration was permitted.

Assessment of lipids

Serum total cholesterol and triglycerides were analysed by an enzymatic method using the appropriate assays (Roche Diagnostics, Almere, The Netherlands) on a Cobas 6000 analyser (Roche Diagnostics) according to the manufacturer's instructions. The HDL cholesterol is determined enzymmatically by cholesterol esterase and cholesterol oxidase coupled with polyethylene glycol to the amino groups. In this oxidation reaction, a purple-blue dye will be formed and its intensity correlates to cholesterol concentration, which can be measured photometrically. Apolipoprotein (Apo) A-I and Apo B were analysed by an immunoturbidimetric method using appropriate assays (Roche Diagnostics). The Friedewald formula was used to calculate low density lipoprotein levels; when triglyceride levels were lower than 4.5 mmol/l, the total cholesterol to HDL cholesterol (atherogenic index) and Apo B to Apo A-I ratios were calculated.

HDL cholesterol protein profiling

To detect HDL composition changes between responders and non-responders during RTX treatment we selected from our cohort the six best responding patients (defined as DAS28 <3.2 and ΔDAS28 >1.2) and six patients with absence of response (defined as DAS28 >5.2 and ΔDAS28 <1.2 6 months after RTX). In this subgroup, a pilot HDL protein profiling study was performed as described previously.25 For coating of antibody, a 5 ml mixture containing 2.8 nM anti-Apo A-I monoclonal antibodies and 3 mM ethylenediamine in phosphate-buffered saline was added per spot of a PS-20 protein chip, and covalent binding of antibodies through primary amine–epoxide chemistry was achieved by incubating the chip in a humid chamber overnight at 4°C. Excess antibody was removed by one wash with distilled water and, subsequently, free amine binding places were blocked by incubating the chip with 1 M Tris buffer (pH 8.4) for 30 min at room temperature. For HDL capture, after mounting of the PS-20 protein chip(s) in a 96-well bioprocessor, 100 ml of plasma aliquots (diluted 1:2 with tris buffered saline) (50 mM tris (pH 7.4), 150 mM NaCl) was applied onto each spot and allowed to bind for 2 h at room temperature on a horizontal shaker. The protein chips were washed four times with tris buffered saline (pH 7.4, 5 min for each wash), followed by a 2-min rinse with tris buffered saline–Tween (0.005%). A final wash step with HEPES solution (5 mM) was carried out to remove the excess salt. All spots were allowed to dry and subsequently 1.2 μl of sinapinic acid (10 mg/ml) in a 50:49.9:0.1% acetonitrile–water–trifluoric acid mix was applied onto each spot. All chips were air-dried and stored at room temperature in the dark until analysis. The measurements were conducted on the same day as the chip processing.

SELDI-TOF analysis

The reliability of the surface-enhanced laser desorption/ionisation time-of-flight (SELDI-TOF) MS technique as such has been under debate recently. Since protein profiling and not the identification of proteins was the major goal we have chosen this approach because of its strong applicability in high-throughput sample measurement. Data on the reproducibility and clinical applicability of the SELDI technique as used here have been extensively reported previously and were the basis of our experimental setup as used in this study.25 ,26

In brief, SELDI-TOF analysis of the captured HDL was carried out using a phosphate-buffered saline IIc protein chip reader (Ciphergen Biosystems, Fremont, California, USA). An automated data collection protocol within the Protein-Chip software (version 3.1) was used. Data were collected up to 160 kDa. Laser intensity was set in a range of 190–220 arbitrary units at a sensitivity of 7, and the focus mass was set at 28 kd, specific for anti-Apo A-I capture. Measurement of the spectra was performed with an average of 100 shots at 13 positions per SELDI spot. Calibration was done using a protein calibration chip (Ciphergen). Spectra were normalised on Apo A-I peak area. Detected peaks having a signal to noise ratio of 5 were recognised as significant peaks.

Statistical analysis

Baseline characteristics of RA patients were expressed as mean (±SD) or median (IQR), where appropriate. Generalised estimating equation (GEE) analysis was used to analyse longitudinal data on lipids and lipoproteins measured at four different time points (baseline, 1 month, 3 months and 6 months). This is a longitudinal linear regression analysis that takes into account the same patient is measured repeatedly over time. An advantage of GEE analysis (in comparison with the commonly used methods for analysing repeated measurements) is that it does not require observations at all occasions on each person and that the intervals between the observations may vary.27 In a first crude analysis, changes in time were analysed compared with baseline with a linear regression analysis. The regression coefficients (βs) of lipid levels represent the (mean) value of baseline or the changes in lipids compared with baseline. In a second step, we adjusted one at a time for the additional covariates age, gender, use of lipid lowering drugs, smoking, prednisolone dosage, relevant comorbidity (ie, diabetes mellitus or thyroid disease), disease duration, erosions and baseline DAS28. To test whether responder status was an effect modifier for specific lipid variables, interaction terms were constructed for lipid variables and responder status. Interaction terms were considered as an effect modifier when statistical analyses revealed p values <0.05. All analyses were performed using SPSS V.15.0 (SPSS, Chicago, Illinois, USA) and p values <0.05 were considered statistically significant.


Characteristics and clinical response of RA patients

Baseline characteristics of RTX treated RA patients (n=49) are shown in table 1. Thirty-two (65%) showed response according to European League Against Rheumatism response criteria. In the total group, the mean DAS28 decreased significantly from 5.7 (±1.3) to 4.5 (±1.3) and 4.7 (±1.6), respectively, after 3 and 6 months (all p<0.001). In responders, 13% were statin users compared with 18% in non-responders. During treatment, statin users had a higher mean DAS28 score (∼0.8 points, p for trend=0.075). None of the patients initiated lipid lowering therapy during follow-up.

Table 1

Baseline characteristics of rituximab treated RA patients

Cholesterol changes in the total group

In the total group, levels of HDL and Apo A-I levels increased and the atherogenic index and Apo B to Apo A-I ratios decreased (p values: 0.010, 0.003, 0.089 and 0.029, respectively) after 1 month of treatment (table 2). The regression coefficients of lipid levels between baseline and follow-up did not change and remained statistically significant after adjustment for other covariates one at a time (including age, gender, smoking, relevant comorbidity (ie, diabetes mellitus or thyroid disorder), statin use, prednisolone dosage, disease duration, erosions and baseline DAS28) (data not shown). Mean atherogenic index decreased from 4.3 to 3.9 (∼9%) 6 months after initiation of RTX (p=0.054). Adjustment for other covariates did not have any influence on the decrease in atherogenic index (p=0.052). Testing for effect modification revealed a difference in effect on lipid levels between responders and non-responders upon RTX (p<0.001).

Table 2

Lipid levels during treatment with rituximab in RA patients (n=49)

Cholesterol changes in responders upon RTX

No significant changes in cholesterol levels were observed in non-responders. In responders (n=32), total cholesterol, HDL cholesterol, Apo A-I and Apo B to Apo A-I ratios changed significantly after 1 month of treatment of RTX (table 3). Adjustment for other covariates did not change the observed effects. After 3 months, triglycerides and Apo B to Apo A-I ratios changed significantly. Six months after RTX, HDL cholesterol changed ∼5%, Apo A-I ∼5%, atherogenic index ∼7% and the Apo B to Apo A-I ratio ∼9% (p=0.035). The observed effects of these cholesterol profiles did not change after adjustment for confounding factors (data not shown).

Table 3

Lipid levels in responders upon rituximab in RA patients treated with rituximab (n=32)

Lipids, disease activity and prednisone

The association between lipid levels and markers of disease activity is shown in table 4. No significant association was observed between prednisone dosage and lipid levels. In contrast, significant associations were observed between changes in DAS28 on the one hand and changes in atherogenic index, total cholesterol, HDL cholesterol, Apo B to Apo A-I ratio or Apo A-I levels on the other hand. For example, an increase of 1 point DAS28 will decrease total cholesterol with 0.11 mmol/l and HDL cholesterol with 0.052 mmol/l, resulting in a raise of the atherogenic index of approximately 0.11.

Table 4

Relationship between lipids and markers of disease activity in RA patients treated with RTX (n=49) during 6 months RTX

SELDI-TOF analysis in responders versus non-responders during treatment

Characteristics of proteome analysed RA patients are shown in table 5. A significant decrease in density of mass charge (m/z) marker 11743 representing SAA-1 in good responding RTX patients when compared with non-responding patients was observed (figure 1). Moreover, seven additional markers (ie, 3893, 7065, 8647, 9401, 9764, 15969 and 87407 m/z) changed in the good responding patients (all p<0.05). In both groups, one patient was using statins.

Figure 1

Changes in high density lipoprotein (HDL) protein signature using surface-enhanced laser desorption/ionisation time-of-flight MS analysis. (A) Depicted are representative profiles (virtual gel-views) of one individual from on-chip captured HDL from baseline (T0) to 6 months after start of rituximab treatment (T3). The two lines indicate the serum amyloid A (SAA) region in the spectra. All spectra were normalised on total ion current and baseline corrected and subsequently translated to the ‘virtual gel-view’ output format. (B) Mean SAA expression (±1 SD) over the good-responding (n=6) and non-responding (n=6) patients as indicated.

Table 5

Baseline characteristics of proteome analysed RA patients


This study shows favourable effects in the atherogenic index and the Apo B to Apo A-I ratio of RA patients treated with RTX. Along with these beneficial effects on the lipid profile, functional alterations are observed in HDL particle protein composition, that is, from proatherogenic to a less proinflammatory and less proatherogenic composition in good responding patients during treatment, as shown by a decline in SAA concentration on the HDL particle during RTX treatment.

This prospective study is by far the largest report that studies alterations in lipid levels during RTX treatment in RA. The first study in pilot setting performed in only six RA patients described an increase in total cholesterol, HDL cholesterol and low density lipoprotein cholesterol levels, especially 2 weeks after RTX initiation.17 Another pilot study in five RA patients observed significant improvement of HDL cholesterol.18 In line with these reports, the present study shows a borderline significant (p=0.054) improvement of the atherogenic index (∼9%) and non-significant increase in HDL cholesterol (∼5%) in the total group after 6 months treatment of RTX. Moreover, in good responding RTX patients a significant improvement of the Apo B to Apo A-I ratio and a (non-significant) improvement of HDL cholesterol and atherogenic index was observed. Moreover, this study demonstrates significant associations between the acute phase reactant C reactive protein and lipid levels; hence, the observed changes are at least partly explained by the decline in inflammation. These results underline the importance of tight control with anti-inflammatory treatment in RA as low disease activity decelerates radiological progression,28 and seems to improve the CV risk profile by an, albeit limited, improvement of the atherogenic index in a patient group prone to CVD development.

In addition to the long term effects during RTX treatment, early effects (ie, 1 month after RTX initiation) on lipid levels were observed and did not sustain after 3 months, as was previous described by Gonzalez-Gay and coworkers.17 Moreover, relationships between high dose prednisolone and a rise in total and HDL cholesterol up to 4 months after initiation of glucocorticoids have been demonstrated.29 This supports the idea that concomitant use of glucocorticoids due to high disease activity prior to RTX initiation increases total and HDL cholesterol levels, which decline when glucocorticoids are tapered.30 Although this hypothesis seems plausible and was previously demonstrated in anti-TNF treated RA patients,31 no association was observed between prednisone dosage and lipid levels. On the other hand, no studies have been performed investigating the (long term) effects of pulsed high dose methylprednisolone on lipid levels. Therefore, it seems plausible that the initial effects on lipid levels during RTX treatment are at least partially explained by concomitant glucocorticoid use as premedication.32

Although the observed changes in lipid levels were small, it is remarkable that this study demonstrates significant improvement of Apo B to Apo A-I ratio and the atherogenic index, as a recent meta analysis of our group described significant parallel increases in total and HDL cholesterol levels and no overall effect on the atherogenic index during TNF blocking therapy.15 In contrast, in the present study we observed a significant improvement of the atherogenic index with a decline to ∼3.8 that is below the threshold (4.0) considered desirable by some.33 Moreover, reduction of HDL during lipid lowering drugs was estimated on approximately 5%, comparable with the effect in this study during RTX treatment, and previous reports demonstrated that 6% increase in HDL levels will result in an absolute risk reduction of ∼4.5% of (non) fatal coronary heart diseases.34 ,35 Although data about patients using statins cannot be directly extrapolated to RA patients among others due to heterogeneity in populations, more pleiotropic and pivotal cardioprotective effects of statins, these results ensure at least no detrimental effect of RTX on lipid levels and favour a mild beneficial long term effect of RTX on lipid levels.

Next to the quantitative changes in lipid levels, we performed SELDI-TOF MS analysis to study HDL particle composition changes. These changes in protein signature are important as new insights reveal HDL protein composition to correlate to HDL function.36In contrast to the functionality of the HDL particles cholesterol levels only provide a measure of the size of the HDL pool. Functional properties of normal HDL are among others downregulating cell adhesion molecules in endothelial cells, regulating TNF production by macrophages, antioxidative and anti-inflammatory capacities by reducing lipid oxidative products.37 These functional components are lost in situations of acute infection or disease status resulting in a more proinflammatory and proatherogenic HDL protein composition.38 ,39 This study demonstrates a proinflammatory HDL particle composition in active RA patients illustrated by a large amount of SAA-1 binding of the HDL particle before initiation of RTX treatment. This observation is in line with the HDL proteome identification study in RA patients performed by Watanabe et al.40 This study described HDL proteome differences in acute phase reactants like SAA, complement factors and thrombotic markers between RA patients with high and low disease activity. Previously, an inverse association between active disease status and an anti-inflammatory HDL compound was described.41 Moreover, this group demonstrated recently an impaired reverse cholesterol transport, a major antiatherogenic function of HDL cholesterol, in active RA patients.42 Intriguingly, we were able to demonstrate that HDL proteome alters to a more anti-inflammatory and (possible) less proatherogenic profile during 6 months RTX as shown by a significant reduction of SAA-1 in the HDL protein content in good responding RTX treated RA patients. This observation confirms the findings of alteration in HDL composition during anti-TNF treatment in patients with ankylosing spondylitis.16 All these observations emphasise the pathogenic role of proinflammatory HDL in atherosclerosis as SAA-enrichment of HDL particles has been postulated to facilitate the binding of HDL to vascular proteoglycan and both Apo A-I, the main lipoprotein of HDL, and SAA have been demonstrated in atherosclerotic lesions.43,,46 However, responders and non-responders of the proteome analysed group differed in baseline characteristics with respect to proclaimed biomarkers for RTX treatment. To what extent these baseline differences fully explain the dynamics in HDL composition cannot be assessed with the current study design and sample size. On the other hand, the observation of a significant reduction of seven additional proteins or protein fragments (which remain to be identified) underlines the dynamics of the HDL protein particle composition as a consequence of changes in inflammatory status during anti-inflammatory treatment. Therefore, we encourage new studies in high CVD risk patients groups like RA to focus on inflammation and functional capacities of HDL in relation to atherosclerosis. Moreover, it remains an intriguing question whether the proteins found at baseline serve as a predictor for RTX response.

In conclusion, this study shows at least no detrimental effect on lipid levels upon 6 months RTX treatment and favours late beneficial effects on the cholesterol profile along with improvement of disease activity. Proteomic HDL particle profiling revealed composition changes from proatherogenic to a less proinflammatory and less proatherogenic composition during RTX. These results emphasise the need for unravelling the function of HDL composition in inflammatory rheumatic disorders in relation to CV outcome.


The authors acknowledge MTHM de Koning and T de Gast for performing laboratory assessments and EE Platek for collecting clinical data.


View Abstract


  • Competing interests HR, HL and AV have no potential conflict of interest or financial support to disclose. MN received research and speaking fees from Roche. BD and MN received research grants from Wyeth, Abbott and Schering-Plough. WFL is an advisory member of Roche, Abbott and MSD.

  • Ethics approval Approval provided by the Ethical Medical Committee University Medical Center.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.