The initial CRP assay carried out using RAP assay is designed around a two channel sensor 'chip'. Sheep anti-CRP was covalently coupled to the test channel using standard EDC/NHS coupling chemistry. Sheep IgG coupled to the other channel demonstrated very low background signal in the appropriately diluted spiked serum samples. Sample and/or standards are passed over these channels in parallel to give a fairly rapid assay of 30 minutes per cycle (Figure 1). CRP is often monitored in autoimmune diseases where samples containing rheumatoid factor have very high incidence 26. The sheep IgG channel can provide a suitable control for both this or anti-animal antibodies that are a potential interference in immunoassays 27282930.

This initial CRP assay carried out using RAP assay was compared with a commercial, validated high sensitivity ELISA by analysing 6 spiked horse serum samples across the range of AHA/CDC guidelines. Good correlation for spiked serum samples above 5 μg/mL was found in all CRP assay formats using RAP (Figure 2) when compared to commercial hsCRP ELISA (Table 1). In particular, the direct sandwich assay also showed good correlation below this level (R² = 0.998; Table 1, Figure 3) The mean difference between the two methods for estimating serum CRP as calculated by Bland-Altman analysis (Figure 3b) was 2.17 μg/mL and the limits of the standard deviation (2SD) is 6.78 μg/mL. The differences in values obtained by the two methods lay within mean +/- 2SD. The methods agree well, whilst the RAP method gave higher values at 44 μg/mL and 116 μg/L, the difference at this level would not hinder classification according to the AHA/CDC guidelines and both methods would infer other sources of inflammation (bacterial, viral).

The detection limit of the procedures was the amount of CRP that could produce a signal in the test (anti-CRP channel) equivalent to the mean value of duplicate zero mg/L CRP injections plus 3 times the standard deviation of the zero standard. For direct capture this was found to be 13 ng/mL, for homogenous sandwich Assay 20 ng/mL and for the direct Sandwich Assay 3 ng/mL.

Precision for the direct sandwich assay was determined using 3 test channels, injected with standard CRP concentrations from 0 to 232 ng/mL (Figure 4). Below 10 ng/mL the coefficient of variation (CV) rose above 10%, above this CRP concentration, a CV of 11.3% decreasing to 4.7% was observed.

An inter-assay, intra-assay precision profile analysis was performed by determining CRP concentrations of spiked serum samples using the sandwich assay in three to five replicates of each sample within test channels and different test channels (Table 2). The coefficient of variation (CV) lay between 3.1 to 12.6% across the range 0.1 to 116 mg/L original concentration of spiked serum. Generally acceptable CV values in diagnostic methods are less than 10%, the smaller the CV the more accurate the classification of sample. However, the level of imprecision found herein is similar to that of commercial hsELISA (e,g, IBL hsELISA, Hamburg, Germany quotes intra-assay CV of 5.5 and 6% for two samples of 22 and 99 ug/L CRP and inter-assay variation of 11.6 and 13.8% for two samples of 22.1 and 90.4 ug/L CRP) and the results still indicate the assay is useful in differentiating the cardiovascular risk levels.

In order to test the possibility of false negative results due to high CRP levels, the standard CRP range was extended approximately two fold higher (231 ng/mL) than the hsELISA range. No hook effect was observed at this level (Figure 4).

Whilst the standard CRP assay was able to provide quantitative results, calibration of a sensor channel response using standards prior to sample is a relatively time consuming process. Since the turnaround time is 10 minutes per sample, then five singlet calibration standards prior to a sample would take one hour turnaround. To deliver a more rapid semi-quantitative assay from a blood sample, a simple, rapid ratio metric assay was thus performed. A normalisation standard corresponding to a blood sample with 3 μg/mL CRP in serum was diluted 1 in 50 then passed over one channel. In parallel, a blood sample diluted 1 in 50 was passed over the other channel (Figure 5). The chosen CRP normalisation concentration corresponds to that of the 90^th percentile and also the borderline between 'average' and 'high' AHA/CDC guidelines. If the signal ratio between the two channels is greater than 1 then a higher CRP level is present and thus 'high' risk, if less than 1, 'low' risk and at 1 is borderline. Spiking of 3 separate blood samples at 'average' (1.5 μg/mL CRP in serum or 0.68 μg/mL in whole blood) and 3 separate blood samples at 'high' (15 μg/mL in serum or 68.18 μg/mL CRP in whole blood) CRP levels were tested. Ratios obtained gave excellent correlation with that expected for a calibrated sensor channel at these levels. For average CRP level blood the value obtained was 0.51 +/- 0.06 (expected ratio 0.5, n = 3) and for high CRP level blood the value was 1.45 +/- 0.2 (expected ratio 1.3, n = 3).

Spiked horse serum was tested for CRP content using a validated commercial hsELISA kit from Kalon Biological (Kalon Biological, U.K.). The assay was conducted according to the manufacturer's instructions, with spiked serum diluted to as low as 1 in 5 to as much as 1 in 10,000 in the supplied sample diluent.

RAP experiments were conducted using automated instruments (Akubio Ltd, Cambridge, UK). The instruments apply the principles of QCM, in that a high frequency voltage is applied to a piezo-electric crystal to induce the crystal to oscillate, and its resonance frequency is monitored in real time. The four-channel instruments comprise two pairs of oscillating crystal sensors mounted in parallel microfluidic flow cells, allowing sample to be flowed across four surfaces simultaneously. As sample is flowed across sensors, binding, if any, is measured as a reduction in the oscillation frequency.

The RAP instruments were fitted with a thermally-stable sensor mounting block providing temperature control, and with microfluidic and electrical connections to the piezo-electric sensors. Buffer flow was maintained with syringe pumps (Tecan UK Ltd, Reading, UK) under software control (Akubio Ltd., Cambridge, UK). Microfluidics comprised separate flow-paths to individual flow cells, combined with a common flow path split to address flow cells simultaneously. Interchange between the different flow paths was either by manual or electronically-operated valves (Akubio Ltd). Disposable AKT◆iv sensor covalent A acrylic sensor cassettes were employed in this work (Akubio Ltd., Cambridge, UK. These contain gold-coated quartz wafers with a carboxylic acid-terminated monolayer coating to provide a surface for protein immobilisation. Each cassette contains two derivatised sensors. These can be addressed independently via a micro-fluidic system that is integrated in to the AKT◆iv sensor cassette. One of the flow cells (channels) can be used as a control for real-time measurements if required. Two AKT◆iv sensor cassettes can be docked into Akubio's RAP instruments, allowing four simultaneous independent measurements to be carried out.

Sensor surfaces were prepared by immobilising sheep anti-CRP onto the 'active' sensor surface and sheep immunoglobulin type G (Sh IgG) onto the 'control' sensor surface using conventional amine coupling chemistry. Immobilisation was performed at room temperature under continuous flow conditions with a running buffer, PBS, between sample injections was at a flow rate of 25 μL/min. Each injection step taking 3 minutes. First sensor surfaces were activated with a 1:1 mixture of 400 mmol/L EDC and 100 mmol/L NHS prepared in 0.22 μm-filtered deionised water, and mixed immediately prior to use (final concentrations; 200 mmol/L EDC and 50 mmol/L NHS). EDC-NHS was injected simultaneously across both sensor surfaces. Sheep anti-CRP and Sh IgG were prepared for immobilisation at 25 μg/mL in 10 mmol/L sodium acetate, pH 4.5, and were injected simultaneously across separate sensor surfaces. Non-reacted surface was then capped with BSA prepared at 100 μg/mL, again in 10 mM Sodium Acetate pH 4.5 and injected simultaneously across all sensor surfaces. Finally, the surface of sensors and microfluidic channels was blocked with 100 ug/mL BSA in Tris Buffered Saline (TBS). At the end of the procedure, between 412 Hz and 420 Hz of anti-CRP and 320 and 340 Hz Sheep IgG were immobilized on individual and approximately 630 Hz after capping/blocking with BSA on individual flow channels. The resulting "sensor chips" were stored at 4°C until required.

Normal horse serum spiked with human CRP was supplied by Kalon Biological (UK). Spiked horse blood was prepared as follows. Spiked whole horse blood collected in EDTA tubes was kept refrigerated and used within 24 hours of collection. The blood was centrifuged in 1.5 mL microcentrifuge tubes at 3000 × g for five minutes at 4°C, the upper layer was then aspirated. The whole blood volume was reconstituted by addition of the spiked serum appropriately diluted in normal serum to give blood spiked with human CRP at 1.5 μg/mL (average CRP) and 15 μg/mL (High).

All assays were performed at room temperature under continuous flow at 25 μL/min with a running buffer of TBS, 0.005% Tween-20. CRP standards were prepared in a sample buffer comprising TBS containing 0.005% Tween-20 and 100 μg/mL BSA from a concentrated stock solution (94.8 μg/mL CRP). CRP spiked horse serum was also appropriately diluted in the same sample buffer from a 1/50 to 1/6000 dilution

After gently mixing, spiked horse blood samples were diluted 1 in 50 in sample diluent (0.1 mg/mL BSA in 10 mM HEPES, 150 mM NaCl, 3 mM EDTA pH 7.4). The normalisation standard was 27 ng/mL CRP in the same buffer (equivalent to 3 μg/mL serum in whole blood diluted 1/50). A standard assay was then performed with Sheep anti-CRP sensor surfaces as previously but using running buffer of 10 mM HEPES, 150 mM NaCl, 3 mM EDTA pH 7.4. For each dual sensor channel chip, standard was passed over one channel, spiked sample over the other. The blood preparation and CRP screen was performed on three separate occasions. Blood samples were mixed by aspiration-dispense prior to loading onto sensor surface.

RAP data was analysed as initial rates of signal generation (Hz/s) upon injection of CRP or antibody onto a test channel containing immobilised anti-CRP. The data was displayed and analyzed using RAP◆id Workbench v1.0.25 (Akubio Ltd., Cambridge, U.K.). Statistics were generated using Excel, and estimation of spiked samples was performed using a 4-parameter plot of the standards and appropriate dilution of the unknown spiked sera. For the semi-quantitative assay, the initial rate of signal at the sandwich step was corrected for baseline slope and the ratio of blood sample signal to normalisation standard signal was estimated.