Expedient methodology for total methotrexate polyglutamation pool determination in human erythrocytes

Abstract

The measurement of methotrexate polyglutamate metabolites in red blood cells has potential to aid in individualization of methotrexate therapy in rheumatoid arthritis and juvenile idiopathic arthritis. In this report a method is presented for rapid analysis of these metabolites in human red blood cells. The analytical procedure is a simple “one pot” pre-column reaction involving the addition of sodium dithionite as a reducing agent followed by 15 minutes of boiling. After centrifugation the supernatant is introduced into a conventional HPLC system equipped with a fluorescence detector, without the need for further workup. By performing the derivatization reaction pre-column, a time consuming (6–14 h) commonly used deglutamation procedure utilizing blank human plasma, becomes obsolete. Using the described procedure the total sample preparation time for a 50 sample run should not exceed 1–1.5 hours. The chromatographic run time per sample is 7 minutes using isocratic elution conditions. The method was found to be linear over the clinical relevant concentration range of 10–500 nM of intra-cellular methotrexate polyglutamates. The intra-run mean accuracy of the target value was between 98.1% and 106.0%. The intra-run precision was between 1.2% and 8.8%.

---

Introduction

The antifolate methotrexate (MTX) (4-amino-10-methylpteroylglutamic acid) administered in low dose, on a weekly basis, is widely used for the treatment of rheumatoid arthritis (RA)¹ and juvenile idiopathic arthritis (JIA).² MTX is the first-line therapy for RA patients exhibiting inadequate disease control from treatment with non-steroidal anti-inflammatory drugs.³ In low-dose this disease-modifying drug is generally well tolerated and considered safe, however, there is a substantial percentage of patients who have poor response to MTX or develop adverse reactions.⁴ This can be partly explained by a poorly understood inter-patient variation in the dose of MTX required to achieve a desirable level of disease control, making it difficult to optimize and individualize therapy. It has recently been shown that the measurement of MTX and its polyglutamate metabolites (MTXPGs) in red blood cells (RBCs), an easily accessible space thought to reflect the intra-cellular state, could be a useful tool in individualizing MTX therapy.^5–8 Whilst the exact mechanism of action of MTX in RA/JIA is poorly understood,^9,10 its therapeutic effect is attributed to the various MTXPGs rather than MTX itself.^11–13

Various methods to measure intra-cellular levels of MTXPGs have appeared recently in the literature. These strategies can be categorized as methods that measure the total amount of polyglutamated MTX (i.e. as a total pool) (MTXPG_total) and methods that measure and quantify each MTXPG individually, up to the hepta-glutamate. Individual concentrations of MTXPGs in RBCs have been measured by HPLC with a post-separation photochemical (PCR(hv)) derivatization followed by fluorescence detection (FD),⁶ and more recently by an ion-pair HPLC method followed by tandem mass spectrometry (LC/MS/MS).¹⁴ Currently, the LC/MS/MS method is the most specific and sensitive assay for detection of MTXPGs. Whilst these methods give a detailed presentation of the intra-cellular MTX metabolome, calibration is required for each MTXPG, necessitating the use of expensive standards. Further, gradient elution separation is required to compensate for differences in chromatographic behavior of the different MTXPGs, typically resulting in a relatively long run and re-equilibration times. Clinically, the relevance of individual MTXPG concentrations or population distribution is not fully understood, and as a result the various polyglutamates are often summed to a total or alternatively, grouped as short-chain (MTXPG_1–2) and long-chain (MTXPG_3–5)^5,6 pools.

MTXPG_total methods utilize the plasma enzyme γ-glutamyl hydrolase to deglutamate the various MTXPGs into MTX (in a similar fashion as folate polyglutamates).¹⁵ Typically, an aliquot of human plasma is added to a RBC lysate as a source of γ-glutamyl hydrolase, followed by a lengthy incubation period ranging from 6 to 14 hours at 37 °C. The amount of MTX after deglutamation (reflecting the MTXPG_total) is generally measured by HPLC-PCR(hv)-FD.^6,16 The post-column oxidative degradation used by these methods generates a fluorescent pteridine derivative by cleaving the C9–N10 bond of MTX(PGs)^17,18 (Fig. 1a). Hence, the only function of the elaborate deglutamation procedure is to circumvent chromatographic resolution between the various polyglutamates. We hypothesized that a pre-column derivatization strategy leading to cleavage of C9–N10 bond of MTX(PGs) would circumvent the tedious deglutamation step and as a result increase sample throughput, reduce the number of sample preparation steps and avoid the addition of blank human plasma and the ensuring incubation period. The presently proposed analytical method would possess the attributes of being rapid and economical, an attractive alternative for the routine monitoring of total intra-cellular MTX related species in JIA and RA patients, while still providing a quantitative basis for individualization of MTX therapy.

Fig. 1 The labile C9–N10 bond in MTX can be cleaved in oxidative (a) and reductive (b) environments. The resulting fluorescent derivative bears either a carboxylic acid or a methyl substituent at the C6 position, depending on the chemistry selection.

Materials and methods

Chemicals and reagents

Methotrexate (MTX) was purchased from Sigma Aldrich (St Louis, MO, USA) and methotrexate polyglutamates (MTXPGs) 2–7 were purchased from Schircks Laboratories (Jona, Switzerland). 2,4-Diamino-6-methylpteridine (DAMP) was obtained from ChemDiv (San Diego, CA). HPLC grade methanol (MeOH), sodium phosphate monobasic and ammonium bicarbonate were purchased from Sigma Aldrich (St Louis, MO, USA). Ammonium acetate and sodium phosphate dibasic were acquired from Fisher Scientific (Fairtown, NJ, USA). Sodium dithionite was purchased from Riedel—de Haën (Seelze, Germany). Demineralized water was produced by Labconco Waterpro PS (Labconco Corporation, Kansas City, MO, USA). Blank red blood cells (RBCs) were drawn from healthy volunteers on site.

Chromatographic separation

Solvent was delivered by a binary pumping system containing two Shimadzu LC6A pumps. The sample was introduced by a Shimadzu SIL-6B auto injector equipped with a 50 µL injection loop and a 50% water/50% ACN wash solution. Detection was accomplished by a Shimadzu RF-10Axl fluorescence detector using an excitation–emission wavelength of 367 nm and 463 nm, respectively, with a bandwidth of 15 nm. Chromatography was performed using a Phenomenex Inertsil ODS-3 analytical column (150 × 4.6 mm) packed with 5 µ 100 Å particles. The analytical column was guarded using a Supelcosil LC-8 guard column (5 µ, 2 × 4.0 mm). Mobile phase A consisted of 10 mM ammonium acetate in water and mobile phase B was 100% methanol. An isocratic mobile phase containing 30% B at a flow rate of 1 mL min⁻¹ was used for the separation.

Stock solutions and buffers

Phosphate buffers (1.0 M) of pH 5.5, 6.0, 6.5, 7.0 and 7.5 were obtained by blending a 1.0 M monobasic sodium phosphate solution with a 1.0 M dibasic sodium phosphate solution to achieve the targeted pH. MTX and MTXPGs were dissolved in 0.10 M ammonium bicarbonate buffer to prepare 1.0 mM master stock solutions. The working stock solutions containing the appropriate concentrations of the analytes were created by further dilution of the master stocks solutions with water. All stock solutions were prepared daily prior to each experiment.

Preparation of erythrocyte (RBC) lysates

Blood samples (∼5 mL) obtained from patients were centrifuged at low speed (2000 rpm) in a Beckman tabletop centrifuge to pellet the RBCs. After removal of plasma, the RBCs were suspended in an equal volume of sterile normal saline, mixed by gentle inversion and subjected to a second low speed centrifugation. The supernatant was discarded and the wash procedure was repeated a second time. After discarding the supernatant, the packed RBCs were divided into aliquots and stored at −70 °C until use.

Derivatization protocol

A general derivatization procedure was established. An 100 µL aliquot of patient or blank RBCs was transferred to an Eppendorf reaction vial and was spiked with one of the analytes. The volume was adjusted to 200 µL using deionized water and/or aqueous calibration solutions of MTX to obtain spiked RBCs of the appropriate concentration, and subsequently homogenized via vortex mixing for 5 seconds. The pH of the homogenate was adjusted by the addition of 600 µL of pH 6.0 phosphate buffer followed by an additional 5 second vortex step. The samples were placed in an in-house constructed sample holder that was able to pressure seal the caps of the Eppendorf reaction vials. The total reaction volume was brought to 1.0 mL by addition of 200 µL of a freshly prepared solution of 10 mg mL⁻¹ sodium dithionite. After sodium dithionite addition to each vial, the lids were closed and the entire holder inverted 5 times, then vortexed for 10 seconds to ensure complete mixing (this step was found to be extremely important in obtaining reproducible results). The sample holder was then placed in boiling water for 15 minutes to facilitate the reduction reaction and precipitate proteins. After the reaction period, the sample holder was refrigerated at 7 °C for 30 minutes. The reaction vials were subjected to centrifugation (13 000 rpm for 5 minutes), and the resulting supernatant transferred to a 1.0 mL autosampler vials for subsequent determination.

Yield determination and validation

In order to determine the yield of the reduction, 100 µL of RBCs (n = 5) were spiked with MTX to 1.0 µM and subjected to the derivatization procedure. The reduction yield was obtained by comparing the spiked RBC samples to an aqueous DAMP standard calibration curve (0.1, 0.5, 1.0, 1.5, 2 µM). Method validation consisted of determination of MTX spiked RBCs at 10, 25, 50, 100, 250 and 500 nM in 5 replicates on four consecutive days.

Results and discussion

Selection of the derivatization reaction and format

The order of operations within an overall methodology has significant analytical implications. For example, the conduct of a derivatization reaction in the pre- versus post-column format defines the nature of the data obtained. Conduct of a post-column derivatization reaction allows for separation of the various analytes of interest prior to formation of an analytical reporter product. In contrast, pre-column derivatization of a family of analytes may lead to the formation of a common product and thus loss of analytical identity. As noted previously, there are clinical situations where determination of the total concentration of all related species is sufficient and desirable, rather than determination of each individual analytical species

Continuing with the concept of total analyte determination, as previously noted, the C9–N10 bond in methotrexate is chemically labile to oxidative^6,16–18 and reductive¹⁹ environments to yield fluorescent pteridine derivatives (Fig. 1). These different processes lead to different products, with oxidative cleavage leading to the formation of 2,4-diaminopteridine-6-carboxylic acid (DAPC) (Fig. 1a) while reduction cleavage leads to 2,4-diamino-6-methylpteridine (DAMP) (Fig. 1b). A prior publication suggests that the reductive product DAMP exhibits more favorable fluorescence properties as compared to the oxidative product DAPC, with a yield of approximately 70% being realized when the analyte is present in plasma.¹⁹

In preliminary studies (results not presented), it was found that sodium dithionite reductive mediated reductive cleavage of the C9–N10 bond may be extended to the situation where MTX is conjugated to a macromolecular carrier via the glutamyl residue. The apparent generality of this reaction leads to the present hypothesis that sodium dithionite reductive mediated reduction could be used to measure MTXPG_total (DAMP serves as a common reporter molecule for each MTXPG). In another consideration with respect to the choice of an oxidative versus reductive process to mediate C9–N10 cleavage, the previous investigation had also shown the reductive procedure to result in significantly less chromatographic interferences as compared to the oxidative procedure. With any bioanalytical method, simplistic and efficient sample preparation is a hallmark of a robust method. In the present case, it was observed that by conducting the reductive procedure directly in plasma (reaction for 15 minutes in boiling water), efficient protein denaturation was accomplished, and after centrifugation, the supernatant could be directly analyzed by HPLC-FL. When this procedure was investigated for the analysis of MTXPG_total in human RBCs, the chromatograms in Fig. 2 were obtained. A DAMP reference standard dissolved in water was used to optimize chromatographic conditions (Fig. 2a). Blank human RBCs showed no significant interference originating from endogenous molecules during the elution window of DAMP (Fig. 2a). Due to the low background a rapid isocratic separation was possible, leading to an analysis time of just 7 minutes per sample. In order to validate the hypothesis that the various MTXPGs would yield the same product (DAMP), the various individual MTXPGs were spiked in similar concentrations and derivatized in human RBCs obtained from individuals not treated with MTX. Identical chromatograms were generated regardless of the derivatized MTXPG (Fig. 2b).

Fig. 2 (a) Illustration of a blank RBC chromatogram (black line), showing no endogenous interference during the elution of the DAMP derivatization product (blue line). (b) 7 individual chromatograms of RBC samples, each spiked with a different MTXPG standard (1–7).

Yield optimization and comparison

Deen and co-workers stated in their plasma method that no increase in fluorescent signal was observed after 15 minutes of reaction time and/or an increase in dithionite concentrations.¹⁹ Similar observations were made with the reaction in the RBC matrix. The influence of the pH on the reduction reaction of MTX with an RBC matrix was investigated at pH 5.5, 6.0, 6.5, 7.0 and 7.5 (Table 1). The highest fluorescence was obtained at pH values of 5.5 and 6.0. In order to maximize the buffer capacity of the phosphate buffer, a pH of 6.0 was selected. It has been shown that patients on MTX therapy show large inter-patient variability in MTXPG distributions (short-chain or long-chain MTXPGs being dominant). Since MTXPG_total is calculated as the sum of all glutamation species from MTXPG₁ (parent) to MTXPG₇, it was essential that derivatization of the various MTXPGs proceeds with similar yields in order to avoid analytical bias due to differences in the MTXPG population distribution. MTXPG_1–7 were individually (n = 5) spiked at 1.0 µM in the RBC matrix and subjected to the reductive procedure (Table 2). The average yield of reaction was 61.6%, with MTXPG₂ giving the highest (65.1%) and MTXPG₇ the lowest (56.6%) yield. The MTXPG₇ reduction yield was significantly lower than the mean using the student t-test at the 95% confidence interval. However, this standard is very hygroscopic and had visually picked up some water during storage. Therefore it was concluded that reaction yield between the different MTXPGs was identical, with observed differences being caused by minor variations in the purities of the commercially obtained MTXPG standards. As a result, the analytical method can be calibrated with reference to MTX thus avoiding the need for specialized expensive MTXPG standards.

Table 1 pH influence on reaction yield

pH	Relative yield	Precision (RSD %)
5.5	1.00	1.5
6.0	0.96	3.2
6.5	0.74	2.2
7.0	0.56	3.4
7.5	0.41	3.3

---

Table 2 Reduction yields of various MTXPGs

	Concentration/µM	Found/µM	Yield (%)	RSD (%)
MTX	1.00	0.62	61.9	1.8
MTXPG2	1.00	0.65	65.1	1.4
MTXPG3	1.00	0.61	61.0	1.1
MTXPG4	1.00	0.64	64.4	2.0
MTXPG5	1.00	0.63	63.5	0.6
MTXPG6	1.00	0.59	59.1	1.3
MTXPG7	1.00	0.57	56.5	2.8

---

Matrix influence

During method development it was observed that sodium dithionite-mediated reduction to yield DAMP was matrix dependent. Unexpectedly, in aqueous solutions the reaction was less efficient (absolute yield was about 5–10%) as compared to RBC matrix (yield about 62%). As this variation was of considerable concern initially, further experimentation was conducted in an effort to establish the nature of this variability and to define conditions leading to reproducibility. RBCs from four separate individual donors were spiked (n = 5) with an equal amount of MTX and subsequently analyzed (Table 3). The various lots of RBCs did not significantly affect the derivatization yield and as a result the ability of the method to quantitate MTXPG_total in individual RBC samples is not compromised. Since the derivatization proceeds more efficiently in RBCs than water, it was established that it is very important to keep RBC volumes between calibrants and samples constant during the derivatization process, or in other words an identical matrix must be maintained for all samples, unknowns or standards, being subjected to sodium dithionite reduction.

Table 3 Yield of reaction in various individual RBC donors

RBC donor	Spiked/nM	Found/nM	Yield (%)	RSD (%) (n = 3)
1	100	109.5	109.5	11.4
2	100	107.7	107.7	1.5
3	100	111.7	111.7	4.3
4	100	107.8	107.8	1.6

---

Method validation

Using MTX the method was validated on four consecutive days at six different concentrations from 10–500 nM, representing a clinically relevant MTXPG_total range reported in RBCs (Table 4).

Table 4 Intra- and inter-day precision and accuracy results

Nominal RBC MTXPG concentration/nM	Intra-run (n = 5)			Inter-run (n = 20)
	Mean observed concentration/nmol L^−1	Mean accuracy of target value (%)	Precision (RSD %)	Mean observed concentration/nmol L^−1	Mean accuracy of target value (%)	Precision (RSD %)
10	10.6	106.0	5.6	10.0	99.7	9.6
25	25.7	102.7	2.3	24.8	99.2	7.5
50	50.0	100.0	8.8	49.6	99.2	6.1
100	98.1	98.1	3.2	98.7	98.7	4.1
250	250.6	100.2	2.6	253.2	101.3	2.1
500	500.0	100.0	1.2	498.7	99.7	1.6

---

The intra-run mean accuracy of the target value was between 98.1% and 106.0%. The intra-run precision was between 1.2% and 8.8%. The limit of quantification was determined to be 10 nM MTXPG_total. The limit of detection, defined as 3 times the signal to noise ratio, was 4 nM MTXPG_total. Four different calibration curves at different days revealed a linear relationship between peak area and concentration of the prepared RBC standards, with correlation coefficients r² > 0.999 for each day. A typical calibration curve was described by the following linear equation: y = 376.7x + 832.1, where y is peak area and x is concentration added in nM. The derivatization procedure results in a 10 fold sample dilution, however, limits of quantification were within assay requirements, so no effort was made to preconcentrate the sample prior to injection.

Stability

The stability of the reporter molecule (DAMP) was investigated at three different concentrations (10.0, 50.0 and 250.0 nM). After sample processing the samples were placed in the autosampler and analyzed 24 hours later (Table 5). Sample deterioration was not observed, indicating that the processed samples are stable for at least 24 hours at room temperature, which is sufficient for analysis. The freeze–thaw stability was investigated in a similar manner (Table 5). Recently, Hroch et al. have reported sample instability when multiple freeze–thaw cycles were applied (method of detection was enzymatic deglutamation followed by HPLC-PCR(hv)-FD).¹⁶ A significantly lower value (47%) was observed, especially after the third freeze–thaw cycle. This instability was not observed when these freeze–thaw cycled samples were analyzed by the presented methodology.

Table 5 Freeze–thaw stability and stability of the derivatization product

Storage condition	Concentration of MTXPGtotal/nM		RSD (%) (n = 3)
	Nominal	Measured
Stability in autosampler (24 h)	10	10.0	13.4
	50	50.2	6.1
	250	250.2	5.8
Freeze–thaw stability (3 cycles)	10	9.4	24.4
	50	48.3	13.3
	250	242.3	5.9

---

Patient samples analyzed by the MTXPG_total assay and LC/MS/MS

Patient samples were analyzed by the MTXPG_total method (Fig. 3) and compared with our previously reported LC/MS/MS methodology (Fig. 4). The LC/MS/MS method was specific for each MTXPG, leading to quantification of individual MTXPGs which were added to provide MTXPG_total. Results obtained by the different methods correlated well (R² = 0.983) and therefore respond comparable to changes in intra-cellular levels of MTXPG_total within patient RBCs. The relationship between the methods could be described by the following equation: y = 1.32x + 5.3, where y is the response in nM measured by the reductive method and x is the response in nM measured by the LC-MS/MS method. Analysis of the equation reveals that the reductive procedure yields on average a 32% higher MTXPG_total value when compared to LC/MS/MS. It has been shown that another MTX metabolite, 7OH-MTX, can accumulate to a significant extent in bone marrow and certain tissues.²⁰ Since red blood cells are generally considered to be a biomarker for the (anti)folate status in the bone marrow,²¹ it was hypothesized that these elevated levels could be due to the presence of 7OH-MTX in RBCs. In order to check the specificity of the reductive method towards 7OH-MTX, RBCs were spiked with MTX and 7OH-MTX respectively. It was found that the reductive method was less sensitive towards 7OH-MTX and more importantly was able to distinguish between the 7OH-MTX and MTX (Fig. 5), eliminating the possibility of measuring elevated MTXPG_total levels due to the presence of 7OH-MTX in the patient samples. While the nature of the discrepancy between the reductive method and LC/MS/MS remains unknown, it is to be noted that interfering backgrounds have been observed when HPLC-FL was used in conjunction with the post-column photochemical derivatization. As a result, when comparing and reporting absolute MTXPG_total concentrations the analysis strategy should be taken into account, but despite the absolute value attained, the approach provides a clear indication of MTXPG_total that should be of clinical value.

Fig. 3 Chromatograms of a patient with a relative high concentration of MTXPG_total (black line) and a patient with lower MTXPG_total (blue line) are demonstrated. A worked up RBC blank is included for reference purposes (red line).

Fig. 4 Correlation between MTXPG_total concentrations in 10 patients obtained by LC/MS/MS and the presented reductive method.

Fig. 5 Separation of the 7OH-MTX and MTX derivatization products in a RBC matrix.

Conclusion

The measurement of RBC MTXPGs has potential to aid in optimization and individualization of MTX therapy in JIA and RA. Despite the clinical advantages of monitoring RBC MTXPGs this tool is currently not commonly used. Analytical methods that allow for the determination of intra-cellular MTXPGs are rather labor intensive and/or require the use of specialized equipment. Furthermore, the throughput of these methods is low and cost of analysis is high, thus making their implications in a clinical environment challenging. In this report a method is presented for rapid analysis of MTXPG_total status in human RBCs. The analytical procedure is a simple “one pot” pre-column reaction involving reagent addition and heating for 15 minutes. After centrifugation the supernatant is introduced into a conventional HPLC system equipped with a fluorescence detector, without the need for further workup. By performing the derivatization reaction pre-column, a time consuming (6–14 h) commonly used deglutamation procedure utilizing blank human plasma, becomes obsolete. Using the described procedure the total sample preparation time for a 50 sample run should not exceed 1–1.5 hours. The chromatographic run time per sample is 7 minutes using isocratic elution conditions. The rapid chromatographic procedure in combination with the uncomplicated sample preparation procedure should allow for same day sample determinations. Clinical results obtained by the reductive procedure correlated well with an earlier published LC/MS/MS method, however, obtained absolute RBC MTXPG_total concentrations were on average 30% higher, likely due to the enhanced selectivity in accord with the more elaborate and expensive instrumentation required for LC/MS/MS determinations.

References

1. L. Stamp, R. Roberts, M. Kennedy, M. Barclay, J. O'Donnell and P. Chapman, Biomed. Pharmacother., 2006, 60, 678–687.
2. A. V. Ramanan, P. Whitworth and E. M. Baildam, Arch. Dis. Child., 2003, 88, 197–200.
3. Arthritis Rheum., 2002, 46, 328–346.
4. B. N. Cronstein, Pharmacol. Rev., 2005, 57, 163–172.
5. T. Dervieux, D. Furst, D. O. Lein, R. Capps, K. Smith, M. Walsh and J. Kremer, Arthritis Rheum., 2004, 50, 2766–2774.
6. T. Dervieux, D. Orentas Lein, J. Marcelletti, K. Pischel, K. Smith, M. Walsh and R. Richerson, Clin. Chem. (Washington, D. C.), 2003, 49, 1632–1641.
7. T. Dervieux, D. Furst, D. O. Lein, R. Capps, K. Smith, J. Caldwell and J. Kremer, Ann. Rheum. Dis., 2005, 64, 1180–1185.
8. P. Angelis-Stoforidis, F. J. Vajda and N. Christophidis, Clin. Exp. Rheumatol., 1999, 17, 313–320.
9. J. Swierkot and J. Szechinski, Pharmacol. Rep., 2006, 58, 473–492.
10. J. A. Wessels, T. W. Huizinga and H. J. Guchelaar, Rheumatology (Oxford, U. K.), 2008, 47, 249–255.
11. C. J. Allegra, J. C. Drake, J. Jolivet and B. A. Chabner, Proc. Natl. Acad. Sci. U. S. A., 1985, 82, 4881–4885.
12. B. A. Chabner, C. J. Allegra, G. A. Curt, N. J. Clendeninn, J. Baram, S. Koizumi, J. C. Drake and J. Jolivet, J. Clin. Invest., 1985, 76, 907–912.
13. J. E. Baggott, W. H. Vaughn and B. B. Hudson, Biochem. J., 1986, 236, 193–200.
14. L. v. Haandel, M. L. Becker, J. S. Leeder, T. D. Williams and J. F. Stobaugh, Rapid Commun. Mass Spectrom., 2009, 23, 3693–3702.
15. C. Pfeiffer and J. Gregory, 3rd, Clin. Chem. (Washington, D. C.), 1996, 42, 1847–1854.
16. M. Hroch, J. Tukova, P. Dolezalova and J. Chladek, Biopharm. Drug Dispos., 2009, 30, 138–148.
17. J. Salamoun and J. Frantisek, J. Chromatogr., 1986, 378, 173–181.
18. J. Salamoun, M. Smrz, F. Kiss and A. Salamounova, J. Chromatogr., 1987, 419, 213–223.
19. W. M. Deen, P. F. Levy, J. Wei and R. D. Partridge, Anal. Biochem., 1981, 114, 355–361.
20. J. E. Baggott and S. L. Morgan, Arthritis Rheum., 2009, 60, 2257–2261.
21. Y. Huang, S. Khartulyari, M. E. Morales, A. Stanislawska-Sachadyn, J. M. V. Feldt, A. S. Whitehead and I. A. Blair, Rapid Commun. Mass Spectrom., 2008, 22, 2403–2412.

---