INTRODUCTION

The collection, separation and transfusion of red blood cells, platelets, whole plasma and fractionated plasma components are mainstays of our health care system. Each of these elements is essential for the preservation of life and the treatment of disease. Despite years of effort, suitable substitutes have yet to be developed.

In the early 1980s, it became clear to the medical community that these life-sustaining and essential therapeutic elements were transmitting life-threatening diseases (1). The primary causative agents of these diseases were identified as the human immunodeficiency virus (HIV) and hepatitis C virus. Because of the delay in development of suitable detection systems, these agents passed undetected into the blood supply. The fact that these potentially fatal diseases were transmitted by blood, and the lack of suitable blood substitutes, posed a tremendous challenge for members of the medical community. The resulting concerns raised by these events triggered a surge of research into methods to purify, reduce or eliminate infectious agents in the world's blood supply.

We are developing a medical device for the reduction of pathogens in platelet concentrates. The technology uses light and the photosensitizer riboflavin (RB) (vitamin B₂). The extent of our interest in RB as a pathogen reduction agent for blood stems from over 70 years of research literature that details vitamin B₂'s chemistry, toxicology and the in vitro and in vivo function in sensitizing the photochemistry of nucleic acids (2–4). Interest in the chemistry of this compound was based on its involvement in metabolic and nutritional functions as well as its behavior in individuals subjected to phototherapy (5–8). Before this research, RB had never been evaluated as a possible sensitizer for use in the ex vivo photochemical treatment of blood products. The presence of RB in our diets (9–11), the extensive knowledge of its photochemical properties and its well-behaved toxicological profile makes it an ideal candidate for a blood additive in this application (12,13). Previous studies have demonstrated the ability of this process to inactivate viruses including Human Immunodeficiency Virus (HIV), Porcine Parvovirus (PPV) and West Nile Virus (WNV) and bacteria including S. epidermidis and E. coli in platelet products while maintaining normal cell quality parameters suitable for transfusion (14).

One of the concerns about using any chemical agent as an additive to blood components for use in pathogen reduction technology (PRT) applications arises from the potential introduction of new chemical agents into the blood supply. These chemical agents may, on their own accord, raise concerns regarding increased toxicological risks that could, in certain circumstances, outweigh the risks associated with potential disease transmission of the blood component. The toxicological profile of RB is very well known and characterized. However, after exposure of RB to light, photochemical reactions can lead to the degradation of the molecule, yielding several photoproducts (15). These photoproducts result primarily from the decomposition of the ribityl side chain in the parent molecule (16,17). Several of these intermediates and breakdown products have been isolated and characterized (16–19).

Exposure of RB in aqueous solution to light leads to rapid photobleaching, measured at the absorption maximum of 447 nm (20). At alkaline pH, lumiflavin is a major breakdown product of RB (20). Under neutral and acidic conditions (the pH at which the pathogen reduction process is performed), lumichrome (LC) is the main breakdown product. Several intermediate by-products produced through metabolic or photochemical degradation of RB have also been identified. These include the 2′ keto-flavin (2KF) and 4′ keto-flavin (4KF) and formylmethyl flavin (FMF) (16). The isolation and characterization of these agents has been hindered by both the low levels present in isolated samples and the low sensitivity of methods of isolation and quantification available to the research community for study of these agents.

The ability to identify these agents in blood components as they naturally occur and the further ability to quantify or characterize them in samples before and after PRT treatment would permit direct examination of the potential impact of PRT on blood component chemistry and toxicology. Such studies, although not definitive, could be used to examine potential consequences in this regard in the blood transfusion setting. Hence, in support of clinical trials and toxicology studies involving this treatment regimen, we have developed a novel high-performance liquid chromatography (HPLC) method for the separation, identification and quantification of RB, LC and other photoproducts of RB in apheresis platelets both before and after PRT treatment. Typically, the measurement of flavins in blood components involves trichloroacetic acid (TCA) precipitation of proteins (21–34). This evolved from the earlier Warburg and Christian methods and was used with high salt (usually ammonium sulfate) both to dissociate flavocoenzymes (flavin mononucleotide [FMN] and flavin adenine dinucleotide [FAD]) from flavoenzymes and more recently to inactivate enzymes that could hydrolyze the flavocoenzymes or even catabolize the released RB. Subsequent extraction and concentration of the flavins was done with phenol or benzyl alcohol.

In this study we describe a sensitive and robust method for the accurate determination of both RB and LC in blood components without the need for protein precipitation or extraction. We also discuss the findings from direct examination of blood components before and after addition of RB and exposure to light as is proposed in the blood sterilization treatment protocol. These results are presented in terms of potential toxicological and medical implications for this blood safety initiative.

MATERIALS AND METHODS

Chemicals

HPLC grade methanol was purchased from Fisher Scientific (Pittsburgh, PA). RB and LC were purchased from Sigma–Aldrich (St. Louis, MO). RB was used without further purification. LC was purified as described below. Reagent grade water was prepared using a Barnstead E-Pure water purification unit (Dubuque, IA). Saline (0.9%) was purchased from B. Braun Medical Inc. (Irvine, CA).

HPLC apparatus and chromatographic conditions

The HPLC system consisted of an Agilent 1100 equipped with a quaternary pump (model G1311A), an autosampler (model G1329A), an autosampler thermostat (model G1329A), a thermostatted column compartment (model G1316A), a fluorescence detector (model G1314A) and a fraction collector (model G1364A). The fluorescence detector flow cell volume was 8 μL. The HPLC column used was a Zorbax® 80Å SB-CN, 4.6 × 250 mm, 5 micron. The precolumn was a Zorbax® 80Å SB-CN 4.6 × 12.5 mm, 5 micron. The nominal backpressure was 96 bar at the beginning of the run. The fluorescence detector settings were changed during each run. Initial fluorescence detector settings were maximized for detection of RB followed by maximum sensitivity for LC. The settings were: excitation 268 nm, emission 525 nm for the first 14.5 min, followed by excitation 260 nm, emission 470 nm. The small change in baseline upon wavelength change did not significantly affect the integration. A manual injector program was set up using four wash vials (reagent grade water) to wash the needle between runs and prevent carryover of RB and LC into the next run. The left- and right-column temperatures were set to 50°C. The autosampler temperature was set to 5°C. Flow rate was 1.0 mL/min. The injection volume was 10 μL. Run time was 25 min.

Reagent grade water and methanol were used as the mobile phase and a gradient program was set up (Table 1) to separate RB and LC from the other photoproducts. A six-point calibration curve was used with RB and LC concentrations ranging from 250 to 1500 nM. Calibration stock consisted of 50 μM RB and 50 μM LC in 0.9% saline, pH 5.0. All stock sample concentrations were confirmed by spectrophotometric analysis of RB and LC determined from extinction coefficients of the purified compounds. The stock was diluted to the appropriate calibration concentrations with 0.9% saline. An excellent straight-line fit was obtained for both RB and LC (r² ≥ 0.999 in both cases). The precolumn was replaced after every 30–60 samples.

HPLC-MS/MS analysis

Photoproducts of RB were isolated in 10 separate fractions using a fraction collector. Injection volume was maintained at 500 mL. A diode array detector was used during the separation with settings of stored range from 190 to 500 nm, peak width <0.05 min, slit 4 nm.

The following table contains the fraction collection start and stop times:

Isolated fractions were analyzed using the diode array detector to identify UV maxima in the individual peaks. Samples were then forwarded to Cardinal Health (Raleigh, NC) for HPLC-MS/MS analysis. They were protected from light and shipped at room temperature. For analysis, samples were transferred to dark glass vials at room temperature. Analysis of all samples was performed using a Cohesive Technologies 2300 HPLC system coupled to a Sciex API3000 triple quadrupole mass spectrometer. Extractions of each of the unknown solutions were performed on a Cohesive Cyclone-P 1 × 50 mm extraction column and on either a 250 × 4.6 mm cyano or a 150 × 4.6 mm cyano analytical column for analytical separations. All mass spectrometric analysis was done using electrospray negative ionization. In cases where more signal was needed, samples were concentrated using a Turbo-vap at 30°C with nitrogen gas. The instrument was protected from light during the evaporation. For all injections, the following MS/MS transitions were scanned, based on analysis of standards: RB, 375.1 > 255.1 amu; KF, 373.1 > 241.3 amu; LC, 240.9 > 198.1 amu; FMF, 282.9 > 240.2 amu.

Apheresis platelet preparation

Single donor platelets (a minimum of 270 mL) were collected by an accredited blood bank facility using a Gambro BCT Inc. TRIMA® Automated Blood Component Collection System. The platelet product was held between 2 and 30 h in the TRIMA collection bag before subsequent processing. The bag containing the platelets was connected to the pathogen reduction illumination–storage bag (ELP™) using a Terumo® Sterile Tubing Welder. After sterile connection, 250 ± 5 mL of platelet product was transferred into the illuminator bag, which contained 28 mL of RB solution (500 μM). The transfer tubing was sealed using a Sebra® hand-held, radio frequency tubing sealer. After connection, the two bags were separated and the original collection bag was discarded. Each final platelet product to be treated contained 1260 × 10³ to 1690 × 10³ platelets/μL suspended in approximately 278 mL of 90% autologous plasma in a 1 liter citrate-plasticized, polyvinyl chloride ELP bag with an illumination surface area of 347 cm²/side.

Plasma derived from buffy coats

Buffy coat platelets were prepared from whole blood that was stored at room temperature for 16–24 h after collection. The platelets were spun down at 16 100 RCF for 3 min and diluted 1:50 with 0.9% saline before analysis.

The PRT process

A complete description of the procedures used for this process has been described previously (Ruane et al.). Briefly, the platelet concentrates were treated separately with 6.2 J/mL of light. The lamp phosphor possesses an output range from 265 to 370 nm. The platelet product was placed in a product chamber where mixing (on a motorized platen) and exposure to light took place. Total illumination ranged from 8 to 10 min. Dedicated fans were used to cool each lamp chamber and product chamber. Product temperatures during illumination ranged from 22°C to 24°C.

Purification of LC

The average purity of reagent LC (Aldrich PN 10,321-7) was found to be approximately 68–85% as determined by negative ion mass spectroscopy and HPLC. LC was purified before use as an analytical standard as follows: reagent grade LC (250 mg) was added to a 250 mL amber volumetric flask and diluted to the mark with reagent grade water. The solution was sonicated (VWR® sonicator model 150HT) at 80°C for 2 h and then vacuum filtered while hot, using a 0.45 μm filter. The filter cake, which consisted of purified LC, was dried at 105°C for 2 h (the impurity is soluble in water and LC is relatively insoluble). The LC purity after this process was found to be greater than 99%. LC purity was confirmed using mass spectrometry, proton nuclear magnetic resonance (¹H NMR), HPLC and elemental analysis. An API3000 triple quadrupole mass spectrometer using negative ionization mode observed the parent M−1 ion 241 m/z concurrent with the LC nominal molecular weight of 242 g/mol in the purified sample. Elemental analysis—found: C 59.04, H 3.81, N 22.95; required: C 59.50, H 4.16, N 23.13. ¹H NMR (D₂O), 2.39, s, 3H; 2.41, s, 3H; 7.61, s, 1H; 7.78, s, 1H.

Preparation of FMF reference compound

FMF was synthesized according to the method of Fall and Petering and characterized using NMR and mass spectrometry (35).

Calibration standard preparation

Analytical standards were prepared by diluting a stock solution of 50 μM RB and 50 μM LC in saline. The stock solution was prepared by adding LC (12.1 mg) to a 1 liter amber volumetric flask and adding approximately 950 mL of pH 5.0 saline. The solution was sonicated using a VWR sonicator (model 150HT) at 70–100°C for approximately 2 h until all LC was in solution. RB (18.8 mg) was then added and the solution allowed to cool while stirring overnight. It was then filled to the mark with pH 5.0 saline. Calibration standards were prepared fresh every 24 h.

HPLC validation for quantitative analysis of RB and LC

Validation of the method for RB and LC, the principal photoproduct of RB was carried out during an 8 day period. Both these agents were present in sufficient quantities in blood products to validate the methodology. Other components (FMF, 2KF, 4KF) were isolated by bulk HPLC and characterized by mass spectroscopy and UV–visible (UV-VIS) spectroscopy. Results from these studies were used to determine concentrations of these agents in blood components, but validation of the methodology for quantitative determination of these components was not performed. Linearity of the calibration curve was 0.999 (r²) or better for all 8 days of validation. For five replicates of both RB and LC at three concentrations (25, 37.5 and 50 μM), the average relative standard deviations were RB = 0.64% and LC = 0.76%. For these 30 samples, the average percent recovery for RB was 97% whereas the average percent recovery for LC was 102%. The limit of quantitation for RB (undiluted) was 16 nM and for LC was 60 nM. Samples were diluted 1:50 with 0.9% saline pH 5.0 before analysis. The range of the method for RB is 0.016–1.500 μM and for LC 0.06–1.500 μM. A typical analysis involved two replicates of each sample. No loss of peak shape, peak height or retention time was noted for runs using as many as 15 samples (30 samples when replicates are accounted for).

During the 8 day validation, all retention times for RB samples were within 7.8 ± 0.2 min and all retention times for LC samples were within 15.4 ± 0.2 min. The average resolution was 10.1 and the average tailing was 0.34, passing the Food and Drug Administration validation requirements of ≥2.0 resolution and ≤2.0 tailing (36).

Reinjection precision analysis was performed to determine whether an analytical run could be restarted in case of instrument failure. Analysis was performed on a calibration curve and on 30 samples, 10 each at three different concentrations. The calibration curve was followed by 30 samples, including blanks and check standards every 10 samples, simulating a routine analysis. After completion of the sequence, the entire sequence was restarted. The difference in measured RB and LC concentrations from the first run and second run was less than ±1.5%.

Interference

Several analytes were run to test for interference including FAD, FMN, lumiflavin and FMF. None of these analytes interfered with the analysis of RB and LC. Results from sample runs with these agents are included in Fig. 1.

RB and photoproduct analysis in platelet concentrates

Ten platelet concentrates spiked with 50 μM RB were treated with 6.2 J/mL light. The concentrations of RB, LC and other photoproducts were measured before and after spiking with RB, after illumination and after 5 days of storage on a Helmer shaker, respectively (22–24°C, shaker speed of 72 ± 5 cpm). All samples were stored in the dark until analysis, whereupon a 4 mL sample was drawn from each platelet product. Each sample was centrifuged at 3000 RCF for 10 min to remove the cellular components. The supernatant was respun at 16 100 RCF for 3 min and the supernatant diluted 1:50 with 0.9% saline. The samples were then placed in amber HPLC vials for analysis. Analysis included characterization of peaks associated with the 2KF, 4KF and FMF as well. Quantification of these latter samples was conducted on the basis of extinction coefficients for the parent molecule and HPLC analysis. Identification of peaks associated with these agents was performed through analysis of isolated fractions by mass spectroscopy–HPLC and UV-VIS absorption characteristics (Table 2). In the case of FMF, direct comparison with a de novo synthesized standard was possible (Table 3).

Additional samples taken from normal donors were analyzed for the presence of RB and RB catabolites or photoproducts. These samples were assayed directly without any RB addition to determine the levels of each component present in a natural state in blood components.

RESULTS

DISCUSSION

The methodology described represents a sensitive HPLC method for the quantification of RB, LC, 2KF, 4KF and FMF in blood components that does not require a protein precipitation or organic solvent extraction step. The method simply removes the cellular components by centrifugation followed by dilution with saline and analysis using a reverse phase cyano-column (Zorbax® 80Å SB-CN, 4.6 × 250 mm, 5 micron). We have found that the low concentration of protein after dilution does not interfere with the assay.

Typically, C18 reversed phase columns are used to analyze RB and related compounds in aqueous solutions (15,18–20,22,24–29,31–34, 36). However, we found that the use of this stationary phase was not satisfactory and separation of photoproducts from RB after photolysis was not feasible. A reverse phase cyano-column was found to be optimal in resolving compounds closely eluting to RB (Fig. 2). Other columns were used and separation was not optimal (Agilent (Zorbax) Columns: SB-C18, XDB-C18, 300 Extend C18, 300 SB-C8, XDB-C8, SB-C8, 300SB-C3, 300SB-CN, Carbohydrate, SB-Aq, Eclipse AAA, NH2, Diol and GF-250).

The measurement of vitamin B₂ in blood components is not new (15,16,19–24,32). Typically, FAD in the blood is converted to the more stable FMN. Free RB and FMN are then measured after their extraction into organic solvents. These methods also require the acidic (usually TCA) precipitation of blood proteins before analysis. This has one significant drawback. RB and related compounds tend to coprecipitate with the protein making accurate determinations difficult. In most cases, although agents such as FMF, 2KF and 4KF have been previously identified, their low levels in naturally occurring blood products and the low sensitivity inherent in previous methodologies prohibited their routine identification and quantification in blood components. The increased sensitivity of the method described in this study permits this direct analysis.

The method described in this article has been developed to support a PRT known as MIRASOL PRT. The technology involves the inactivation of pathogens in platelet concentrates using 50 μM RB and light. During the course of this study, we measured the amount of RB, LC, FMF, 2KF and 4KF in 10 apheresis platelet products before and after treatment. The platelet concentrates were also stored at 22–24°C on a Helmer shaker for 5 days, whereupon the amounts of each of these agents were reanalyzed. We were able to accurately determine the amount of RB converted to photoproducts during the process. A total of 78.9% RB remained immediately after irradiation. This amount increased upon 5 days of storage to 82.0%, presumably because of photoproducts (not LC) reverting back to RB during this period. The amount of LC also increased upon storage from 6.2% to 7.5%. The total mass balance of both compounds on Day 5 was 89.5%. The remaining 10.5% is made up of other photoproducts that have been positively identified and quantified through mass spectroscopic and other analytical methods. Figure 2 shows an overlay of HPLC chromatograms of RB in a diluted platelet product (1:50), before illumination (a), after illumination (b) with the PRT process and Day 5 after treatment (c). The photoproducts β and γ have been identified as 2KF and 4KF, respectively. Their UV-VIS absorbance spectra (both are nearly identical to that of RB) and mass spectra are consistent with this hypothesis. Results from HPLC, UV-VIS analysis and mass spectra data are depicted in Table 2. FMF (δ) has been positively identified by comparison with an authentic sample. Results from this analysis are depicted in Table 3. Quantification of these photoproducts has been conducted using the extinction coefficients for the parent molecule and are included in Table 4.

Hustad has reported a value of 6.9 nM (2.7–42.5 nM) for RB concentration in plasma (n = 63) (37). In a later report, Hustad reported the average RB concentration in the plasma of 124 senior citizens as 15.3 nM (5.4–28.4 nM) (25). Capo-chichi reported in 2000 that the median concentration of RB in 10 infants and 10 adolescents to be 20.9 nM (12.7–53.4) and 18.5 nM (8.2–57.8), respectively (25). We now report for n = 30, an average value of 23.9 nM (8.6–79.6 nM) (Fig. 3). The higher mean value observed in this study might be a reflection of the fact that no additional protein precipitation or extraction step was required in this procedure resulting in higher retention of starting levels of RB, or might be due to partial leakage of erythrocytes still present in the buffy coat platelet products.

Photolysis of RB during the MIRASOL process results in the formation of four photoproducts: 2KF, 4KF, FMF and LC. These four photoproducts were found to be present in apheresis platelets that had not undergone any photochemical treatment, although at a much lower concentration (Table 5, Fig. 4). It is also important to note in this context that the concentrations measured in the platelet products will be diluted by a factor of 16- to 20-fold upon infusion of the products into a patient's blood stream, thus lowering the difference between concentrations determined in these products and the levels naturally circulating in blood. The demonstration of the existence of these agents in naturally occurring blood products suggests that the introduction of a RB-based PRT process will not introduce new agents into the blood supply, which are not already present to some extent. The consequence of increased levels of these agents in blood products resulting from this process is being evaluated in separate toxicology studies, which include short-term and long-term exposure to each of the agents described and characterized in this study (38–45). The presence of these agents in our blood, the ubiquitous nature of RB exposure, its presence in our diets and our ability to metabolize it and manage its inherent photochemistry suggests a low risk profile for this product. These features in combination with the pathogen reduction capacity of this system have the potential to greatly enhance the safety of blood products presently offered in routine clinical practice.