Biospecimen Integrity

Protein Oxidation as a Metric of Blood Plasma/Serum Integrity

Prolonged handling and storage of blood plasma and serum (P/S) samples above their freezing point of ‑30 °C can contribute to sample measurements that do not accurately reflect biological reality in vivo.  Within the scope of tools for assessing the integrity of P/S samples, nearly all candidate markers are founded on quantification of a nominal protein via a molecular recognition-based assay such as an ELISA.  As such, the indication for loss of specimen integrity lies in an apparent loss of the target protein beyond the normal human reference range.  In our view, this paradigm is less than desirable as the measurement does not directly reflect the molecular insult that has occurred.

Instead, we tap into spontaneous chemical oxidation of common P/S proteins as endogenous markers of storage history and overall specimen integrity. We have found that protein thiol and thioether groups on specific abundant P/S proteins are susceptible to different forms of oxidation at temperatures above ‑30 °C (Figure 1).  By using a simple LC-MS based technique for quantifying the relative degree of oxidation we are able to detect oxidative changes that take place over a matter of hours at room temperature and days at ‑20 °C using less than 1 microliter of P/S (Figure 2).  When implemented, such markers for the “early detection” of P/S sample degradation will help eliminate false biomarker discoveries in research programs that rely on the use of biobanked P/S samples. This is illustrated in Figure 3 below.

Figure 1: Charge deconvoluted electrospray ionization (ESI)-mass spectra of human albumin (panel a) and apolipoprotein A-I (panel b), illustrating how improper sample storage may lead to artifactual protein oxidation.  For albumin, ex vivo oxidation takes the form of S-cysteinylation and in apolipoprotein A-I it takes the form of methionine sulfoxidation. Reproduced with permission from Borges et al Mol Cell Proteomics 13(7): 1890-1899, 2014.


Figure 2:
Panel a) Increasing relative abundance of S-cysteinylated albumin in plasma over time at -80 °C, -20 °C, and room temperature (25 °C). Experimental details are provided in Borges et al Mol Cell Proteomics 13(7): 1890-1899, 2014. S-cysteinylated albumin hits an upper limit (plateau) because there are more albumin equivalents present in plasma/serum than there are free equivalents of cysteine (as cysteine and cystine). Once oxidative processes have forced albumin to consume all of the free cysteine equivalents in P/S, the fraction of albumin in the S-cysteinylated form stops rising.  Panel b) Increasing apolipoprotein A-I (ApoA-I) oxidation in plasma over time at -80 °C, -20 °C, and room temperature (25 °C). Samples are the same as those in panel a (albumin and ApoA-I are measured in the same assay). Total weighted ApoA-I oxidation can reach a theoretical maximum value of 1. (Panels a and b Reproduced with permission from Borges et al Mol Cell Proteomics 13(7): 1890-1899, 2014.)

By measuring the relative abundance of the oxidized (S-cysteinylated) form of albumin once, then intentionally driving it to its maximum value (see plateau phase in Figure 2) by incubating a P/S sample overnight at 37 °C, then measuring it again, we are able to take the difference in these measurements, called “ΔS-Cysteinylated Albumin”, to readily detect exposure of any P/S sample to the thawed state. We recently used this approach to detect a biospecimen integrity discrepancy between the control group and stage I cancer group of a nominally pristine serum sample set with an ideal paper trail (Figure 3)—a finding which ultimately prompted a novel disclosure that the -80 °C freezers in which the control set of specimens had been stored had lost power for 3-4 days during a natural disaster.


Figure 3: Ability of S-Cysteinylated Albumin (panel a) and enhanced ability of ΔS-Cysteinylated Albumin (ΔS-Cys Albumin; panel b) to detect sample integrity discrepancies in highly pedigreed serum samples collected, handled and stored under the same SOP. The yellow highlighted range in panel b indicates the theoretical/calculated range that should be observed in pristine samples based on the known ranges of albumin, free cysteine and free cystine in nominally healthy human P/S samples. Samples that lie below this range are either statistical outliers or (more likely) have experienced ex vivo oxidation from exposure to the thawed state that must have occurred prior to the first measurement of S-Cysteinylated Albumin (panel a). Panel (c) provides a “behind the scenes” glimpse at the second measurement of S-cysteinylated albumin and shows a lack of difference between the two sample sets with regard to the maximum degree to which albumin can be S-cysteinylated (oxidized) via ex vivo incubation for 18 hrs at 37 °C.

We are working to determine the rate law for the set of biochemical reactions involved in formation of S-cysteinylated albumin and modeling its application in actual P/S samples. Once successful, this will allow us to measure ΔS-Cysteinylated Albumin in any sample and project an estimated time frame over which the sample has been exposed to room temperature-equivalent conditions.

Blind Challenges for ΔS-Cysteinylated Albumin

Group-Wise Blind Challenge: Six unique, freshly collected plasma (n = 3) and serum (n = 3) samples were divided into ten sets containing the same six aliquots in each set. Each set of six samples was then either kept as a control at -80 °C or mistreated under one of several conditions (as a set; see Figure 4). An analyst was given the samples and told which samples belonged in which set, but not told anything about how any of the sets were treated (or how many of the sets were controls and how many were mistreated). The analyst’s goal was to determine which set(s) were kept as controls (at -80 °C) and which were mistreated in some way.  RESULT: Raw data for ΔS-Cysteinylated Albumin (ΔS‑Cys‑Albumin) are shown in Figure 4. Using ANOVA with Dunn’s posthoc test and alpha = 0.20, nine out of the ten sets described in Table 1 were identified correctly. It was not possible to detect exposure of samples to 4 °C for 8 hours (Group 7).  All controls and all other mistreatment conditions shown in Table 1 were identified correctly as either “controls” or “mistreated” sets. Notably, ΔS‑Cys‑Albumin was sensitive enough to detect exposure to conditions as mild as 25 °C (room temperature) for only 2 hours.

Individual Sample-Level Blind Challenge: Ten aliquots each of the 3 plasma and 3 serum samples were made and then randomized to provide a total of 60 samples. After randomization, each individual sample was then either kept as a control at -80 °C or mistreated under one of several conditions (see caption to Figure 5). An analyst was given the randomized samples with no information whatsoever besides a random code assigned to each sample and instructed to identify which samples had been mistreated in some way and which were controls. RESULT: Raw data for ΔS‑Cys‑Albumin are shown in Figure 5. A cutoff of 3 standard deviations below the mean for samples presumed to be controls was employed by the analyst to distinguish putatively mistreated samples from putative controls. All samples were correctly identified, on an individual basis, as being either controls or mistreated in some way (see caption to Figure 5 for details on mistreatment conditions).

Figure 4: Group-wise blind challenge results. Each group of six samples consisted of an aliquot of the same plasma (n = 3) and serum (n = 3) samples. All groups except Group 7 (mistreated by keeping at 4 °C for 8 hrs) were correctly identified by a blinded analyst as either “Control” or “Mistreated” sets. Notably, ΔS-Cys-Albumin was sensitive enough to detect exposure to conditions as mild as 25 °C (room temperature) for only 2 hrs (Group 4).


Figure 5: Individual sample-level blind challenge results. a) ΔS-Cysteinylated Albumin was measured in freshly collected matched plasma and serum from 30 cardiac patients to establish a conservative reference range for the marker in the general population. b) Ten aliquots each of 3 unique plasma samples and 3 unique serum samples were made. Samples were randomized and mistreated on an individual basis under one of the following conditions: -80 °C (Controls, n = 12); -20 °C for 60 days (n = 6); -20 °C for 90 days (n = 6); 4 °C for 7 days (n = 6); 4 °C for 14 days (n = 6); 25 °C for 24 hrs (n = 6); 25 °C for 48 hrs (n = 6); 25 °C for 72 hrs (n = 6); 25 °C for 7 days (n = 6). Each individual sample was correctly identified as either “Control” or “Mistreated” by a blinded analyst using the three standard deviations (3 SDs) cutoff shown. Notably, the cutoff and outcome was the same when the twelve samples with the highest values of ΔS-Cys Albumin in the blind challenge were employed as the “reference range”.