Volumetric sampling using shaped filter paper for collection of a defined amount of blood
We describe here a simple and precise approach to the volumetric sampling of DBSs that demonstrates all the advantages of DBS collection while overcoming the sampling issues of hematocrit and homogeneity. Our approach involves the absorption of a liquid capillary blood sample onto a precisely shaped strip of Whatman filter paper, wherein the volume absorbed is controlled by the properties and size of the strip. This approach offers simpler sample collection than traditional DBS card collection and the entire collected sample is analyzed, which allows better integration into sample processing procedures.
Several types and sizes of filter paper were analyzed for suitability for the sampling device, including Whatman 903, CF12, and 415, Munktell TFN, and Ahlstrom 226 in 7.5 mm × 7.5 mm and a 5 mm × 10 mm sizes. We found no significant differences between the filter papers tested by size in the filling times and volume (data not shown). The expression of 22 protein biomarkers across 4 strips collected from the same person were assessed in Whatman 903 5 mm × 10 mm strips, Asanté™ 60 mm2, and Asanté™ 100 mm2 strips (1819-100, 1820-100, Sedia Biosciences, Beaverton, USA). The %CV were 12%, 19% and 16%, respectively (Supplemental Table S6). Thus, the 5 mm × 10 mm size was selected due to the smaller volume requirements compared to the other sizes tested and the Whatman 903 paper was selected due to the low %CV, accessibility, and common use among other CE/IVD available collection devices. The developed volumetric sampling device consists of a 30 mm × 5 mm Whatman 903 filter paper strip secured in a holder that exposes a 10 mm × 5 mm section that collects a 30 µL volume of blood (Fig. 1). The device is dipped into blood samples and absorption occurs by wicking. The strip is defined as full when no visible white areas are observed in the 10mmx5mm exposed area. The filling time for each strip was assessed in 20 strips and found to take 20.70 ± 3.55 s to fill (Fig. 2A). Filling time was dependent on the individual subject and the area of the strip that was immersed in the blood. The strip holder ensures that the tips of each strip do not touch each other or their surroundings during collection and drying, which prevents blood contamination and transfer. The filled strips were allowed to dry for at least 2 h and dried strips were stored at room temperature sealed in a biohazard bag with a desiccant pack.
Variation in the fill time and volume of blood absorbed onto the volumetric sampling device. (A) complete filling time for 20 strips. (B) amount of blood absorbed by 20 strips. Solid red line: represents the average. Dotted red lines: represent the standard deviation.
The exact volume of blood absorbed by the device was assessed by weighing the blood absorbed in 20 strips. Each strip was found to absorb 29.52 ± 2.48 µL of blood (Fig. 2B). The coefficient of variation in volume of blood absorbed was 8.39%, which is much less than the 30% variation in the volume of blood collected using a traditional DBS card and punch approach39 and is comparable to the 5% and 8.1% variation found in similar volumetric sampling devices31,32. We then assessed whether it was possible for the device to be overfilled. Ten strips were allowed to absorb blood until visibly saturated, then the strips were exposed to 0, 1, 2, 3, or 4 additional drops of blood. The volume of blood absorbed by the device was then assessed by weighing as described above. There was no significant difference in the volume of blood absorbed in strips that were exposed to additional drops of blood compared with strips not exposed to additional blood (p = 0.22), indicating that overfilling the strip did not occur in this situation (data not shown).
One of the advantages of DBSs is the ability to obtain high quality samples from remote locations and/or developing countries and from those with limited access to clinics, such as military patients or patients with limited mobility such as the elderly. Our volumetric sampling device allows for remote, in-home blood collection and sample shipment by mail, which facilitates DBS expansion into these populations. To investigate the accuracy of remote collection from naïve users, we assessed operator variability in the volume of blood sampled. The accuracy of blood volume collected and ease-of-use of our device was assessed in samples collected by 16 operators with no experience in the collection of blood samples and who had not previously used this or any DBS collection device. The operators were given written instructions for the use and collection of a DBS with the device. Each operator attempted to fill 5 strips. Of the 80 samples collected, 63 (78.75%) were properly filled with blood according to the instructions. These results are in good agreement with the sampling errors associated with other established techniques of DBS collection using naïve collectors40. Operator feedback indicated that the instructions were relatively clear and easy to follow.
To verify the volumetric sample device developed here overcame the hematocrit issues seen in traditional DBS cards, we analyzed the expression of 200 proteins as well as hemoglobin concentration in 21 matched serum, DBS, and dried serum samples (DSS) samples. The hemoglobin concentration was 32.00 ± 1.57 µg/mL, 148.70 ± 11.41 mg/mL, and 24.30 ± 1.30 µg/mL in serum, DBS, and DSSs, respectively, which is consistent with published levels41. The average correlation of the expression of 200 proteins across DBS/DSS, DBS/Serum, DSS/Serum, Hemoglobin/DBS, Hemoglobin/DSS, and Hemoglobin/Serum groups in matched serum, BDS, and DSS sample with hemoglobin concentration were 0.63, 0.56, 0.67, − 0.08, − 0.07, and − 0.10, respectively. These data show that the correlation of hemoglobin concentration with the expression of each of the 200 proteins assessed in DBS, DSS, and serum was negligible. Indicating that hematocrit as evidenced by the hemoglobin concentration is not an issue with determining the quantitative protein expression using our volumetric sampling device.
We have demonstrated that this volumetric sampling device is capable of collecting an accurate, reproducible volume of blood that can be used for quantitative analysis, independently of hematocrit and homogeneity issues. This device can also be used to provide quantitative analysis from samples collected from populations not normally included in clinical study. While this device addresses some of the major disadvantages of DBSs for clinical quantitative analysis, a thorough comparison of the DBS method against venous blood draw is needed. Preferably, this comparison would be conducted across a wide range of protein analytes before DBSs may be deemed comparable or not to liquid serum and/or plasma.
Validation and optimization of volumetric DBS device procedure for quantitative proteomics analysis
A thorough method validation requires a strict protocol covering all relevant assay steps. Thus, a comprehensive workflow for preparing DBSs collected with our volumetric sampling device and analysis with a quantitative proteomic immunoassay for future use in a clinical setting is presented, including (1) the collection of blood, (2) drying and storage, (3) elution of DBSs from the strips, (4) optimization of the procedure, and (5) analyses of DBS eluates.
One of the advantages of DBSs is their stability in storage, enabling the shipment of the dried samples at ambient temperature. We investigated the stability of 14 relatively abundant serum proteins (listed in Supplemental Table S1) in DBSs using an accelerated stability test protocol. A total of 25 DBSs were collected with our volumetric sampling device and stored at − 80 °C, 4 °C, ambient temperature (AT), 37 °C, 50 °C, and 60 °C for 0, 1, 4, 8, and 12 weeks. The expression levels of the 14 analytes were quantitatively assessed and compared with the expression levels from a corresponding freshly collected DBS (Fig. 3). Average protein retention over 12 weeks was 100%, 99%, 95%, 62%, 48%, and 40% for DBSs stored at − 80 °C, 4 °C, AT, 37 °C, 50 °C, and 60 °C, respectively. Our results show that a small volume of blood (30 µL) can be used to quantitatively assess 14 proteins and these samples can be stored at ambient temperature for up to 3 months with no significant loss of protein stability.
Average stability of 14 proteins in DBS samples over 3 months at varying temperatures. The average retention in the expression levels of 14 relatively serum abundant proteins in DBS samples stored at − 80 °C, 4 °C, ambient temperature (AT), 37 °C, 50 °C, and 60 °C for 1 week, 4 weeks, 8 weeks, and 12 weeks were compared with the expression levels from a corresponding freshly collected DBS sample. Data are expressed as the mean ± SEM.
Once the blood is collected and stored, the next step in the method is elution of the DBS from the sampling device. For many clinical applications, accurate measurement of multiple proteins is needed. The limitation of a small sample volume requires efficient use of all DBS material collected. Similarly, the extraction efficiency of the proteins from the densely dried sample is paramount. Various proteins and molecules diffuse at different rates, more or less efficiently, from the collection device and may cause a variable and thus less accurate recovery rate42,43,44. Thus, optimization of the elution method used is necessary for the end assay application. To optimize and validate the elution of proteins from the entire DBS collected with our volumetric sampling device, we analyzed 4 different elution buffers: 1 × PBS (EB1), 1 × PBS containing 0.1% Tween-20 (EB2), 1 × PBS containing 10% casein and 50% BSA (EB3), and 1 × PBS containing 10% casein, 50% BSA, and 0.1% Tween-20 (EB4). The completely filled 5 mm × 10 mm exposed section of the strip was cut using straight, clean surgical scissors and placed into a 1.5 mL conical tube using forceps cleaned with alcohol. The elution buffer was added to the conical tube at a 1:10 ratio and eluted for 4 h at room temperature on a shaker, vortexing for 10 s every 30 min. Next, the liquid was transferred to a clean 1.5 mL conical tube leaving the strip behind and centrifuged at 14,000 rpm for 10 min. The supernatant was collected into a clean 1.5 mL conical tube. We then quantitatively assessed 40 proteins (listed in Supplemental Table S2) in 3 DBSs each. These 40 proteins were chosen partially based on the wide variation in expression levels in serum reported in the literature. Ten of the 40 total proteins assessed were not detectable using any elution buffer in these particular samples. The 30 remaining proteins were detectable in at least 1 elution buffer. Of the remaining 30 proteins, 22, 27, 23, and 21 were detectable using EB1, EB2, EB3, and EB4, respectively (Fig. 4A). The concentrations of 5 of those proteins, ADAM12, B7-H3, CD48, Kallikrein 5 and Pentraxin 3 with elution using all 4 buffers are also shown (Fig. 4B). Both the concentrations and variation (shown as standard error across samples) were different for each protein with each elution buffer. Overall, for the 30 measurable proteins, EB2 had the least variance in protein concentration across samples tested compared to the other elution buffers. Our results also indicated that EB2 allowed for the quantitative measurement of the most proteins. Thus, this elution buffer was used in all following assays.
Optimization of Elution buffer composition for protein extraction in DBS samples. 4 different elution buffers: 1 × PBS (EB1), 1 × PBS containing 0.1% Tween-20 (EB2), 1 × PBS containing 10% casein and 50% BSA (EB3), and 1 × PBS containing 10% casein, 50% BSA, and 0.1% Tween-20 (EB4) were used to elute 3 DBS samples each. (A) The number and percentage of 40 proteins that were quantitatively assessed and measurable with each elution buffer. (B) Concentrations and variance as evidenced by the standard error of 5 proteins compared across the 4 different elution buffers.
We further optimized the sample dilution of the DBS eluates for use on the quantitative immunoassay platform using the provided assay sample diluent. A total of 15 DBSs were collected and eluted using EB2. Five different dilutions of the eluate were tested in triplicate: 5 ×, 10 ×, 20 ×, 50 × and 100 ×, with quantitative assessment of the same 40 proteins as above. Results show that the total number of measurable proteins was 37, 33, 29, 25, and 21 for the 5 ×, 10 ×, 20 ×, 50 × and 100 × dilutions, respectively. In choosing the optimal dilution, we took into consideration the following: background signals across the multi-plex array (which were increased in the 5 × dilution group, data not shown), the amount of protein expression retained as the eluate dilution increased (Fig. 5), and the total sample volume restriction, which would make quantification of a large number of proteins untenable at low eluate dilutions. We concluded that a 20 × eluate dilution yielded the best signal to background ratio. Thus, a 20 × eluate dilution in EB2 was used for all further assays.
Optimization of sample dilution of DBS eluates for use on a quantitative immunoassay platform. 5 different eluate dilutions: 5 ×, 10 ×, 20 ×, 50 × and 100 × were used to quantitatively assess 40 proteins. The % expression loss compared with the 5 × dilution is shown for a subset of 10 proteins.
To investigate the accuracy and reproducibility of the quantitative proteomics analyses (intra-assay variability) we collected a DBS using our volumetric sampling device from a single individual at a single time-point and preformed the entire validated method indicated above 10 separate times. We correlated protein expression across the replicates and found the mean r2 correlation value between replicates was 0.860 with a range of 0.621–0.975 (Fig. 6). Average within-subject % CV across the quantitative assessment of the measurable proteins (29 out of 40) was 23.77% across 10 tests (Table 1). We also assessed variability across individual samples based on collection order. Ten individual DBSs were collected from a single individual and numbered based on the order they were collected. A pooled sample of the first 5 DBSs collected and of the last 5 DBSs collected were also assessed. The levels of protein expression for each sample were compared to each other and to the pooled samples. We found no difference in protein expression based on collection order or between independent and pooled samples (Fig. 6). These results indicate that collection order does not compromise reproducibility, and that any number of samples can be collected and pooled to generate enough sample volume to test any number of proteins, with subject comfort and physical bleeding capacity being the only limiting factors.
Accuracy and reproducibility of the quantitative proteomics analyses A. correlated protein expression across sample (SA) replicates 1 through 10 and pooled samples (SA1-5 and SA6-10).
Analysis of 1000 protein analytes in DBSs with an established quantitative multi-plex immunoassay
To establish confidence in and implementation of DBSs in clinical applications, validation of a large number of proteins that span all levels of expression, using an approach that has been previously clinically established, is necessary. Although we are not validating the quantitative multi-plex used here clinically, the potential to do so for a subset of identified proteins of interest is available in the future.
First, we tested 21 matched serum, DBS, and DSSs using quantitative multi-plex immunoassay to determine the expression profile of 200 proteins. Of these 200 proteins, 18 were undetectable (i.e., below the limit of detection (LOD) in at least 2/3 of all samples per group) in all sample types (G-CSF, GM-CSF, I-309, IFNγ, IL-1α, IL-2, IL-4, IL-6, IL-7, IL-8, IL-10, IL-11, IL-12p70, IL-17, TNFβ, BMP-4, GRO, and IL-2 Rγ). Another 9 were undetectable in DBS (IL-5, IL-13, MCSF, TNFα, bFGF, BMP-7, FGF-7, IGF-1, and XEDAR), 6 were undetectable in DSS (IGF-1, XEDAR, IL-1β, GITR, IL-21R, and NRG1-β1), and 18 were undetectable in serum (IL-5, IL-13, MCSF, TNFα, IL-1β, NRG1-β1, IL-1ra, MCP-1, MIP-1β, TGFβ3, VEGF, CTACK, IL-9, IL-29, I-TAC, LIF, CD30, and Contactin-2). As shown in Fig. 7, there is a moderate correlation among the proteins assessed across the three sample types. DSS/serum had the highest correlation, with 69% of the measurable proteins having a Pearson’s r value greater than 0.5. The percentage of measurable proteins having a Pearson’s r value above 0.5 for DBS/DSS and DBS/serum are 60% and 55%, respectively. However, among all matched sample group comparisons, the correlations ranged wildly, with a number of individual proteins that were not correlated (Pearson’s r value < 0.3) and a number of individual proteins that were highly correlated (Pearson’s r value > 0.7). The concentrations and Passing–Bablok regressions for each of the measurable proteins is shown in Supplemental Table S4 and Supplemental Table S5, respectively. These results indicate that a sweeping comparison of sample types for expression of all proteins is likely to be inaccurate and that each sample type should be assessed independently for clinical suitability, or a larger validation of each individual sample type on its correlation with another should be conducted for each independent protein.
Example Passing–Bablock regression of the expression of 200 proteins across groups in matched serum, DBS, and DSS samples for MCP-4.
Next, we tested unmatched serum and DBSs in parallel using a quantitative multi-plex immunoassay to determine the expression profile of 1000 proteins45. The list of the 1000 proteins analyzed can be found in Supplemental Table S3 and the validation linearity and recovery data for the assay can be found in the manufacturer’s protocol. Two hundred normal serum samples and 72 normal DBSs were collected. Of the 1000 proteins analyzed, 410 were measurable (i.e., above the limit of detection (LOD)) in at least 2/3 of all the DBSs assayed, whereas 491 were measurable in at least 2/3 of all the serum samples assayed (Fig. 8A). A similar number of targets, 283 and 272 targets in DBSs and serum samples, respectively, were below the LOD in at least 2/3 of all samples analyzed. 318 and 226 targets in DBSs and serum samples, respectively, did not clearly fall in either condition (measurable or below the LOD) in at least 2/3 of all samples, making it difficult to clearly define the measurability of those targets. The distribution of concentrations of all the measurable proteins in both serum and DBSs is shown in Fig. 8B,C. To date, there are over 100 protein biomarkers approved by the US FDA for clinical use in serum/plasma46. In total, 40 of the FDA-approved protein biomarkers were present in the tested 1000 protein dataset. We tested whether these proteins are also measurable in DBSs. Of the 40 FDA-approved markers tested, 18 were measurable in at least 2/3 of all DBSs tested, whereas 28 were measurable in at least 2/3 of all serum samples tested. Expression levels for these 18 and 28 targets, respectively, were reasonably well correlated with normal serum values reported in the literature despite the differences in sample type (in the instance of DBSs), individuals tested, and quantification methods (R2 value of 0.81 and 0.64 for DBSs and serum samples, respectively) (Fig. 8D). Validation in single-plex assays for clinical use is often required for biomarkers identified in large scale, multi-plex analyses. Similarly, there are a number of well-known biomarkers, including cancer markers CA72-4 and CA125, that are often used independently in single-plex assays. Therefore, we tested 5 matched DBS and serum samples on single-plex ELISAs for CA72-4 and CA125 (Table 2). The results showed good correlation between serum and DBS for these 2 proteins (adjusted R-squared = 0.7002 and 0.9717 for CA125 and CA72-4, respectively) suggesting that for some markers, DBS may be capable of being used as an alternative sample type for clinical analyses for those patients in which serum collection is contraindicated using the serum values. However, many of the 1000 proteins tested were found to not correlate well between sample types, indicating that DBS should be treated independently from serum when assessing clinical suitability. Similarly, when determining if the clinical serum values are suitable for a DBS, each protein should be validated for correlation independently.
Assessment of measurable proteins in normal DBS and serum samples in a quantitative multi-plex immunoassay. (A) Summary of the number of 1000 proteins that are below the LOD, within the detectable range, above the LOD, or that didn’t fall completely in either of the aforementioned categories in at least 2/3 of all samples assessed in DBS and serum, respectively. (B) Distribution of the concentration of measurable proteins in DBS and serum samples by violin plot. (C) Scatterplot of the distribution of the concentration of measurable proteins in DBS and serum samples. (D) Correlation of the 18 DBS and 28 Serum measurable FDA-approved protein biomarkers in at least 2/3 of the 72 DBS and 200 serum samples assessed with each other and reported serum literature values.
