Sample Processing Considerations for Protein Stability Studies of Low Concentration Biofluid Samples using Differential Scanning Calorimetry

Abstract

Background:

The analysis of biofluid samples with low protein content (e.g., urine or saliva) can be challenging for downstream analysis methods with limited sensitivity. To circumvent this problem, sample processing methods are employed to increase the protein concentration in analyzed samples. However, for some techniques, like differential scanning calorimetry (DSC) that characterizes thermally-induced unfolding of biomolecules, sample processing must not affect native protein structure and stability.

Methods:

We evaluated centrifugal concentration and stirred cell ultrafiltration, two common methods of sample concentration characterized by a low risk of protein denaturation, with the goal of establishing a protocol for DSC analysis of low concentration biospecimens.

Results:

Our studies indicate that both methods can affect protein stability assessed by DSC and, even after optimization of several parameters, the obtained DSC profile (thermogram) suggested that sample processing affects the structure or intermolecular interactions of component proteins contributing to altered thermal stability detectable by DSC. We also found a relationship between changes in thermograms and low protein concentration, indicating that diluting biospecimens to concentrations below 0.1 mg/mL can perturb the intermolecular environment and affect the structure of proteins present in the solution.

Conclusion:

Dilution of samples below 0.1 mg/mL, as well as concentration of samples with low protein content, resulted in affected thermogram shapes suggesting changes in protein stability. This should be taken into account when concentrating dilute samples or employing techniques that lower the protein concentration (e.g., fractionation), when downstream applications include techniques, such as DSC, that require the preservation of native protein forms.

Graphical Abstract

1. INTRODUCTION

Differential scanning calorimetry (DSC) is a biophysical technique that monitors the heat change (excess specific heat capacity) in a fluid sample as it is heated and provides a profile (thermogram) associated with the thermal unfolding of biomolecules. Fifteen years ago, DSC was applied for the first time for the analysis of clinical specimens, such as blood plasma, and was found to have utility for the diagnosis and monitoring of patients [1, 2]. Since then, several groups have observed dramatic changes in plasma thermograms for multiple diseases, including cancer (cervical, ovarian, endometrial, colorectal, gastric, lung, melanoma, brain) [3-10], autoimmune (systemic lupus erythematosus, rheumatoid arthritis, psoriasis) [11, 12], and other diseases (diabetes, Lyme disease, chronic obstructive pulmonary disease) [12-15]. Moreover, thermogram changes allow for the distinction of healthy subjects from affected patients and the separation of patients with different diagnoses [4, 16, 17]. The unique feature of thermograms is that they reflect the overall biomolecular makeup of the sample (e.g., protein concentration, structure, modification, interactions) at the time of collection. Therefore, DSC is distinct from the majority of clinically available biomarkers, which are measures of a single analyte, as opposed to the complex biological milieu resulting in a pathological state.

To date, mainly plasma [2, 18, 19], serum [1, 14], cerebrospinal fluid (CSF) [20], and tissue extracts [9] have been used for DSC analysis. These samples are characterized by high protein concentrations, which allow for their easy analysis using DSC. However, urine is among other biofluids where DSC might find potential utility. Over the last 30 years, much research has been focused on the evaluation of proteomic [21, 22] and metabolomic [23, 24] profiles of urine in healthy and disease states that could be applied for the diagnosis of patients with both renal and systemic diseases [22, 25-27]. By using different approaches, researchers were able to identify from 115 [28] to more than 1800 unique proteins [26, 29], confirming that urine could be a desirable material for diagnostic purposes. The advantage of urine over other biofluids, like plasma and CSF, is an easy collection that is completely non-invasive, allowing for repeated measurements over a long period of time. However, the disadvantage of urine is its low protein concentration that is estimated to be no higher than 0.1 mg/mL for healthy individuals, increasing up to 25 mg/mL for patients with kidney damage, such as with lupus nephritis (LN). The low protein concentration of urine could limit the use of DSC for its analysis because of the high noise-to-signal ratio that could result in uninterpretable thermograms. This limitation could be overcome by concentrating the samples, an approach widely used in proteomic studies. However, the main concern is maintaining the proper structure of the component proteins during this process, an issue that is inconsequential for the analysis of non-native protein samples by bottom-up proteomics but is crucial for obtaining reliable thermograms of the unfolding of native protein forms. Therefore, in the current manuscript, we evaluated multiple methods for protein concentration, where our focus was on evaluating concentration protocols suitable for the analysis of biospecimens with low protein concentrations, such as urine samples, using DSC. However, this manuscript could also be applicable to other low protein concentration biofluids and other analytical techniques that characterize native protein structure and stability. We excluded concentration techniques that present a high risk of protein denaturation or selective removal of some protein subtypes, such as heating or freeze-drying of samples [30, 31], or precipitation of protein from solution using organic solvents/dyes, and focused our attention on two commonly used methods in proteomic studies, centrifugal concentration [32] and stirred cell ultrafiltration [33]. Both methods rely on the use of semi-permeable membranes that allows for the separation of the original solution into two fractions. The separation of solution components uses centrifugal force (centrifugal devices) or inert gas pressure combined with gentle stirring of the sample (stirred cell ultrafiltration). All molecules above the membrane’s molecular weight cut-off (MWCO) are retained inside the device (stirred cell or centrifugal concentrators), whereas water and all components below the cut-off pass through the membrane and are collected in the waste compartment. To optimize concentration conditions, we evaluated the effect of concentration time, and additionally, for centrifugal concentrators, also the effect of different membranes, centrifugation conditions as well as sample starting concentration. Additionally, we tested the impact of centrifugation stress on samples with low protein concentrations (0.1-0.5 mg/mL). The effects of centrifugation stress and the different concentration methods on thermally-induced protein unfolding were then assessed by DSC. Results were compared with unprocessed samples to evaluate whether the concentration affects the structure or intermolecular interactions of component proteins contributing to the thermal stability profile detectable by DSC.

2. MATERIALS AND METHODS

2.1. Sample Preparation

Lyophilized plasma (MilliporeSigma, Burlington, MA, USA) was rehydrated by adding 1 mL of deionized water, and then incubated for 1 h at room temperature, with inversion mixing every 15 min, followed by rotary mixing overnight at 4°C. Any further dilutions and dialysis were performed using phosphate buffer (1.7 mM KH₂PO₄, 8.3 mM K₂HPO₄, 150 mM NaCl, 15 mM sodium citrate, pH 7.5). SnakeSkin dialysis tubing (ThermoScientific, Waltham, MA, USA, 3.5K MWCO) was used for dialysis of plasma samples, with buffer changes after 3h, 4h, 4h and a final overnight dialysis period, for a total dialysis time of 25 h at 4°C. Samples were recovered from dialysis and filtered through 0.45 μm centrifuge tube filters (Pall, Port Washington, NY, USA) to remove particulates. The final dialysis buffer was also filtered (0.2 μm; Pall) and used for all sample dilutions and as a reference solution for DSC studies. During concentration studies, several concentrating devices were used, including Vivaspin 2 and Vivaspin 15R, both with 2K MWCO hydrosart membranes (Sartorius, Bohemia, NY, USA), Vivaspin Turbo with a 3K MWCO polyethersulfone (PES) membrane (Sartorius), and Microsep Advance Centrifugal Devices with a 3K MWCO PES membrane (Pall). A summary of all experiments is presented in Figure 1. During concentration, all samples were mixed between centrifugation intervals by gently pipetting up and down. Additionally, during the concentration of 50 mL of healthy pseudourine samples, fresh aliquots were added in a volume equal to the volume that had passed through the concentrator membrane until the entire 50 mL volume had been loaded into the concentrator.

Figure 1.

Summary of all parameters evaluated in the experiments.

Regenerated cellulose 3K MWCO membrane discs were used with a stirred cell ultrafiltration device (Amicon, MilliporeSigma). Discs were first soaked in deionized water for 1 h at room temperature, with the water changed three times during this time. After assembly, phosphate buffer (50 mL) was used to wash the membrane. Next, 50 mL of sample was loaded into the device, and the sample was concentrated for 40–60 min using nitrogen gas at a pressure of 30 psi.

2.2. Collection of DSC Data

Data were collected using a Nano DSC Autosampler System (TA Instruments, New Castle, DE, USA), maintained, and serviced according to the manufacturer's procedures. Instrument performance was assessed using lysozyme (a biological standard) and was within the manufacturer's specifications. All samples were loaded into the instrument autosampler and thermostated at 4°C until analysis. Thermograms were recorded from 20°C to 110°C at a scan rate of 1°C/min with a pre-scan equilibrium step of 15 min at 20°C. Duplicate scans were collected for each sample. Also, each experiment was performed to ensure that DSC analysis was completed within a seven-day window after initial plasma preparation. During analysis, we collected and examined buffer scans at the beginning and end of each sample set, and after single or consecutive sample scans, to determine acceptable reproducibility.

DSC data were post-processed using Origin version 7.0 (OriginLab Corporation, Northampton, MA, USA). Raw sample scans were corrected for the instrumental baseline by subtraction of a suitable buffer scan. Thermograms were normalized for total protein concentration determined colorimetrically using the bicinchoninic acid protein assay kit and microplate procedure from Pierce (Rockford, IL, USA), with absorbance readings taken with a Tecan Spark plate reader (Tecan U.S., Research Triangle Park, NC, USA). Following normalization, thermograms were corrected for non-zero baselines by application of a linear baseline fit. All 43 final thermograms were plotted as excess specific heat capacity (cal/°C.g) versus temperature (°C) and represent the average of duplicate DSC scans, except for 2 low concentration samples where only one DSC scan is presented because the baseline subtraction of the duplicate DSC scan was unreliable as a result of excessive baseline curvature coupled with a high noise-to-signal ratio of the scan.

3. RESULTS

To evaluate the effect of protein concentration on thermograms, we used human plasma that was diluted to a concentration of 0.05 mg/mL (mimicking the protein concentration of healthy urine) and 10 mg/mL (mimicking the protein concentration of LN urine), further referred to as pseudourine (healthy and LN, respectively). Although it is not common to use diluted plasma as a urine mimic, we chose plasma for multiple reasons. First, the high starting protein concentration of plasma (~60–80 mg/mL) allowed us to test a wide range of samples with different final protein concentrations. We were able to compare the results obtained for highly diluted samples, as well as diluted samples that were subsequently concentrated and/or subjected to centrifugal stress, with reference thermograms of plasma samples analyzed at higher concentrations. Second, several high abundant plasma proteins are found in urine, typically at low levels, but higher concentrations of albumin and globulins are detected in patients with kidney dysfunction when impairment of the renal filtration system results in plasma proteins leaking through damaged nephrons. Third, the conditions for the processing of plasma samples for DSC analysis have been investigated by several groups resulting in established protocols demonstrating the reproducibility of the results [4, 5, 9, 34]. Our previous results indicate that plasma dilutions from ~0.4 to 4 mg/mL do not affect the thermogram shape [34], indicating that there is no discernible change in the thermal stability profile of plasma proteins. Thus, in the current study, we focused most of our attention on assessing the effect of sample processing on thermograms of samples with lower protein concentrations, applicable to the preparation of urine specimens for DSC analysis. We used different concentrator units with either hydrosart or PES membranes, and loading volumes of 2 and 15 mL using both fixed-angle and swing-out rotors, or a stirred cell ultrafiltration device. We compared thermograms of samples that had undergone concentration with unprocessed samples. This allowed us to compare concentrated samples with plasma samples that were unaffected by the concentration process. In addition, we investigated the effect on thermograms of different starting concentrations of plasma, different concentration intervals, and the total time of concentration, as well as the effect of centrifugation without concentration. A summary of all evaluated parameters is shown in Figure 1. For experiments evaluating different types/numbers of centrifugal concentrators or centrifuge type/rotor (Figures 2-4, Suppl. Figure 1), except for unprocessed healthy pseudourine samples (~0.05 mg/mL), all samples were diluted to ~0.5–0.7 mg/mL for DSC analysis, set to the lowest concentration attained after each concentration experiment. In the remaining experiments, for evaluating the length/interval of centrifugal concentration (Suppl. Figure 2), ultrafiltration (Figure 5), and centrifugation versus centrifugal concentration of low protein samples (Figure 6), samples were analyzed directly by DSC (without diluting to a common concentration) and compared with the original unprocessed sample as well as a higher concentration unprocessed plasma sample (0.8 – 1.6 mg/mL) that was within the range previously shown to not affect thermogram shape [34].

3.1. Comparison of Hydrosart and PES Membrane Centrifugal Concentrators

In the first experiment, we compared two concentrator units with hydrosart (Vivaspin 2, Santorius, 2K MWCO) and PES (Microsep Advance Centrifugal Devices, Pall, 3K MWCO) membranes. Both units had a volume of 2 mL. One milliliter of LN and 50 mL of healthy pseudourine were concentrated using several cycles of 20–25 min centrifugation at 4 °C, 2000x g (swing-out rotor, Eppendorf 5810R). The total time of concentration of the healthy pseudourine sample (psHU) was 11 h 10 min (Pall) and 12 h (Vivaspin 2), and 1 h (Pall) and 40 min (Vivaspin 2) for the LN pseudourine sample (psLNU; Figure 2A).

Figure 2. DSC thermograms of psLNU and psHU samples concentrated using small-volume centrifugal concentrators with hydrosart and PES membranes.

(A). Schematic representation of the experiment. (B). Thermograms of a psLNU sample before and after concentration using centrifugal concentrators with PES and hydrosart membranes. The unprocessed psLNU thermogram is used as a reference. (C). Thermograms of a psHU sample before and after concentration using centrifugal concentrators with PES and hydrosart membranes. The unprocessed psLNU thermogram is used as a reference. Note: Except for the unprocessed psHU sample, samples were diluted to a common concentration for DSC analysis, set to the lowest concentration attained after each concentration experiment. This allowed us to compare concentrated samples with plasma samples that were unaffected by the concentration process. The concentration of each sample analyzed by DSC is shown in parentheses.

We found that the membrane type had a minimal effect on thermograms of psLNU samples as they all had the same shape as the unprocessed psLNU reference profile (Figure 2B). As expected, psHU, because of the low protein concentration, had a high noise-to-signal ratio (Figure 2C); however, we observed an altered thermogram shape compared with the unprocessed plasma sample (Figure 2C). This observation was consistent among all performed experiments suggesting that diluting the plasma samples to obtain a low protein concentration mimicking that of healthy urine (~0.05 mg/mL) impacts the stability of the proteins in the obtained solution. Moreover, we found that proteins present in psHU were affected by the concentration process as thermograms of concentrated psHU had an altered shape, with a shift of the main transition from 64°C to 70°C as well as the appearance of an additional small transition ~65°C. Additionally, we noticed that the type of membrane affected the results. More substantial thermogram changes were observed for psHU concentrated using centrifugal concentrators with a hydrosart membrane compared to concentrators with a PES membrane (Figure 2C).

3.2. Evaluation of the Effect of Centrifugation on Thermograms of Concentrated Samples

Since long centrifugation times have been suspected to be the leading cause of the thermogram changes of psHU samples, we used concentrators with higher starting volumes, which allowed us to decrease the concentration time from ~12 h to 7 h 20 min (Vivaspin 15R, 2K MWCO hydrosart membrane) and 4 h (Vivaspin Turbo, 3K MWCO PES membrane; Figure 3A). However, as can be seen in Figure 3B, this did not result in an appreciable improvement in thermogram shape closer to that of the unprocessed plasma sample. In addition, similarly to the previous experiment, we noticed that changes in the thermogram shape were more pronounced for samples concentrated using hydrosart membranes, which could result from the longer centrifugation times needed for concentration of samples that was almost double when compared with PES membrane concentrators.

Figure 3. Effect of shorter centrifugation time on thermograms of psHU samples concentrated using large-volume centrifugal concentrators with hydrosart and PES membranes.

(A). Schematic representation of the experiment with large-volume concentrators (1 unit per sample). (B). Thermograms of a psHU sample before and after concentration using large-volume centrifugal concentrators with PES and hydrosart membranes. The unprocessed plasma thermogram is used as a reference. Note: Except for the unprocessed psHU sample, samples were diluted to a common concentration for DSC analysis, set to the lowest concentration attained after each concentration experiment. This allowed us to compare concentrated samples with plasma samples that were unaffected by the concentration process. The concentration of each sample analyzed by DSC is shown in parentheses.

To further decrease processing time, we increased the number of units used per sample from one to two units, and also evaluated different types of rotors and centrifugation speeds (Figure 4A), trying to reach the maximum rpm settings recommended by the manufacturer when using a particular centrifugal concentrator with a particular type of rotor (swing-out and fixed-angle). When concentrating the sample using the Beckman Allegra 25R centrifuge with a fixed-angle rotor using Vivaspin 15R (2K MWCO hydrosart membrane concentrator), it still took more than 5 h to concentrate 50 mL of psHU to ~2 mL (Figure 4B). In contrast, for samples processed using the Vivaspin Turbo concentrator with a 3K MWCO PES membrane, we were able to limit the centrifugation time to 4 h 10 min when using the Beckman Allegra 25R centrifuge with a fixed-angle rotor, and 1 h 40 min when using the Eppendorf 5810R or Beckman Allegra 25R centrifuges with swing-out rotors (Figure 4C). Nevertheless, all obtained thermograms showed a right-shifted main transition, and those changes were more prominent with longer centrifugation times (Figure 3A). We obtained similar results using concentrators with PES and hydrosart membranes for concentration times <7 h; however, the Vivaspin Turbo concentrator with a PES membrane allowed us to substantially reduce the time of concentration. Based on this observation, we used this concentrator for all further experiments.

Figure 4. Effect of different centrifugation conditions on thermograms of psHU samples concentrated using large-volume centrifugal concentrators with hydrosart and PES membranes.

(A). Schematic representation of the experiment in which psHU samples were concentrated using different rotors, different centrifugation speeds, and large-volume concentrators with PES and hydrosart membranes (2 units per sample). (B). Thermograms of a psHU sample before and after concentration using large-volume concentrators with a hydrosart membrane and a fixed-angle rotor. The unprocessed plasma thermogram is used as a reference. (C). Thermograms of a psHU sample concentrated before and after using large-volume concentrators with a PES membrane using different rotors (fixed-angle and swing-out) and different centrifugation speeds. The unprocessed plasma thermogram is used as a reference. Note: Except for the unprocessed psHU sample, samples were diluted to a common concentration for DSC analysis, set to the lowest concentration attained after each concentration experiment. This allowed us to compare concentrated samples with plasma samples that were unaffected by the concentration process. The concentration of each sample analyzed by DSC is shown in parentheses.

Interestingly, further reduction of the centrifugation time to ~1 h by increasing the number of units used per sample to four (Suppl. Figure 1A) resulted in a thermogram of the concentrated sample that resembled the unprocessed psHU thermogram, with a less pronounced transition ~65°C and high-temperature shoulders ~75°C than psHU samples with longer concentration times (Suppl. Figure 1B). However, as mentioned previously, both thermograms for the unprocessed and 1 h concentrated psHU samples differed from those obtained for the unprocessed plasma sample, illustrating the importance of the starting protein concentration of the sample.

3.3. Effect of Different Centrifugation Intervals on Thermograms of Concentrated Samples

Our evaluation of the total concentration time of psHU samples ranging from 12 h to 1 h revealed a consistently altered thermogram shape compared with concentrated psLNU samples. We hypothesized that changes in thermogram shape could result from stress related to the centrifugal forces during prolonged centrifugation times. Thus, we evaluated a total centrifugation time < 1 h utilizing short intervals (4 or 20 min) of centrifugation instead of continuous centrifugation for 30-40 min (Suppl. Figure 2A). Moreover, in this experiment, no fresh aliquots were added to the concentrator to exclude the effect of repeated cycles of sample concentration/dilution. We hypothesized that a longer total centrifugation time (40 min) might result in greater centrifugal stress and a greater alteration of the thermogram shape than a shorter total centrifugation time (20 min). Moreover, longer centrifugation intervals (20 min) might result in the greater local concentration of the sample than shorter intervals (4 min), which could confer a “protective” effect and a less affected thermogram shape. DSC analysis revealed that 1 × 20 min concentration resulted in the least affected thermogram compared with the unprocessed psHU sample. Thermograms of the other conditions were similar (Suppl. Figure 2B), and overall, there was no consistent pattern associated with shorter/longer intervals or shorter/longer total centrifugation times.

3.4. Stirred Cell Ultrafiltration

The lack of satisfactory results using centrifugal concentrators prompted us to evaluate an alternative membrane-based method of concentration using a stirred cell ultrafiltration device (Amicon), which allowed us to exclude the potentially adverse effect of centrifugal stress (Figure 5A). As expected, the unprocessed psHU sample was characterized by a high noise-to-signal ratio that resulted in undefined features of the thermogram compared with the unprocessed plasma sample (Figure 5B), similar to previous experiments with centrifugal concentrators. In addition, thermograms of stirred cell concentrated psHU samples resembled those obtained using centrifugal concentrators, characterized by a right-shifted main thermogram transition from ~64°C to 70°C and a shoulder ~65°C. These results indicated that concentration via ultrafiltration instead of centrifugation did not ameliorate the impact on the thermogram features.

Figure 5. DSC thermograms of psHU samples concentrated using a stirred cell ultrafiltration device.

(A). Schematic representation of the experiment. (B). Thermograms of psHU samples before and after concentration using a stirred cell ultrafiltration device. The unprocessed plasma thermogram is used as a reference. Note: The concentration of each sample analyzed by DSC is shown in parentheses.

3.5. Investigating the Effect of Starting Protein Concentration of Samples on Thermograms

Our results indicated that all evaluated methods of sample concentration resulted in thermogram features of psHU samples that differed from the thermograms obtained for unprocessed plasma and psLNU samples. However, we observed that simply diluting plasma to concentrations corresponding to healthy human urine already affected the thermogram. Given that short concentration times that did not exceed ~1 h resulted in thermograms of concentrated samples that were similar to the thermogram for psHU, our results suggested that a more detailed examination of initial sample concentration and centrifugation/concentration regimens on the thermogram would be of interest. We decided to test the effect of low protein concentration on thermograms to identify critical concentrations where thermogram features are affected. For this purpose, we analyzed samples with concentrations of 1, 0.5, 0.2, and 0.1 mg/mL. As shown in Figure 6A, although lower protein concentration increases the noise-to-signal ratio, the overall thermogram shape is unaffected. However, the noise-to-signal ratio impacted overall signal strength in the 0.1 mg/mL sample for which the thermogram amplitude was visibly lower than the other concentrations. These results suggest that the critical concentration of samples that impact the thermogram shape is below 0.1 mg/mL, as we have observed for psHU samples (~0.05 mg/mL).

Figure 6. The impact of initial protein concentration and centrifugation on thermograms.

(A). Thermograms of samples with different initial total protein concentrations (0.1–1.6 mg/mL). (B). Schematic representation of the experiment comparing the effect of centrifugation and centrifugal concentration of samples with starting protein concentrations of 0.1, 0.2, and 0.5 mg/mL. Thermograms of 0.5 mg/mL (C), 0.2 mg/mL (D) and 0.1 mg/mL (E) samples before and after centrifugation or centrifugal concentration using centrifugal concentrators with a PES membrane. The unprocessed plasma thermogram is used as a reference. Note: The concentration of each sample analyzed by DSC is shown in parentheses.

3.6. Impact of Centrifugation on Low Protein Concentration Thermograms

Since we observed differences in the shape of the thermogram for unprocessed pseudourine samples and that the thermogram was further affected by centrifugal concentration, our next step was to investigate whether stress conditions, such as centrifugation or combined centrifugation and concentration, would negatively impact the thermogram shape for samples with slightly higher starting concentrations (Figure 6B). To our surprise, total centrifugation times of 20 or 40 min or centrifugal concentration of up to 40 min did not have any significant impact on thermogram shapes of the 0.5 mg/mL (Figure 6C) sample. For 0.1 mg/mL (Figure 6E) and 0.2 mg/mL (Figure 6D) samples, we observed a modest effect on the main thermogram transitions ~65°C (slightly right-shifted and lower amplitude) and ~70°C (variation in the amplitude) for centrifuged samples but that centrifugation combined with concentration was “protective” and no impact on the thermogram was observed. These results indicated that for samples with a total protein concentration of at least 0.1 mg/mL, centrifugation alone can have some impact on protein stability in the lower concentration range (0.2 mg/mL or lower) but that centrifugation/concentration provides a protective effect that does not substantially impact protein stability as assessed by the thermogram shape.

4. DISCUSSION

Samples with high protein concentration are required in a wide variety of applications involving the functional or structural characterization of samples. Among those methods is an array of biophysical techniques, including DSC. The crucial aspect of DSC is that it requires the protein present in the sample to be in its native form, so protein unfolding during sample heating can be monitored and recorded. Since this method also requires samples with relatively high concentration, when analyzing specimens characterized by low protein concentration (e.g., urine or saliva) or samples whose protein concentration has been decreased because of processing (e.g., chromatography fractions), samples may need to be concentrated before analysis.

Several different techniques have been established to increase protein concentration [35]. One class of such methods is based on the precipitation of proteins from solutions using organic solvents [36, 37] or dyes [38]. Unfortunately, those methods are not easily adaptable to the analysis of large sample volumes, can result in protein denaturation, or lead to selective removal of some types of proteins (e.g., glycoproteins) [39]. A similar issue is observed for methods based on heating [30] or freeze-drying [31] of samples as the changes in temperature can also lead to denaturation of proteins, and thus obtained samples are not suitable for DSC analysis. Other commonly used techniques are based on ultrafiltration using either centrifugal or stirred cell concentrators. Those methods are easily adaptable for large sample volumes, are widely used (e.g., preparation of urine samples for mass spectrometry analysis), and the conditions of sample concentration minimize the risk of sample denaturation during the process. Therefore, in the current manuscript, we evaluated the utility of ultrafiltration-based techniques for the concentration of samples for DSC analysis with the goal of establishing a protocol that could be routinely used for the analysis of biospecimens with low protein concentrations.

In order to optimize the choice of concentration devices, we tested different types of concentrators, two types of membranes, devices of various sizes, and several centrifugation conditions (Figure 1). For testing, we used plasma samples diluted to ~0.05 mg/mL (healthy pseudourine; psHU), which allowed us to compare the thermogram of the concentrated sample with the thermogram of plasma that did not undergo concentration. We then prepared the samples based on our established DSC protocol (dialysis for normalization of buffer conditions) and set the concentration of all samples to a common concentration (~0.5–1 mg/mL) for direct comparison of thermograms. To our surprise, we found that all evaluated conditions of sample concentration using the centrifugal concentrators resulted in thermogram features of psHU samples that differed from the thermogram obtained for unprocessed plasma. Switching to a stirred cell ultrafiltration device did not improve the outcome, as thermograms of psHU samples concentrated under such conditions were characterized by similarly altered thermograms that differed from unprocessed plasma. However, we observed that simply diluting plasma to concentrations corresponding to healthy human urine resulted in an alteration of the thermogram shape. Given that short concentration times that did not exceed ~1 h resulted in thermograms of concentrated samples that were similar to the thermogram for psHU (with both unconcentrated and concentrated psHU thermograms differing from the thermogram for unprocessed plasma), our results suggested that there may be a critical protein concentration below which the solution structure and/or interactions of proteins change, thus affecting thermogram shapes. It is important to note that the critical concentration could differ between samples, especially those originating from biofluids containing a complex mixture of proteins/peptides, as there are many sample-specific factors, such as protein-protein interactions and posttranslational modifications, which could impact potentially deleterious effects of low concentration conditions on the stability of individual proteins.

In the next set of experiments, we investigated the impact of initial sample concentration and centrifugation/concentration regimens on the thermogram shapes. We tested concentrations of 0.1, 0.2, 0.5, and 1 mg/mL and found that the obtained thermogram shapes of unprocessed 0.1, 0.2, 0.5, and 1 mg/mL samples were similar to unprocessed plasma, with lower concentrations exhibiting higher noise-to-signal ratio. Importantly, the high noise for samples with low protein concentration significantly impacted DSC analysis because of challenging baseline selection. After centrifugation, we observed that the 0.5 mg/mL sample did not differ substantially from unprocessed plasma. However, we found a small negative impact of centrifugation on the samples with 0.1 and 0.2 mg/mL protein concentrations. In both cases, centrifugation caused small shifts in the main thermogram transition ~65°C and changes in the amplitude of both transitions (~65°C and ~70°C). In contrast, the same centrifugation conditions used during sample concentration did not impact the thermogram shapes, suggesting a protective effect against centrifugal stress of increasing the protein concentration in samples.

Although our studies focused on the effect of sample processing for DSC analysis, our observations are important for any technique where the native protein solution form needs to be maintained. The results should also be considered for studies where the initial protein concentration is decreased through sample processing, such as sample fractionation, especially when substantial differences are observed between analyzed samples. One such example is a study combining immunoaffinity chromatography and DSC to identify disease-induced changes in the human blood plasma proteome [40]. The sample processing protocol involved dilution of the samples during removal of human serum albumin (HSA) using immunoaffinity chromatography, followed by sample concentration using centrifugal concentrators. The authors noticed that although some of the analyzed samples still contained HSA, the corresponding thermograms were “missing” the transition assigned to HSA unfolding and hypothesized that HSA unfolding was masked by the unfolding of other plasma proteins. However, it can not be excluded that extensive dilution and concentration of samples affected protein structures or interactions, and that the HSA-associated transition was temperature-shifted in a similar way as we observed in our studies. It is therefore important to give careful consideration to the processing of low concentration biofluid samples for downstream protein stability studies.

CONCLUSION

The results of our analysis indicate that plasma samples with a total protein concentration below 0.2 mg/mL are sensitive to centrifugation stress that affects protein stability, as determined by DSC. These effects can be circumvented for 0.1 and 0.2 mg/mL samples by increasing the protein concentration during centrifugal concentration. This protective effect of increasing protein concentration is not observed for samples with a starting concentration below 0.1 mg/mL. Using both centrifugal concentrators and stirred cell ultrafiltration, the concentration of samples with a starting protein concentration similar to healthy urine (~0.05 mg/mL) resulted in affected thermogram shapes, suggesting changes in protein thermal stability during sample processing. Although our studies focused on the use of samples for DSC analysis, changes in protein structure can impact multiple downstream applications requiring preservation of the native protein structure. Studies using biofluids with a high protein content that can be analyzed directly or those that do not involve analysis of native proteins may not face these challenges. However, for biofluid samples with a low protein content where sample concentration is used before analysis or where sample handling involves a decrease in the protein concentration through fractionation or other methods, protein structure or intermolecular interactions may be affected by these processes. This is especially crucial when comparing results of the DSC analysis of samples with distinct starting concentrations, such as urine samples with low protein content obtained from healthy volunteers and urine samples with higher protein concentrations obtained from patients with kidney damage, as the observed differences could be a combination of both the impact of the disease as well as the sample concentration process. Therefore, handling protocols for samples with low protein concentrations need to be evaluated for possible perturbation of protein stability before embarking on studies that involve the characterization of native proteins.

LIST OF ABBREVIATIONS

CSF

Cerebrospinal Fluid

DSC

Differential Scanning Calorimetry

HSA

Human Serum Albumin

LN

Lupus Nephritis

PES

Polyethersulfone

psHU

Healthy Pseudourine

psLN

Lupus Nephritis Pseudourine