Human serum albumin (HSA) was purchased from Sigma Aldrich (St. Louis, MO). Two different preparations of HSA with differing advertised purity were received as lyophilized powder. They were fatty acid and globulin free, ≥99% pure HSA (product number: A3872, lot number: SLBD740V) and ≥96% pure HSA (product number: A1653, lot number: SLBG2676V). The former was advertised as purified from the latter, but not necessarily the same batch. According to the supplier the 3% greater purity of the HSA₉₉ sample is due to the absence of slight amounts of FA and globulins purified away from HSA₉₆. Hereafter, the 99%-pure and 96%-pure preparations of HSA are referred to as HSA₉₉ and HSA₉₆, respectively. The latter is the well-known HSA fraction V isolated by the method of Cohn et al. [30] . Although fraction V yields a large relative percentage of HSA, the preparation has been shown to be contaminated by ~2% globulins, primarily α₁-globulins such as orosomucoid and long chain FA present at a Mole/Mole ratio of FA/HSA < 1. A typical value of this ratio = 0.51 was reported [2] . Although an independent analysis of the specific contaminants in our HSA₉₆ preparation was not performed, the Certificate of Analysis of HSA₉₆ for our lot provided by the supplier reported not more than 2% of protein contaminant (as expected for fraction V). Thus, the majority of the remaining contaminant is nominally attributed to FA. Still the overall composition of the 3% contaminant of HSA₉₆ is not precisely known. To convey this fact, henceforth, the 3% impurity of HSA₉₆ is referred to as the “FA- and globulin-like contaminant” (FA/G-LC), with the belief that by mass the FA/G-LC is roughly comprised of 2% globulins and not more than 1% FA. In analogy several authors reported FA-free and FA-containing HSA species that have been referred to as defatted and un-defatted HSA [1] [15] [18] .

Standard buffer solutions contained 10 mM potassium phosphate and 150 mM NaCl, pH = 7.4. Total ionic concentrations of buffers were verified by electrical conductivity measurements. After preparation and prior to use, buffer solutions were stored at 4˚C. All solutions and buffers were prepared with Nanopure deionized water. All chemicals and reagents were molecular biology grade or higher.

Stock solutions of HSA were prepared at 1.0 mM by dissolving 0.067 g of HSA (67 kDa) in 1 mL buffer. Generally, reconstituted protein solutions were stored at 4˚C for at least 24 hrs before use. In the case of HSA₉₆, samples that were somewhat unstable under storage were examined within 48 hrs after preparation. For DSC melting experiments, a portion of the HSA stock solution was diluted in buffer to a final concentration of 1.5 - 2.0 mg/mL.

Biotin was attached to multiple sites on HSA by targeting the primary amines of lysine residues. Of the 59 lysine residues in the primary structure it is believed that nearly 20 of these are likely to be sufficiently exposed, and therefore accessible for covalent attachment of Biotin in the 3-D structure of HSA [26] . Biotin was attached to HSA using the EZ-Link Sulfo-NHS-Biotin kit (product number 21217 from Thermal Fisher Scientific) according to the supplier’s instructions. For the attachment reactions, a 10 mM stock solution of Biotin was prepared by dissolving 3.2 mg Biotin in 500 µL of ddH₂O. A solution containing a 1:1 molar ratio of HSA:Biotin was prepared by adding 3.02 µL Biotin stock solution per mL of a HSA solution at 2 mg/mL. In this manner HSA was incubated with various amounts of Biotin at a variety of molar ratios, HSA:Biotin, up to 1:50. For these reactions appropriate amounts of the Biotin stock solutions were added to HSA in the presence of the Biotin linking reagent and incubated for four hours at 20˚C. Generally, HSA samples prepared with different levels of attached Biotin were stored at 4˚C for at least 24 hours before removal of unattached free biotin. In this way Biotin was attached to both HSA₉₉ and HSA₉₆ preparations at a variety of HSA:Biotin ratios from 1:1 to 1:50.

Biotin was covalently attached to HSA targeting the disulfides of cysteine residues. Presumably, under neutral conditions, only one reduced sulfhydryl is available for biotin attachment at cysteine-34 of the HSA structure [2] , [31] . In targeting the single reduced sulfhydryl at cysteine 34 of the HSA primary structure in these reactions, biotin was assumed to attach to only a single site on HSA. In attachment reactions HSA was incubated at 21˚C for four hours in the presence of increasing amounts of attachment reagent according to instructions provided by the supplier. For these reactions the EZ-Link Maleimide-PEG2-Biotin kit (product number 21902 from Thermal Fisher Scientific) was employed according to the supplier’s instructions. Sample preparations were incubated at increasing molar ratios of HSA:Biotin, from 1:1 to 1:10. Attachment reactions were performed in 4 mL at a concentration of HSA = 1.6 mg/mL in the presence of increasing amounts of attachment reagent. In this way, Biotin was attached to both HSA₉₉ and HSA₉₆ preparations.

When attachment reactions were complete, free (unattached) Biotin was removed using a Zebra spin column (product #A9892 from Thermal Fisher Scientific). In this procedure, 1.5 mL of each attachment reaction solution (total volume 3 mL) was added directly to a spin column, and the sample was washed three times with 2 mL buffer. The washed sample was then retrieved and pooled.

Protein concentrations were determined at several steps. 1) Prior to performing attachment reactions and; 2) After application of the Biotin removal kit, at each attachment ratio, prior to measuring DSC thermograms. Protein concentrations were determined using the BCA Protein Assay Kit (product #23225, Thermal Fisher Scientific).

Samples were analyzed by electrophoresis on SDS polyacrylamide gels prior to, and after measurement of thermograms (not shown). Approximately 5 ug of protein sample was suspended in gel sample buffer (25 mM Tris, 192 mM glycine, 0.1% SDS, 20 mM DTT pH = 8.3) and loaded onto a BioRad Any KD™ Mini- PROTEAN^®15 well gradient polyacrylamide gel. Samples were heated at 95˚C for approximately 3 minutes before loading on the gel. Gels were run in gel sample buffer until the marker lane was well separated (35 minutes); then fixed in 50% methanol, 10% glacial acetic acid solution and stained with Coomassie Blue, then de-stained in 50% methanol, 10% glacial acetic acid; further soaked in 5% acetic acid solution, removed and photographed. The gel loading buffer for the samples contained 20 mM DTT to insure samples were in a reducing environment.

All DSC melting experiments were performed using a CSC Model 6100 Nano II-Differential Scanning Calorimeter (formerly Calorimetry Sciences Corporation, Provo UT, now TA Instruments). In a DSC melting experiment the excess heat capacity, ∆C_p, is measured as a function of increasing temperature. Gasket tipped mico- pipettes provided an airtight means for protein/buffer injections and care was exercised in loading both buffer and sample solutions in the DSC to prevent introduction of air bubbles. The average of five buffer scans collected over the temperature range from 0 to 100˚C served as the buffer baseline for analyzing scans of protein samples. For all DSC experiments the temperature scan rate was 1˚C/min and the protein concentration was approximately 2 - 2.5 mg/ml. To prevent degassing of solutions upon heating, all measurements were made at a pressure of 3.0 atm. Following each run, before a fresh buffer and sample were loaded in the DSC, sample and reference cells were carefully cleaned with a 95% ethanol solution.

Results of preliminary experiments (not shown) determined that the melting transition of HSA₉₉ spans the temperature region between 50 and 90˚C, and this is the temperature range for all displayed thermograms. To encompass this region, the temperature range used for measuring DSC thermograms of HSA₉₉ (and HSA₉₆) was from 45 and 95˚C.

DSC data was analyzed using the CpCalc 2.1 software package provided by the DSC manufacturer. The standard analysis procedure was comprised of six essential steps. 1) Measured heat in µW was converted to raw molar heat capacity, ΔC_p-raw(T) [cal/K∙mol] versus temperature curves. 2) These raw curves were then normalized for protein concentration (2.0 - 2.5 mg/mL), molecular mass (66.5 kDa), partial specific volume of the protein molecules (0.733 cm³/g) [32] and sample cell volume producing normalized ΔC_p(T) versus T curves (ΔC_p-N(T) versus T). 3) The measured buffer baseline scan (ΔC_p-buffer(T) versus T) curve was subtracted from the normalized (ΔC_p-N(T) versus T) sample curve producing a buffer corrected, ΔC_{p-buffer corrected}(T) versus T curve. 4) Next, the baseline of the buffer corrected ΔC_{p-buffer corrected}(T) versus T curve was determined using a polynomial function constructed by connecting the lowest and highest temperature points on the buffer corrected ΔC_{p-buffer corrected}(T) versus T curve (generally the range from 45˚C - 90˚C). 5) This fitted baseline was then subtracted from the ΔC_{p-buffer corrected}(T) versus T curve producing finally, a concentration normalized, buffer and baseline corrected ∆C_{p-N-buffer-baseline-corrected}(T) versus T curve which is the DSC melting curve, or thermogram. 6) The experimentally observed calorimetric transition enthalpy, , was evaluated from the integrated area under the measured thermogram. The entropy, , was evaluated from the integrated area of the ∆C_{p-N-buffer-baseline-corrected}(T)/T versus T curve. In the analyses that were performed the observed calorimetric free-energy, , was determined using the Gibb’s relationship, i.e.

The transition temperature, t_m, is the temperature of the peak height maximum on the thermogram. The t_m and evaluated measured thermodynamic parameters, and provided quantitative characterizations of the temperature-induced melting transitions of the HSA samples. Peak heights and peak widths at half height also served as an additional means of semi-quantitative comparison of thermogram shapes.

Analysis of DSC data in this fashion was performed under the simplifying assumptions that the melting transition of HSA occurs in a two-state manner, and that the overall change in the value of ΔC_p at the initial and final temperatures is zero [33] . Other than the choice of baseline and assumption that the ∆C_p = 0, no other assumptions governed evaluation of the thermodynamic transition parameters. Results indicated the impact of these assumptions was not very severe, as evaluated parameters were in reasonable agreement with published results; and relative comparisons provided quantitative insight into effects of biotinylation on HSA structure and stability. Thus, we assumed the observed evaluated parameters corresponded to the denaturation transition of the HSA species in their standard state. The standard state thermodynamic parameters evaluated from thermograms of modified and unmodified HSA₉₉ and HSA₉₆, i.e. generally denoted, , and, were employed to interpret their observed contrasting behaviors in response to covalent attachment of biotin.

More sophisticated models that include the possibility of multi-state transitions and a non-zero ΔC_p have been applied to analyze DSC measured denaturation curves of HSA and provided evaluations of thermodynamic transition parameters [1] [13] . Obviously depending on specific features of the particular analytical model employed to analyze DSC thermograms, actual values of the thermodynamic transition parameters obtained from the analysis could differ. However, relative values of the evaluated parameters should provide a more consistent basis for comparisons of results obtained using different models for thermogram analysis.

The average of multiple measurements (>10) of the excess heat capacity, ΔC_p, versus temperature (thermograms) for 99% pure HSA (HSA₉₉) and 96% pure HSA (HSA₉₆) samples, are shown in Figure 1. The distinctly different shapes of the thermograms for HSA₉₉ and HSA₉₆ in Figure 1 demonstrate the capability of DSC measure-

ments to discern effects on HSA of the presence of the FA- and globulin-like contaminant (FA/G-LC), advertised as comprising the 3% impurity in the case of HSA₉₆. Error bars indicate the experimental variation from the average over at least three independent experiments. The curve for HSA₉₉ is the leftmost curve shown in Figure 1. As indicated by the error bars, in multiple DSC experiments the thermogram for HSA₉₉ was highly reproducible, with a nearly symmetric melting transition, a single major peak and slight high temperature asymmetry, and average t_m = 63.3˚C ± 0.2˚C. The peak width at half height of the HSA₉₉ thermogram is not more than 7˚C. The t_m and curve shape of the HSA₉₉ thermogram are similar to those in published reports for defatted HSA also measured by DSC [1] [12] [17] . These reports have onset of the HSA melting transition at around 50˚C with a span of 35˚C and completed by about 85˚C. Preparations of HSA₉₉ made from freshly diluted concentrated stock solutions (prepared from dry powder) and stored for over a year at 4˚C produced thermograms that were indistinguishable from those obtained from freshly prepared HSA₉₉ concentrated stock solutions. Accordingly, this stability of the HSA₉₉ thermogram confirms the structure of this preparation is quite stable during storage.

The average thermogram measured for HSA₉₆ is also displayed in Figure 1 (right-most curve). When the thermograms for HSA₉₆, and HSA₉₉ are compared, the thermogram for HSA₉₆ is shifted to higher temperature by more than 5˚C and overall the transition seems to be comprised of at least two sub-transitions centered at 68.8 and 78.8˚C. There is also a very minor peak inflection on the low temperature side of the main transition at the same temperature as the t_m of HSA₉₉. Thermodynamic transition parameters evaluated from the thermograms of HSA₉₉ and HSA₉₆ are summarized in Table 1 and discussed later.

The bi-phasic melting transition observed for HSA₉₆ is remarkably similar to results of published DSC melting experiments on HSA in the presence of endogenous long chain fatty acid (LCFA) [1] [15] . Measuring under nearly identical conditions (buffer and solvent, scan rate, protein concentration) these authors observed a bi- phasic melting transition attributed to coexistence of two different conformational states of HSA in the presence of LCFA. To indicate their relative affinity for binding LCFA these different forms were dubbed the “fatty acid- poor” or “defatted” and “fatty acid-rich” or “un-defatted” structures of HSA. Reportedly the bi-phasic melting transition resulted from redistribution of LCFA between the two forms, which differed in LCFA binding

affinity and temperature dependent stability, during the denaturation process. The observed similarity between our results and the published results [1] [15] leads to the supposition that the 3% impurity in the HSA₉₆ preparation (FA/G-LC) affects HSA structure and stability in a manner similar to that reported for LCFA [1] . In support of this supposition, the reported relative amounts of LCFA present in the FA-poor (0.084 mole of LCFA per mole HSA) and FA-rich (1.43 mole LCFA per mole HSA) HSA preparations are comparable to ours and span the value of 0.51 mole of FA per mole HSA assumed for the composition of the 3% FA/G-LC in our HSA₉₆ preparation.

Because of the remarkable similarity between the thermogram for HSA₉₆ and published thermograms for defatted and un-defatted HSA measured under very similar solution and experimental conditions, the bi-phasic thermogram of HSA₉₆ in Figure 1 is also attributed to the presence of two protein populations. In analogy with previous reports [1] [15] , and for reference in future discussions, the first peak on the HSA₉₆ thermogram at 68.8˚C is designated HSA_96-L; while the second peak at 78.8˚C is referred to as HSA_96-H. The former is thought to more resemble the structure and stability of HSA₉₉ (but arguably not precisely the same). This distinction is evidenced by the small shoulder peak on the low temperature side of the HSA₉₆ thermogram attributed to the HSA₉₉ species. This small inflection occurs precisely at the t_m of HSA₉₉ (63.3˚C), and likely indicates a slight amount of HSA₉₉ also present in the HSA₉₆ preparation. Presumably, both HSA_96-L and HSA_96-H bind FA to a lesser or greater extent, respectively [16] . Similarity of the melting characteristics of the HSA₉₉ and HSA₉₆ preparations with published results also provided assurance of the fidelity and initial native state of these HSA preparations prior to covalent attachment of biotin. Before describing results of experiments on biotinylated HSA₉₉ and HSA₉₆, several additional aspects of their behavior are described.

The HSA₉₆ preparation was not as stable in storage as HSA₉₉. Examination of the HSA₉₆ thermograms at different times after preparation indicated the average HSA₉₆ thermogram displayed in Figure 1 was reproducible for at least one week (but not for much longer). On multiple occasions over the course of this study we observed the HSA₉₆ thermogram was not always reproducible for samples stored at 4˚C for longer than a week.

According to their thermograms, spontaneous conversion of HSA₉₆ to HSA₉₉ was observed. Accepting that HSA₉₆ contains a mixture of HSA_96-L and HSA_96-H the thermograms of HSA₉₆ (Figure 1) were found to differ for different preparations, and in particular to be quite sensitive to age of the sample (after those preparations) when the thermograms were measured. With this lability of the HSA₉₆ samples, reported thermograms for HSA₉₆ (displayed in Figure 1) were obtained for protein samples that were made from HSA₉₆ stock solutions prepared fresh from powder and examined within (at most) one week following preparation. Generally the time after preparation when thermograms were measured for HSA₉₆ and its biotinylated derivatives was 24 to 72 hours.

Although a systematic study of the state of the HSA₉₆ preparation at different times after preparation was not performed, typical results are shown in Figure 1. There, the thermogram of HSA₉₉ (freshly prepared and less than a week old, solid line) is compared with that obtained for a preparation of HSA₉₆ stored for three weeks prior to examination by DSC (dotted line). Clearly, the thermogram for the older HSA₉₆ is very similar to that of HSA₉₉, evidencing conversion. Such conversion was observed in as little as one week for some HSA₉₆ samples stored at 4˚C.

Experiments investigating the reversibility of the HSA₉₉ melting transition were also performed (data not shown). For this investigation thermograms were measured for HSA₉₉ samples first heated in separate experiments to temperatures from 30 to 70˚C, cooled at the same rate, and then re-heated from 20 to 100˚C. Results showed when HSA₉₉ was first heated to temperatures up to 60˚C, just at the edge of the melting transition, cooled and re-melted, the thermogram was not different from the thermogram of unheated HSA₉₉; with a single peak, t_m = 63.3˚C. This result was taken to indicate that up to about 60˚C, corresponding to early onset of the melting transition, the measured thermogram was unaffected by prior heating and cooling, and was reversible. In contrast, when the HSA₉₉ solution was pre-heated to temperatures above t_m, i.e. up to 80˚C, cooled and re-melted, the observed thermograms for the pre-heated samples were very different from those observed for the unheated (or reversible) curves. These results were in concordance with those reported in published studies of HSA, that HSA melting is reversible up to ≈62˚C [12] [13] . From experiments performed under conditions similar to ours these same authors found at temperatures above 75˚C, HSA resides in an irreversibly denatured, predominantly molten globule state. Reportedly, the rate of the irreversible reaction was considerably slower at lower temperatures (below 74˚C) [12] . Under experimental conditions comparable to ours these authors asserted that even up to 74˚C the denaturation reaction is predominantly reversible. This is a major operating assumption.

Figure 2 shows the thermograms for HSA₉₉ and HSA₉₆ modified by biotinylation at a single site. Protein samples that produced the results shown in Figure 2 were prepared by using the Maleimide-activated reagent targeting the single reduced sufhydryl at cys-35 of the HSA primary structure. Products of the reactions were either HSA₉₉ or HSA₉₆ with biotin attached, denoted HSA_99-M(x:1) or HSA_96-M(x:1), respectively (M = biotin attached with Maleimide). In Figure 2, thermograms are displayed for HSA_99-M(x:1) (a) or HSA_96-M(x:1) (b) incubated at molar ratios of biotin:HSA₉₉ = 1:1, 3:1 and 10:1 (x = 1, 3, 10). Error bars show experimental variation from the average curve (symbols) for HSA_99-M(x:1) at each incubation ratio. As seen in Figure 2(a), the effect of attachment on the HSA₉₉ thermogram is a peak height reduced by nearly 40% and shift of t_m upward in temperature by over 5˚C. At x = 1 the peak width at half height increases by more than a factor of two. After these most dramatic changes initially at a 1:1 ratio, thermograms measured at biotinylation ratios of x = 3 and x = 10 are not greatly different. They have essentially the same t_m with only slight increases in peak widths on the high temperature side; and peak heights are the same (within error). Results of these changes associated with biotin attachment also manifest in the measured thermodynamic parameters for HSA_99-M(x:1) summarized in Table 1. The evaluated parameters, t_m, and for HSA_99:M(x:1) show little variation between the different biotinylation ratios of x = 1, 3, 10.

In the same manner as for HSA₉₉, biotin was covalently attached to HSA₉₆ (prepared fresh) and the thermograms were measured within 48 hours of sample preparation. Thermograms collected for the covalently modified HSA₉₆ are shown in Figure 2(b), and denoted HSA_96-M(x:1). Effects of single site attachment of biotin to HSA₉₆ are dramatic, and there is a very clear difference between the HSA₉₆ thermogram and the HSA_96-M(x:1)

thermograms. Responses to biotin attachment at a single site on HSA₉₆ might be unexpected for several reasons. 1) As indicated by the bi-phasic thermogram in Figure 1, the preparation of HSA₉₆ is comprised of two different forms that have been denoted HSA_96-H and HSA_96-L, present in their respective mole fractions. 2) The effect of the biotinylation reaction on the relative population of these two states of HSA is unknown. 3) The bi-phasic nature of the HSA₉₆ thermogram indicates HSA_96-H is more stable against denaturation than HSA_96-L.

Presumably, when the attachment reaction is performed using HSA₉₆, both components, HSA_96-H and HSA_96-L are targeted for biotinylation. At incubation ratios of x ≥ 1, the two-peaked thermogram of HSA₉₆ in Figure 1 with t_m’s at 68.8 and 78.8˚C is consolidated into a broad, single-peaked transition with an intermediate t_m = 72.3˚C. The peak height is roughly the same as the major peak at 68.8˚C on the thermogram of HSA₉₆ alone (solid line).

Consolidation of the thermogram in Figure 2(b) seems to arise from a shift up in temperature of the major peak attributed to labeled HSA_96-L (HSA_96-L-M(1:1)). The higher temperature peak corresponding to labeled HSA_96-H (HSA_96-H-M(1:1)) seems to have shifted down a few degrees and become absorbed into the consolidated peak brought about by the up-shifted HSA_96-L-M(1:1) peak. The peak widths at half height of the thermograms for the HSA_96-M(x:1) samples in Figure 2(b) are reduced from that of HSA₉₆ (~20˚) to approximately 15˚C. Again as expected for a single biotinylation site, once saturation occurs (at x =1) incubation at increased ratios of biotin has little effect on the thermogram shape, or t_m. In effect, there are no significant differences in the thermograms for HSA_96-M(x:1) at x ≥ 1. Just as observed for HSA_99-M(x:1) thermodynamic transition parameters for HSA_96-M(x:1) summarized in Table 1 are essentially the same at all attachment ratios examined. Although they show a similar response to biotinylation the actual thermograms of HSA_99-M(x:1) and HSA_96-M(x:1) are different. The peaks of the HSA_99-M thermograms are sharper with an average t_m = 68.5˚C compared to the broader symmetric peaks of the HSA_96-M(x:1) thermograms with average t_m = 73.8˚C. As shown in Table 1, and discussed in detail later, evaluated thermodynamic parameters for the thermograms of HSA_96-M(x:1) and HSA_99-M(x:1) also displayed different responses to biotin attachment.

The primary structure of HSA contains 59 lysine residues. Examination of the HSA crystal structure and results of cross linking studies suggested nearly half of the 59 lysine residues in HSA could be adequately exposed and therefore presumably accessible for biotinylation [26] [27] . HSA samples that produced the thermograms shown in Figure 3 and Figure 4 were prepared by using the NHS-activated biotin reagent targeting the primary amines of lysine residues for attachment with increasing amounts of biotin. Thermograms of NHS biotinylated HSA₉₉, denoted HSA_99-NS(x:1), are shown in Figure 3. NS corresponds to biotin attached with NHS. Different thermograms shown in Figure 3 are for samples with increasing amounts of attached biotin.

Thermograms in Figure 3(a) are for HSA_99:NS(x:1) samples where the attachment ratios varied from x = 1 to 10. Figure 3(b) shows the thermograms obtained for HSA_99:NS(x:1) samples for which the incubation ratio varied from x = 10 to 50. Differences in the thermograms of HSA_99-NS(x:1) and HSA₉₉ seen in Figure 3 indicate a very clear effect of biotinylation. With increased attachment ratios from x = 1 to 20 there is a corresponding shift of the thermograms to higher temperature with incrementally increased t_m. The peak heights of the modified thermograms are all about the same and reduced by about 15% from that of HSA₉₉. Peak widths at half height of the modified thermograms increase by nearly 50% from the thermogram of HSA₉₉ alone. Figure 3(b) shows the thermograms obtained for HSA_99-NS(x:1) at incubation ratios from x = 10 to 50. Up to x = 20, the thermogram t_m is shifted up in a titratable fashion by more than 9˚C. At higher incubation ratios examined (x = 40, 50) there are no further changes in t_m. At incubation ratios greater than x = 10, the peak heights and widths at half height vary somewhat. Even with this variation, evaluated thermodynamic parameters in Table 1 are very consistent.

Thermograms of HSA₉₆ biotinylated at multiple sites are shown in Figure 4, and denoted HSA_96-NS(x:1) where again x is the incubation ratio corresponding to the relative amount of biotin attachment. Thermograms were measured for x = 1, 3, 5, 8, 10, 15, 20, 40, 50. A sampling of these is shown in Figure 4. Thermograms shown in Figure 4(a) were measured for HSA_96-NS(x:1) samples prepared at incubation ratios varying from x = 1 to 10. Thermograms in Figure 4(b) were obtained for HSA_96-NS(x:1) samples prepared at higher incubation ratios from x = 10 to 50. There is a stark difference between the HSA₉₆ thermogram (thin black line) and the HSA_96-NS(x:1) thermograms. Most notably, at the lowest attachment ratio (x = 1) the formerly two-peaked HSA₉₆ thermogram (thin black line in Figure 4(a)) becomes single peaked with t_m = 66.2˚C (bold dash line Figure 4(a)). For x ≥ 1 the HSA_96-NS(x:1) thermograms have the same shape but are shifted up in temperature, with increased t_m at higher attachment ratios up to about x = 40, with no changes thereafter (Figure 4(b)).

It is important to recall and consider that the initial preparation of HSA₉₆ from which the HSA_96-NS(x:1) samples were prepared was comprised of a mixture of two forms of HSA (referred to earlier as HSA_96-H and HSA_96-L). Accordingly as seen in Figure 4 for HSA₉₆ the effect of biotinylation at lysine residues is different from that for single site biotinylation at cys-35 (Figure 2(b)). On the thermograms of HSA_96:NS(x:1) (Figure 4) there is no sign of the high temperature transition at 78.8˚C, denoted HSA_96-H, suggesting the attachment reaction itself must result in conversion of the originally present more stable structure, to a less stable one. The converted less stable form, HSA_96-L, when modified at a 1:1 ratio becomes HSA_96-NS(1:1) which is quite qualitatively similar to HSA_99-NS(1:1), with an identical t_m. Similar comparisons can also be made between HSA_96-NS(x:1) and HSA_99-NS(x:1) at increasing values of x (discussed later).

Thermograms of HSA₉₆ samples incubated under the same conditions as the attachment reactions (four hours at 21˚C), without the attachment reagents, were indistinguishable from the thermogram of freshly prepared HSA₉₆ samples (not shown).

In summary, results in Figure 1 through Figure 4 clearly demonstrate the ability of DSC thermograms of HSA to detect the presence or absence of a slight amount of contaminant that contains globulins and FA, denoted earlier as FA/G-LC (Figure 1); detect changing states of HSA samples over time (Figure 1); differentiate covalent modification of HSA at a single site (Figure 2); or at multiple sites (Figure 3 and Figure 4). For both species of HSA, the effect of single site modification is a thermogram reduced in peak height with a small shift up in t_m. For multiple site modification the thermogram displays a significant temperature shift up with increasing amounts of biotinylation.

DSC measurements provided evaluations of the thermodynamic parameters of HSA melting transitions. The enthalpy, , entropy, , and standard state free-energy at 37˚C, , and corresponding t_m’s of the melting transitions of the different HSA samples of this study are summarized in Table 1. Comparison of the evaluated thermodynamic parameters in Table 1 reveals differences between HSA_96-NS(x:1) and HSA_99-NS(x:1). Differences between the two samples at lower attachment ratios up to x = 20 suggest the HSA_96-L-NS(x:1) samples at the lower attachment ratios reside in a different state than that of HSA_99-NS(x:1). That is, HSA_96-L-NS(x:1) can also be bound by components of the FA/G-LC contaminant present in the HSA₉₆ preparation. The measured t_m’s of HSA_96-NS(x:1) are plotted versus attachment ratio, x, in Figure 5 (open symbols) and compared with the behavior of HSA_99-NS(x:1) (filled symbols). At lower attachment ratios, HSA_96:NS(x:1) is actually less stable than HSA_99-NS(x:1), indicating the proposed HSA_96-L-NS(x:1) structure (with the presence of FA-containing ligands) is less stable than

its HSA_99-NS(x:1) counterparts at the same values of x. At higher incubation ratios (≥1:20) t_m’s of the thermograms level off to around 75˚C.

Figure 5 shows for HSA_96-NS(x:1) that t_m also approaches 75˚C at higher x values, but does so more slowly than HSA_99-NS(X:1). In Figure 5 the t_m’s of the HSA_99-NS(1:x) preparation (at least for x ≤ 20) indicate it is apparently more stable against thermal denaturation than HSA_96-L-NS(1:x) at the same x. The thermodynamic parameters evaluated for HSA_96:NS(x:1) and HSA_99:NS(x:1) (and) versus x displayed the same qualitative behavior as displayed for versus x in Figure 6 (not shown). For comparisons that follow values of t_m and versus x are utilized. Observed trends in t_m and with covalent attachment of biotin to HSA₉₉ and HSA₉₆ provide an indication of the effects of biotinylation on thermodynamic stability. As shown in Figure 5 and Figure 6, these parameters show very different responses of the two HSA forms to covalent modification by biotin. In Figure 6 the values from Table 1 are plotted versus attachment ratio for the HSA_99-NS(x:1) and HSA_96-NS(x:1) preparations. This plot reveals differences at the lowest attachment ratios, the plots reaching near equivalence around x = 15. Interestingly, is greater for the HSA_96-NS(1:x) preparation, even though the t_m is lower than that for HSA_99-NS(x:1) at comparable x. This observation could be due to effects of irreversible aggregation in the denaturation region.

HSA₉₉ (without the FA/G-LC contaminant) aggregates more readily in the denaturation region, i.e. has a higher rate of intermolecular aggregation mediated by the melted state than the populations of HSA_96-H and HSA_96-L that comprise the HSA₉₆ preparation [1] . FA is known to mediate aggregation in the denaturation region, thus the greater stability of HSA_96-H. By the same argument, attachment of biotin to HSA₉₉ to form HSA_99-NS(1:x) must also act to decrease irreversible aggregation, as the stability of HSA_99-NS(1:x) is increased by more attachment. Then Figure 6 suggests irreversible aggregation in the denaturation region is increasingly discouraged by increased amounts of biotinylation. The result is values of for HSA_99-NS(x:1) that are smaller than for HSA_96-NS(x:1) at the lower attachment ratios (as plotted in Figure 6), but equivalent at higher ratios.

Prior to comparison, discussion and interpretation of tabulated thermodynamic parameters it should be noted that parameter evaluations were made under the assumption that HSA melting from the native to the denatured state predominantly occurs in a two-state manner. Several published studies have reported results of measurements on HSA and thermodynamic transition parameters of the HSA melting transition [1] [13] [17] . In these reports different methods of data analysis, baseline treatment, model assumptions and curve fitting were employed. Ours is a standard approach, and arguably one of the least sophisticated that has been applied, with several simplifying assumptions. As discussed below relative comparisons of the parameters within the data set were considered to be quantitatively significant, and provided additional insight into the consequences of biotin attachment on the different HSA samples. Melting transition parameters evaluated by DSC analysis of HSA₉₉,

HSA₉₆ and their modified versions with biotin attached, HSA_99-M(x:1), HSA_99-NS(x:1), HSA_96-M(x:1) and HSA_96-NS(x:1) are summarized in Table 1.

A few general features of the parameter sets are noteworthy. 1) For both preparations of unmodified HSA alone, HSA₉₉ and HSA₉₆ measured values (= 163.3 kcal/mole for HSA₉₉), (= 279.5 kcal/mole for HSA₉₆) approach, or are within, the range of reported values for defatted and un-defatted HSA [1] [13] [17] . This general agreement indicates our simplifying assumptions of a two-state melting transition with ΔC_p = 0, and variations in baseline analysis procedures must have a small effect on the quantitative significance of the evaluated thermodynamic parameters. 2) Response of the t_m of HSA to covalent attachment of biotin is different for HSA₉₉ than for HSA₉₆. For HSA_99-M(x:1) at (x = 1), the t_m increases by over 5˚C, and does not change further at higher attachment ratios. The average (over all attachment ratios) is t_m = 68.7 ± 0.4˚C. Response of the thermodynamic parameters, and for HSA_99-M(x:1) concurs with that for the t_m. For HSA_99-M(x:1) increases to the same value at ratios of x = 1, 3 and 10 with an average value of 16.8 ± 0.9 kcal/mol. In contrast, under the same conditions for HSA_96-M(x:1) samples at attachment ratios of x = 1, 3 and 10 there is a nearly 13% decrease in and values, but values for modified HSA_96-M(x:1) are identical to that of unmodified HSA₉₆ with average = 26.1 ± 0.9 kcal/mol. To summarize for the cases of HSA_99-M(x:1) and HSA_96-M(x:1), both show no difference between the attachment ratios (x = 1, 3, 10), consistent with single site attachment. For HSA_99-M(x:1) biotin attachment results in an increase in stability () while HSA_96-M(x:1) displays the same stability as HSA₉₆.

In contrast, opposite responses of and, and in particular, to attachment of biotin to amines of lysine residues were found for HSA_99-NS(x:1) compared to HSA_96-NS(x:1). For HSA_99-NS(x:1) the t_m increases incrementally with increasing attachment ratios x > 20 for a total increase of nearly 10˚C, and levels off thereafter (see Figure 5). Examination of Table 1 indicates the thermodynamic parameters for HSA_99-NS(x:1) and HSA_96-NS(x:1) tend in opposite directions. In comparison consider the behavior of values versus attachment ratio as plotted in Figure 6. As shown there, for HSA_99-NS(x:1) at x = 1, increases and continues to do so in an incremental fashion up to an attachment ratio of x = 15 where after it levels off and is constant up to the highest ratios (x = 40, 50). In contrast for HSA_96-NS(x:1), at an attachment ratio of x =1, decreases from the value for HSA₉₆ and decreases again for x = 3 and 5; dramatically increases at x = 8, then levels off at the higher ratios (see Figure 6). The opposite responses of the values for HSA_99-NS(x:1) and HSA_96-NS(x:1) as a function of increased attachment ratio reveals an effect of the presence of the small amount (3%) of FA/G-LC on modified HSA. The contrasting observations are: 1) Attachment increases stability () of the HSA_99-NS(x:1) samples compared to HSA₉₉ alone. This stability enhancement occurs in an incremental fashion up to 1:15 were presumably saturation of available attachment sites is reached. 2) Attachment dramatically decreases stability () of the HSA_96-NS(x:1) samples at the lowest attachment ratios (x = 1 - 10); thereafter at increased attachment ratios (x ≥ 10) converges to a constant value of higher stability, but with less stability than HSA₉₆ alone.

In order to provide an interpretation of the thermodynamic measurements of the two forms of HSA and consequences of biotinylation of them, consider the following simple model of thermal denaturation of HSA written as the following reaction,

where the transition is assumed to occur in a pseudo two-state manner from the native intact structure, , to the denatured state,. This reaction occurs with an observed standard state equilibrium constant,. The observed standard state free-energy is given by. At temperatures below t_m the standard state free-energy, and the native structure is favored.

The equilibrium reaction in Equation (1) can be represented in terms of the corresponding chemical potentials as,

where and are the hypothetical standard state chemical potentials for the native state (N) and the denatured state (D), respectively. The standard state free energy, , is given in terms of the differences of the standard state chemical potentials of the native and denatured states, i.e.

The observed standard state equilibrium constant for the reaction is also given by,

At temperatures below t_m,. Therefore,. The values of determined at 37˚C for both HSA₉₉ and HSA₉₆ are given as their values in Table 1.

In terms of the specific samples at hand, for HSA₉₉,

Similarly for HSA₉₆,

If the denatured state for both species is arbitrarily assigned as the reference state, , then and. Henceforth for convenience we write and.

The general treatment above requires slight modification to properly consider the HSA₉₆ preparation which, as stated earlier and evidenced by Figure 1, is primarily comprised of two different populations of the protein, termed HSA_96-H and HSA_96-L. To account for this mixture we write,

with equal to the mole fraction of HSA_96-H in the HSA₉₆ preparation. Here it is assumed if a small amount of HSA₉₉ is present, it is so in a negligible amount (<5%). As such, writing Equation (8) in this manner ignores the slight presence of HSA₉₉ (reckoned as the small inflection on the low temperature side of the HSA₉₆ melting transition in Figure 1). Equation (8) assumes that other than HSA_96-H the remaining protein is in the HSA_96-L form.

As stated previously we assume HSA_96-L and HSA_96-H that comprise the HSA₉₆ preparation are in many ways analogous to the defatted (HSA_L-F) and un-defatted (HSA_H-F) forms of HSA previously studied and reported [1] [15] [16] . Their different forms were observed under conditions very similar to ours (heating rate, buffer composition, protein concentration, relative fraction of FA). From their analysis they reported a value of x_H in Equation (8) around 0.3 [1] [16] . They also reported measured average values for the transition enthalpies for the HSA_L-F and HSA_H-F species. These were, ΔH_L-F = 4.18 cal/g for the defatted species and ΔH_H-F = 5.16 cal/g, for the un-defatted species [1] . The molecular weight of HSA (66,500 g/mol) provides ΔH_L-F = 280.06 kcal/mol and ΔH_H-F = 345.72 kcal/mol. Reported t_m values for HSA_L-F and HSA_H-F were 65.7˚C and 78.3˚C, respectively. These combined with ΔH_L-F and ΔH_H-F values provide an estimate on the transition entropies ΔS_L-F and ΔS_H-F for

the reported defatted and un-defatted species. That is, and

. If these reported enthalpy and entropy values are assumed for

the standard state, then estimates on the standard state free-energy at 37˚C, , of 23.9 kcal/mol and 39.6 kcal/mol were obtained for the HSA_L-F and HSA_H-F structures, respectively. In analogy with Equations (7) and (8) these values yield the chemical potentials and for the defatted and un-defatted forms of HSA. The difference,. In the discussion that follows it is arbitrarily assumed for the example shown that the values for HSA_96-L and HSA_96-H are approximately 75% of reported values for the presumably analogous forms, HSA_L-F and HSA_H-F. That is we assume and, with the same relative difference . This is probably not an unrealistic assumption based upon differences in data analysis and reported thermodynamic parameters compared to ours [1] . For both sets of values the relative difference remains the same.

With these assumptions, estimates on the value of x in our experiments can be made by rearranging Equation (8),

Using the values given above for and and the measured value of we find.

In a separate graphical analysis of the HSA₉₆ thermogram (not shown), it was assumed the composite curve in Figure 1 collected for the HSA₉₆ sample was comprised of just two major components, HSA_96-H and HSA_96-L. The graphical curve analysis returned an estimate of which is in agreement with what was calculated assuming the values of and above, but about 20% larger than reported [16] for the relative fractions of the analogous forms HSA_L-F and HSA_H-F. In the remaining development a value of is assumed for the initial composition of HSA₉₆.

Consider the following cases. For the HSA₉₉ preparation,

where represents the generic attachment reaction that depends on the site of attachment, Bio = M or NS. Consider single site attachment first, i.e. Bio = M. In analogy with Equation (6) the observed standard state free-energy of the melting reaction for HSA_99-M is given by,

Again, letting and, which is compared to . As a result, , and the attached biotin increases the chemical potential of the HSA_99-M(x:1) structure compared to HSA₉₉. In the above we have used the average value of evaluated from the average over the values obtained for the HSA_99-M(x:1) samples at x = 1, 3, and 10.

For the HSA₉₆ preparation the thermogram is comprised of two overlapping transitions. The lower temperature transition seems most affected by attachment and is attributed to HSA_96-L with biotin attached, i.e. HSA_96-L-M(x:1), and assumed to be similar (but not identical) to HSA_99-M(x:1)

It is possible that during the attachment reaction some of the labeled HSA_96-H molecules (HSA_96-H-M) convert to the HSA_96-L-M form. To account for these considerations the measured chemical potential of the HSA_96-M preparation is written as,

Rearranging Equation (10) and solving for provides an estimate on the fraction of HSA_96-H-M molecules that remain following the attachment reaction, i.e.

We assume that biotinylation of and has the same effect as biotinylation of, and similarly write the analogous chemical potentials, i.e.

and

And the average of the measured values of evaluated at x = 1, 3 and 10, corresponds to the chemical potential. With the values of, and ,

This indicates that the fraction of molecules in the HSA_96-H-M form decreases from 0.53 to 0.24 when the single site labeling reaction takes place. The result is approximately 50% of the HSA_96-H in the original HSA₉₆ preparation being converted in the single site attachment reactions.

Alternatively, consider the fraction of molecules in the HSA_96-H-M state (presumably now labeled) does not change during the attachment reaction, i.e.. How does the resulting value of change compared to due to the attachment reaction? Rearranging Equation (10) and solving for,

This is considerably smaller than So the effect of biotinylation on the HSA₉₆ preparation is either to convert some fraction of the HSA_96-H-M molecules to the HSA_96-L-M form, or reduce the stability of HSA_96-H-M compared to HSA_96-H.

Results of the above analysis and behavior of the measured thermograms are consistent with the following scenario. In the process of the attachment reaction a portion of the HSA_96-H molecules get biotinylated and convert to HSA_96-L-M essentially becoming HSA_99-M, which is much less stable than HSA_96-H but more stable than HSA₉₉ alone. Thus, there are two opposing effects of biotin attachment on protein stability. Biotin attachment to the HSA_96-L and HSA_96-H structures to create HSA_96-L-M and HSA_96-H-M increases their stabilities. But with the attachment reaction a fraction of the more stable, HSA_96-H molecules are converted to the less stable HSA_96-L form, then subsequently biotinylated to become HSA_96-L-M ≈ HSA_99-M. In combination, the loss of a fraction of the HSA_96-H molecules through conversion results in a decrease in stability and reduces the chemical potential. This destabilization is overcome by attachment of biotin to HSA_96-L to create HSA_96-L-M resulting in an increase in chemical potential. This reduction and increase offset. As a consequence both the HSA_96-M and HSA₉₆ preparations have the same chemical potential.

Evidently, biotinylation increases stability of HSA_96-L-M while reducing stability of HSA_96-H-M compared to their unlabeled counterparts, HSA_96-L and HSA_96-H, respectively. It is possible that binding of ligands that comprise the FA/G-LC contributes to enhanced stability of HSA_96-H that must exist in a more stable conformation than HSA_96-L. HSA_96-L may bind to the FA/G-LC ligands and have a structure and stability like HSA₉₉ bound by the FA/G-LC ligands and have a different conformation than HSA_96-H. Clearly, the 3% contaminant, FA/G-LC, in the HSA₉₆ preparation has quite a profound effect on the structure and stability of the different forms of HSA that comprise the HSA₉₆ preparation.

For the case of multiple attachment, consider for the HSA₉₉ preparation,

where for multiple attachment Bio(x:1) = NS(x:1). For HSA_99-NS(x:1),

Again, letting,. These values of are given in Table 1, and plotted versus attachment ratio, x, in Figure 6. Higher x results in increased stability of the HSA_99-NS(x:1) structure compared to HSA₉₉. For the HSA₉₉ preparation the effect of biotinylation is to incrementally increase (and t_m) with increased attachment ratio, x. These results for HSA_99-NS(x:1) provide the basis for dissecting the effects of multiple attachment on the HSA₉₆ preparation.

On the HSA_96-NS(x:1) thermograms in Figure 4 there is no sign of the high temperature transition attributed to the aforementioned HSA_96-H structure. Apparently this would indicate an effect of the attachment reaction is conversion of all molecules to the HSA_96-L-NS(x:1) type structure. Based on this observation, we assume the thermograms of the HSA_96-NS(x:1) samples are comprised of contributions of the labeled HSA_96-L-NS(x:1) which is similar, but not identical (especially at low attachment ratios) to HSA_99-NS(x:1); and contributions from binding of the FA/G-LC to that structure, i.e.. In this case the chemical potential for HSA_96-NS(x:1) is written as,

or

If we assume in Equation (11) that then the observed differences, are the contributions of binding of the FA/G-LC to the chemical potential of HSA_96-NS(1:x), a state believed to be HSA₉₉ with attached biotin and bound by the FA/G-LC. Both biotinylation and FA/G-LC binding are stabilizing interactions.

Values of are plotted versus attachment ratio (x) in Figure 7. As shown there, the value of is highest at the lowest attachment ratio where the contribution from FA/G-LC binding is greatest. As attachment ratio increases rapidly decreases to essentially zero at the higher biotinylation ratios. This decrease presumably corresponds to occlusion of FA/G-LC binding to HSA_96-L-NS(x:1) by increased biotin attachment at the higher ratios. Given the above analysis and interpretation derived therefrom, we can provide an explanation of the apparently contrasting behaviors of the HSA_99-NS(1:x) and HSA_96-NS(1:x) species. Consider the following scenario for HSA_96-NS(1:x). At the lowest attachment ratios, the protein resides as HSA_96-L with biotin attached to a few sites and some FA/G-LC bound to other potential binding sites. This results in a structure with a stabilizing component in the biotin label (as indicated for HSA_99-NS(x:1)) and a further stabilizing component due to binding of FA/G-LC to HSA_96-L-NS(x:1). As the attachment ratio increases more biotin is attached, displacing formerly bound FA/G-LC at new biotinylated attachment sites, and occluding further FA/G-LC-binding there. At higher levels of biotinylation, destabilization brought about by the loss of FA/G-LC binding is more than compensated for by the increase in stability afforded by the increased amount of attached biotin, resulting in a more stable structure with increased t_m and greater that approaches that of HSA_99-NS(x:1) for x > 20. Thus, the behavior of HSA_96-NS(x:1) can be summarized as follows. The attachment reaction results in conversion of the HSA_96-H species to HSA_96-L. At the lowest attachment ratios (x < 20), the HSA_96-L-NS(x:1) structure is stabilized by FA/G-LC binding and attachment of biotin. At higher attachment ratios stability of the biotinylated structure increases, but FA/G-LC is released resulting in a conversion of the structure from HSA_96-L-NS(x:1) to one more like HSA_99-NS(x:1). If this is the case, the difference in the values for HSA_96-NS(x:1) and HSA_99-NS(x:1) in Equation (11), at the lowest attachment ratio is a direct gauge of the added stabilization afforded by FA/G-LC binding to the HSA_96-L ≈ HSA₉₉ structure.

As mentioned previously, published melting studies of HSA reported observations, at temperatures approaching t_m, of a second reaction attributed to irreversible aggregation of the protein in the denaturation region [1] [12] [13] [15] . They noted the potential existence of a slow irreversible denaturation step(s) but concluded potential contributions to the overall measured thermogram were minimal. Irreversibility occurs due to intermolecular association of denatured HSA polypeptide strands. Such aggregation is not readily reversible in the transition region and can affect reversibility of the HSA melting transition [16] . Influence of irreversible denaturation on the primary melting transition of HSA can be minimized (somewhat) through appropriate choices of experimental parameters [12] [13] [16] . In effect, under optimally designed experimental conditions the temperature scanning rate is such that kinetics of the irreversible denaturation reaction are relatively much slower, and only begin to contribute significantly to the HSA melting transition at temperatures above the t_m. Since our experiments of HSA were carried out under similar conditions to those previously reported [1] , it was assumed the measured HSA melting transitions are, for the most part, reversible. Inherent in the analyses that were performed was the choice of the “universal” reference state, i.e. the denatured state. Therefore, the possibility of heterogeneous denatured states, differing in character in addition to their ability to reversibly renature from the denatured state to the native state, is not considered.

Contributions of the irreversible transition to the measured ΔC_p values can also manifest in the thermogram baseline and therefore be reduced or subtracted out (somewhat) in the baseline analytical procedure. The irreversible reaction could also manifest in evaluated thermodynamic parameters of the melting transition, which assumes a truly equilibrium (reversible) process with parameters evaluated from the area under the baseline corrected thermogram. This process might influence absolute values of thermodynamic transition parameters, ΔH_cal and ΔS_cal, which may be lower (or higher) than what has been reported previously, depending on how much of the baseline encroaches on the melting of transition of HSA from the native to denatured form, and how much irreversibility affects the transition and choices made as part of the baseline analysis procedure. In our analysis as much as 30% of the total endotherm of HSA can be absorbed in the baseline. Our baseline analysis treatment was performed in a systematic fashion in precisely the same manner for all thermograms. Within the inherent

uncertainties underlying the operative analysis, relative values of the thermodynamic parameters are considered to be more quantitatively significant. Fortunately the analysis that was performed involved relative comparisons of the transition thermodynamic parameters.