This study examined the influence of pre-processing conditions on the formation and composition of the PC around GNPs with the ultimate goal of facilitating the identification of new molecular targets (Fig. 1).
Enrichment of proteins in cell lysates by urea extraction and in serum/plasma by immunocentrifugation
Initially, ovarian cancer cells and ovarian cancer patient plasma/serum samples were used as protein sources for the formation of PC around GNPs. Prior to PC formation, preliminary studies were performed to characterize protein types and levels in cell lysates and plasma samples. To obtain proteins from cancer cells, two different cell lysis buffers were used: (i) the commercial RIPA buffer, and (ii) urea. The protein patterns derived from using the two buffers were compared. Ovarian cancer cells (TykNu) were cultured in media containing 10% fetal bovine serum, lysed with either RIPA or urea and the lysates were analyzed by gel electrophoresis. Equivalent amounts of protein, determined using the BCA method (Fig. 2a), were separated by SDS-PAGE and visualized using Coomassie Blue. There were significant differences in the relative molecular weights of proteins generated by RIPA and urea lysis (Fig. 2b). These data suggest that cell lysis conditions play a critical role in determining the types of protein generated. These results are supported by the report of Marini et al. who showed that when the soluble and insoluble fractions of mitochondria from cancer cells are enzymatically digested by trypsin, Glu-C, or chymotrypsin, each lysis condition produces a distinct protein signature (Marini et al. 2020; Nierenberg et al. 2018).
Human plasma/serum, including the plasma/serum from ovarian cancer patient used herein, generally consists of 90–95% HAPs and 5–10% LAPs. The LAPs fraction represents a potential source of biomarkers/NMTPs. Since HAPs can mask LAPs from detection by mass spectrometry, HAPs are generally depleted by passage through a HAP-specific binding column prior to analysis (Millioni et al. 2011). Thus, HAPs were initially depleted (> 90%) by passage of patient plasma/serum through columns containing immobilized HAP-specific antibodies. The binding efficiency of the HAPs to the columns, for serum and plasma, respectively, was 91.6 ± 4 and 91.7 ± 5.5%, showing that HAPs are efficiently depleted by a single pass through the column (Fig. 2c). The depletion of HAPs was confirmed by SDS-PAGE, and demonstrated the absence of HAP bands across the molecular weight range including bands corresponding to the following specific HAPs: apolipoprotein A-II (17.4 kDa), apolipoprotein A-I (28.3 kDa), haptoglobin (40 kDa), α1-acid glycoprotein (41–43 kDa), α1-antitrypsin (52 kDa), albumin (66.5 kDa), transferrin (80 kDa), and IgA (160 kDa). LAPs bands were enriched in the HAPs-depleted flow through (Fig. 2d). These data confirm that LAPs can be enriched and detected by pre-processing, which would theoretically enhance identification of potential NMTPs.
Characterization of the protein corona arising from the interaction of gold nanoparticles with cell lysates, plasma, or enriched low-abundance proteins from plasma
Next, PCs were formed by the interaction of the various protein preparations with 20 nm GNPs to determine any observable differences in their physicochemical properties, i.e., size and charge. 20 nm GNPs were synthesized by the citrate reduction method (Hossen et al. 2019) and characterized using UV–Visible Spectroscopy (UV–Vis), dynamic light scattering (DLS), ζ-potential measurements, transmission electron microscopy (TEM) and the cyquant proliferation assay. A surface plasmon resonance (SPR) band around 522 nm in the UV–Vis spectrum indicates the formation of spherical ~ 20 nm GNPs (Additional file 1: Fig. S1A–C) (Hossen et al. 2019). The hydrodynamic diameter (HD) and surface charge of GNPs were determined by DLS and zeta potential measurements, respectively; GNPs had an HD of 24.7 ± 3.0 nm and a net negative charge of 42.4 ± 3.2 mV (Fig. 3a, b). TEM confirmed that GNPs of approximately 22.5 ± 3.4 nm in diameter were synthesized by this method (Additional file 1: Fig. S1B, C). Previously, we demonstrated that GNPs possess unique self-therapeutic properties (Arvizo et al. 2013). GNPs bind to a number of heparin-binding growth factors (HB-GFs) at their HB domain and inhibit HB-GF function by altering protein conformation; GNPs of 20 nm had the highest efficacy. This inhibition of protein function resulted in inhibition of tumor growth in both ovarian and pancreatic cancer. Thus, proteins that bind to GNPs may have critical roles in tumor growth and GNPs could be used as a tool to identify such proteins as potential NMTPs. Therefore, before proceeding with NMTP identification, it was important to verify the self-therapeutic property of the GNPs. The self-therapeutic property of GNPs to inhibit proliferation was tested against TykNu ovarian cancer cells. GNPs decreased proliferation of TykNu cells in a dose-dependent manner, as evaluated by the cyquant proliferation assay, indicating that the synthesized GNPs are biologically active (Fig. 3c). This result was consistent with our previous finding that GNP inhibits ovarian cancer cell proliferation (Arvizo et al. 2013; Xiong et al. 2014). Moreover, the findings of others also support the concept that GNPs can be utilized not only to identify therapeutic targets in various diseases, such as diabetic retinopathy, macular degeneration, and rheumatoid arthritis, but also to inhibit angiogenesis in these models (Kim et al. 2011; Tsai et al. 2007).
The synthesized and characterized GNPs were incubated with the RIPA and urea cell lysates to determine if the generated PCs differed in their formation and composition. The two lysates were individually incubated with GNPs at various amounts, and the HD and charge of the resulting PC-coated GNPs were determined. When GNPs were incubated with increasing amounts of the lysates (i.e., 1, 5, 10, 25, 50, 100 and 200 µg), their HD increased (Fig. 3d). Following incubation with the RIPA lysate at the increasing protein amounts, the GNPs had size distributions with z-averages of 74.5 ± 2.1, 69 ± 9.5, 70.5 ± 14.2, 78.3 ± 2.9, 79.7 ± ,1.9, 79 ± 3.8 and 62.2 ± 2.7 nm; for the urea lysate, the sizes were 90.5 ± 8.2, 115.4 ± 3.8, 90.3 ± 4.6, 93.9 ± 3.8, 85.9 ± 4.2, 97.5 ± 2.2 and 74.9 ± 2.0 nm. The increased size of GNPs following incubation with lysate proteins indicates PC formation, and is consistent with the findings of others showing that when biologically active NPs are incubated with biological fluids, they become covered with a protein layer and have an associated increase in size (Brun et al. 2014). In addition, the charge of the NP is a second parameter that indicates whether proteins have bound to the surface. Naked GNPs had a ζ-potential of − 43.6 ± 1.2 mV; the urea lysates alone had charges of -12.7 ± 2.4, − 28.5 ± 3.7, − 31.6 ± 2, − 28.1 ± 5.0, − 31.1 ± 2.7, 25.6 ± 1.46 and − 24.7 ± 2.37 mV; while, the RIPA lysates had charges of − 15.5 ± 2.2, − 33.9 ± 5.9, − 36 ± 9.2, − 31.3 ± 5.6, − 25.4 ± 5.7 and − 25.4 ± 2.7 at the increasing protein amounts, respectively (Additional file 1: Fig S1A–D). Interaction of GNPs with increasing protein amounts of the urea lysate resulted in charges of − 28.7 ± 1.3, − 22.9 ± 2.6, − 18.6 ± 1.0, − 15.7 ± 1.2, − 13.3 ± 1.7, − 8.4 ± 0.5 and − 6.7 ± 0.7 mV, respectively; while, the corresponding values for GNPs treated with RIPA lysates were − 29.6 ± 1.5, − 23.5 ± 1.7, − 19 ± 1.6, − 15.8 ± 0.9, − 13.3 ± 1.3, − 12.7 ± 1.2 and − 9.6 ± 1.8 mV, respectively (Fig. 3d, e). These results suggest that proteins gradually bound on the surface of GNPs, thereby reducing the charge, and the charge fell more on the addition of increasing protein amounts. Taken together, the increase in size paired with the decreasing charge of GNPs when incubated with lysate proteins demonstrates that the lysate proteins successfully bound on the GNP surface to form a PC.
To further confirm the formation of a PC around the GNPs, aggregation studies were performed by treating the GNP–protein mixture with 150 mM NaCl. Treatment with 150 mM NaCl disrupts the repulsive electrical double layer around unmodified NPs, thereby inducing aggregation. The ability of NaCl to aggregate NPs decreases when the NP surface is coated with proteins (Giri et al. , 2014; Hossen et al. 2020). NaCl was added at a concentration of 150 mM to the formed urea-PCs around GNPs, and the size and surface charge were measured before and after NaCl addition. The ζ-potential of the uncoated GNPs decreased from − 42.4 ± 3.2 to − 20.8 ± 7.8 mV and their size increased from 24.7 ± 3.0 to 584.9 ± 85.0 nm upon the addition of NaCl, indicating their aggregation (Fig. 3f). However, the size of the GNPs with PCs did not significantly change on addition of NaCl (86 ± 2.9, 130.2 ± 10.4, 95.8 ± ,7.5 90.6 ± 2.3, 88.9 ± 5.7, 94.7 ± 14.3 and 78 ± 2.2 nm); in contrast, the surface charge of PCs decreased at the 1–50 µg protein levels (− 9 ± 3.7, − 8.8 ± 2.3, − 7.6 ± 0.5, − 7.5 ± 0.5, − 9.5 ± 0.6 mV) but not at 100 or 200 µg (− 7.4 ± 1.4 and − 6.6 ± 0.8 mV) (Fig. 3f). These aggregation studies demonstrated that there were no appreciable changes in either size or charge following NaCl treatment of the PC-coated GNPs derived from incubation with 50–200 μg of protein, indicating that 50 μg of lysate proteins is the saturation concentration for 1 ml of 20 nm GNP as synthesized.
Similarly, the GNPs incubated in the original plasma samples and the plasma-derived eLAPs had size distributions of z-average 76 ± 3.2 and 41 ± 1.5 nm and ζ-potentials − 7.3 ± 0.5 and − 6.3 ± 2.8 mV, respectively (Fig. 3g, h). The reduced size of the eLAPs incubated GNPs, i.e., 41 versus 76 nm, may result from the attachment of comparatively smaller proteins around GNPs from the eLAPs; the slightly decreased charge indicates the eLAPs probably lead to more stable PCs. Aggregation studies with NaCl showed no appreciable change either in size or charge, suggesting stabilization of the GNP surface by the eLAPs proteins (Fig. 3f). Moreover, it is also likely that a dynamic equilibrium is reached where high affinity proteins are enriched on the GNP surface replacing the low affinity HAPs resulting in decrease in size and charge.
Impact of cellular protein extraction process on protein corona composition around gold nanoparticle
Following the confirmation of a stable PC and its physicochemical characterization, the composition of the PCs formed following incubation of GNPs with 200 µg of either urea- or RIPA-lysate proteins was determined using nanoLC–MS/MS. Unbound proteins were removed by centrifugation and GNPs were washed once with water prior to analysis, and equal amounts of bound protein were trypsin-digested for analysis by nanoLC–MS/MS. Nine-hundred and fifty seven common proteins were identified, 232 proteins were found in only the RIPA-derived PCs and 628 proteins in only the urea-lysate PCs. The details of the identified proteins are provided in Additional file 2: Tables S1 and Additional file 3: Table S2. Many of the proteins were significantly enriched (2- to 22-fold) in the urea sample compared to RIPA (Fig. 4a–c). The common proteins (957) are most likely abundant proteins and served as a detectable lysate pool to compare the differential properties of the attached proteins and to assess the enrichment of proteins on the GNP surface. In addition to these common lysate proteins, the LAPs extracted by RIPA and urea are bound on the GNP surface according to their affinities. These LAPs were classified into three gene ontology (GO) categories: cellular components, biological functions, and molecular functions (Fig. 4b–d and Additional file 1: Fig. S2). Urea-lysate proteins for biological processes were mainly related to localization, stimulus, metabolic processes and reproduction; whereas among the cellular components, envelope, microbody, mitochondria, and membrane were most abundant. In addition, electron carrier activity is a major pathway related to molecular function (Fig. 4b–d and Additional file 1: Fig. S2). The average abundance of the peptides from these proteins was higher in the urea-lysates than in the RIPA (Fig. 4e). Thus, cellular protein extraction using urea may increase the abundance of NMTPs.
Identification of new molecular target proteins in the protein corona around gold nanoparticle from enriched low-abundance proteins
To analyze PC composition, plasma from ovarian cancer patients was selected in preference to serum since plasma proteins are more stable than serum proteins. Initially, the relative amount of protein in human plasma (HP), the HP-derived PCs, and the eLAPs-derived PCs was assessed by SDS-PAGE. The PCs from eLAPs showed enrichment of various proteins across the molecular weight range (Fig. 5a and Additional file 1: Fig. S3). NanoLC–MS/MS was used to characterize the proteins in eLAPs, and in the PC of GNPs incubated in either human plasma or eLAPs. A total of 170 proteins were identified; 18 new proteins were enriched in eLAPs, and an additional four proteins were identified in the PC around GNPs incubated with eLAPs. The four proteins identified in eLAPs-PCs were gasdermin-B, dermcidin 2, phospholipid transfer protein isoform c and complement C4-B preproprotein, and their presence suggests enrichment of specific proteins on the surface of GNPs. The detailed list of the identified proteins is provided in the supplementary data (Additional file 4: Table S3). The differences in the protein composition of the various samples are shown in the Venn diagram in Fig. 5b, and most of the promising proteins for future clinical application are shown in Fig. 5c. In addition, the heatmap directly compares the protein composition of the samples in terms of fold change of abundance of protein matched peptides (Fig. 5d). To avoid a bias towards the ratio greater than 1, the log of the concentration ratio is shown and allows the data of ratio less than 1 to be compressed between 0 and 1, while the data for fold changes are symmetric (Fig. 5e). Taken together, these data demonstrate that pre-processing of human plasma by immunocentrifugation, followed by incubation with GNPs to generate a PC, will allow enrichment of LAPs by approximately 13%, thus facilitating the identification of NMTPs that are either undetectable or at very low abundance in ovarian cancer patients’ plasma.
Proteins that were enriched in the PC around GNPs incubated in eLAPs included gasdermin-B (47 kDa), dermcidin isomer 2 (16 kDa), kallistatin (58 kDa), phospholipid transfer protein isoform c (80 kDa), EGF-containing fibulin-like extracellular matrix protein 1 (55 kDa), and selenoprotein p isoform 1 (50–60 kDa) (Fig. 5b-f and Additional file 1: Fig. S4). All of these proteins are involved in tumorigenesis (Hergueta-Redondo et al. 2014; Bancovik et al. 2015; Albers et al. 2012; Zhou et al. 1992; Gao et al. 2010; Wang et al. 2020; Han et al. 2017; Seeliger et al. 2009; Chao et al. 2017). Proteomic analysis confirmed that these proteins were not detected in the original plasma samples, possibly due to their low abundance. Since GNPs have high affinities for specific proteins, the eLAPs may concentrate around the surface of GNPs. For example, gasdermin-B, dermcidin isomer 2 and phospholipid transfer protein isoform c were undetectable in either the original plasma or the eLAPs themselves, but they were found in the PC derived from eLAPs.
Gasdermin-B is a 47-kDa protein which is expressed in several tumor types, including hepatocarcinoma, and gastric, cervical and breast cancers, and its over-expression is related to tumor progression (Hergueta-Redondo et al. 2014). Over-expression of dermcidin isomer 2, a 11-kDa protein, occurs in multiple human tumor types, including melanoma, cutaneous tumors, and breast, prostate, pancreatic, and hepatocellular carcinomas (Bancovik et al. 2015). Phospholipid transfer protein isoform c (PLTP) is associated with lipoprotein metabolism and lipid transport in the vascular compartment (Albers et al. 2012). Kallistatin is a serine proteinase inhibitor (serpin) that has diverse functions in apoptosis, inflammation and tumorigenesis (Zhou et al. 1992; Gao et al. 2010) and EGF-containing fibulin-like extracellular matrix protein 1 is a potential biomarker for diagnosis, prognosis and therapeutic assessment in osteosarcoma, glioma, bladder cancer, human pancreatic adenocarcinoma and pleural mesothelioma (Wang et al. 2020; Han et al. 2017; Seeliger et al. 2009). In addition, the contribution of selenoprotein p isoform 1 (SELENOP) to tumor formation and progression has recently been established (Chao et al. 2017). Most of these proteins identified by the proteomic analysis of PCs have not been studied in ovarian cancer. Thus, these findings create an exciting avenue of further study to validate and apply the identified NMTPs to the management of ovarian cancer.