Identification of protein changes in the blood plasma of lung cancer patients subjected to chemotherapy using a 2D-DIGE approach

A comparative analysis of blood samples (depleted of albumin and IgG) obtained from lung cancer patients before chemotherapy versus after a second cycle of chemotherapy was performed using two-dimensional difference gel electrophoresis (2D-DIGE). The control group consisted of eight patients with non-cancerous lung diseases, and the experimental group consisted of four adenocarcinoma (ADC) and four squamous cell carcinoma (SCC) patients. Analyses of gels revealed significant changes in proteins and/or their proteoforms between control patients and lung cancer patients, both before and after a second cycle of chemotherapy. Most of these proteins were related to inflammation, including acute phase proteins (APPs) such as forms of haptoglobin and transferrin, complement component C3, and clusterin. The variable expression of APPs can potentially be used for profiling lung cancer. The greatest changes observed after chemotherapy were in transferrin and serotransferrin, which likely reflect disturbances in iron turnover after chemotherapy-induced anaemia. Significant changes in plasma proteins between ADC and SCC patients were also revealed, suggesting use of plasma vitronectin as a potential marker of SCC.


Introduction
Lung cancer is the most common cancer in the world and responsible for most cancer-related mortality worldwide [1]. Symptoms of lung cancer are usually very difficult to recognise until the disease is in an advanced, non-curable state. It is estimated that only 16% of all patients will survive five or more years after diagnosis. Late diagnosis is a significant factor contributing to poor lung cancer prognosis [1]. For this reason, the development of biomarkers for effective prognosis is of utmost importance [2]. PLOS  Biomarkers are biological compounds that can be used to distinguish a pathological from a normal status. Several biomolecules are used as potential cancer biomarkers, such as DNA, methylated DNA, RNA, miRNA, low molecular weight metabolites, xenobiotics, and proteins [3][4][5]. For example, circulating miRNAs, which are short, noncoding RNA molecules, can bind and interfere with mRNAs that are important for tumour expression pathways, and are also used for the detection of lung cancer [6]. Long, noncoding RNAs are also involved in tumourigenesis [7]. At the molecular level, proteins represent the most important functional unit that is directly responsible for a phenotype. For this reason, almost all Food and Drug Administration (FDA) approved cancer biomarkers are proteins [1,8].
There are several potential noninvasive and convenient sources for biomarker identification, such as body fluids, including serum or plasma, urine, sputum, tears, pleural effusion, and volatile organic compounds in exhaled breath concentrate [9]. Blood plasma is a convenient, noninvasive, inexpensive, and clinically relevant source of substances that can be screened for potential biomarkers. So far, several serum biomarkers for lung cancer have been identified, including fragments of cytokeratin 19 CYFRA 21-1, carcinoembryonic antigen (CEA), squamous cell carcinoma antigen (SCC-Ag), stem cell factor (SCF), neuron-specific enolase (NSE), progastrin-releasing peptide (ProGRP), epidermal growth factor receptor (EGFR), and vascular endothelial growth factor (VEGF) [1,8]. Unfortunately, the performance of individual markers has been disappointing due to their low sensitivity and specificity [1]. Therefore, it is unlikely that a single biomarker for any particular form of cancer can be identified; rather, a multi-marker approach is recommended [10][11]. For this reason, a search for new potential biomarkers is highly justified in order to identify potential candidates for incorporation into multi-marker algorithm biomarkers [12]. Two-dimensional difference gel electrophoresis (2D-DIGE) is variation of two-dimensional electrophoresis (2DE), and has found to be the preferred method for proteomic analysis of lung cancer due to its high reproducibility, sensitivity, comprehensiveness, and high throughput [13][14]. In 2D-DIGE, samples are labelled with distinct fluorescent dyes before electrophoretic separation; the two main approaches include minimal labelling and saturation labelling [15]. The former is a highly sensitive technique that employs fluorophores (Cy5, Cy3 and Cy2) with affinity to primary amino groups, covering only a small fraction (~3%) of each protein [16]. The use of three fluorophores enables electrophoretic co-separation of the sample and control, thereby greatly reducing spot matching errors within one replicate (one gel), and facilitates the creation of an internal pooled standard prepared from the mixture of all specimens used in the analysis. Saturation labelling was introduced by Kondo et al. [17] and is based on the complete labelling of sulfhydryl residues of cysteines using the Cy3 and Cy5 dyes. Due to its high sensitivity, saturation labelling is especially useful when sample amounts are limited, such as when using microdissected tissues [17]. However, only two fluorophores are currently available for this method, and therefore two gels must be run in order to maintain the internal standard concept [18]. For most 2D-DIGE studies of lung cancer, minimal labelling has been employed, and this approach has been recommended for serum proteomics [19]. However, saturation labelling has been successfully used as well [20][21][22].
The objective of this study was to compare protein abundance in blood plasma samples obtained from cancer patients before chemotherapy and after the second cycle of chemotherapy, using the 2D-DIGE approach. The control group was selected from patients with noncancerous lung diseases. Moreover, we sought to determine whether particular lung cancers (adenocarcinoma [ADC] or SCC) are related to differences in blood plasma protein abundance.

Sample collection
Informed consent was obtained from eight control patients (four men and four women, age range 52-76 years, mean age 63.2 ± 8.4 years) and eight non-small cell lung cancer (NSCLC) patients (four diagnosed with SCC and four diagnosed with ADC) before and after a second cycle of chemotherapy (five men and three women, age range 55-81, mean age 66 ± 8.2 years). Blood was collected from NSCLC patients one week before the first cycle of chemotherapy and one week before the third cycle of chemotherapy. The obtained venous blood samples were collected into tubes containing EDTA, and plasma was separated by centrifugation at 1,300 × g for 20 min at room temperature. Next, the plasma was transferred to sterile test tubes and re-centrifuged at 3,000 × g for 15 min at room temperature to remove residual cellular components. The plasma was transferred to cryotubes and stored at -80˚C until the day of analysis. Table 1 shows the clinical characteristics of the samples included in this work. The study was approved by the Ethics Committee of the Medical University of Bialystok (No. R-I-003/ 262/2004), and informed written consent for specimen collection was obtained from each patient before chemotherapy.

Immunodepletion of albumin and IgG from plasma
Blood plasma was depleted using albumin (Alb) and IgG depletion spin traps (GE Healthcare, Uppsala, Sweden). Blood plasma (50 μL) was combined with 50 μL binding buffer (0.15 M NaCl buffered with 20 mM phosphate buffer, pH 7.4). The samples were then applied to spin columns equilibrated with binding buffer. After a 5-min incubation period to allow the binding of Alb and IgG, the unbound protein fraction (depleted blood plasma)  [23]. The labelling reaction was performed in the dark on ice for 30 min. Experimental samples of blood plasma from the control patients and lung cancer patients before and after a second cycle of chemotherapy were labelled with Cy3 and Cy5 according to the scheme presented in Table 2.
For 2D DIGE protocols [24], the calculated minimum number of gels to be run for our 2D DIGE experiment for 8 patients is 12 gels. This was calculated according to the formula: No: of gels ¼ no: of groups � no: of biological reps 2 ðsince two groups on each gelsÞ For our experiment, the number of groups was 3 (control, before chemotherapy, and after chemotherapy), and the number of biological replicates (patient samples) was 8, so number of gels was (3 x 8)/2 = 12.
Cy2 dye was used to label a pooled sample comprising equal amounts of each of the samples within the experiment, and acts as an internal standard. An equal amount of Cy2-labelled pooled standard was loaded on each gel for normalisation and to correct for gel-to-gel variability. After the labelling reaction, differentially labelled samples (50 μg of Cy2, Cy3, and Cy5-labelled samples) were mixed together according to the scheme presented in Table 2. Rehydration buffer was then added to each sample mixture to reach a final volume of 450 μL. Differentially labelled samples were then loaded on 24-cm Immobiline DryStrips, with a 3 to 10 nonlinear gradient pH range (GE Healthcare) and rehydrated for 12 h. Proteins were then separated by isoelectric focusing and SDS-PAGE, as described above.

Image acquisition and analysis
After electrophoresis, the gels were scanned with a Typhoon 9500 FLA scanner (GE Healthcare) using the parameters suggested by the manufacturer for 2D-DIGE experiments. The scanned images were analysed with DeCyder Differential In-Gel Analysis version 5.02 software (GE Healthcare) to identify the fluorescence intensities of the spots. The DeCyder biological variation analysis module was used to detect protein spots, simultaneously matching all 24 protein spot maps from 12 gels using the following parameters: the estimated number of spots was set to 10,000 and the minimum spot size was set to 3,000. Protein spots with a p-value <0.05 by one-way ANOVA analysis, which showed an increase or decrease in relative intensity, were considered to be differentially abundant proteins. Only spots that were successfully matched on >80% of the gel images were considered. To properly select and identify the spots, gels were stained using CBB-G250 after 2D-DIGE, followed by spot excision and identification using matrix-assisted laser desorption/ionisation time-of-flight/time-of-flight (MALDI-TOF/ TOF) mass spectrometry (MS).

Protein identification by mass spectrometry
Protein spots indicated by statistical analysis were excised from the gels, put in Eppendorf tubes, and washed with 50 μL of 50 mM ammonium bicarbonate. The wash was discarded and the spots were washed again with 50 μL of 50 mM ammonium bicarbonate in 50% acetonitrile solution, and incubated for 5 min in room temperature. After discarding the wash and drying the spots, 2 μL of 0.2 μg/μL modified sequencing grade trypsin (Promega, Madison, WI, USA) solution and 2 μL of 50 mM ammonium bicarbonate were added and the samples incubated for 12 h at 37˚C. After digestion, the spots were placed in 100 μL of 0.1 trifluoroacetic acid (TFA) and desalted with Zip-Tip C-18 pipette tips (Millipore, Billerica, MA, USA; [25]). Each Zip-Tip was first washed with 100% acetonitrile and then equilibrated with 50% acetonitrile in

Validation of 2D-DIGE results by Western blot
Western-blot technique was used to validate the results obtained in proteomics study. We used V3 stain-free workflow which eliminates the need for stripping and reprobing the blot for housekeeping proteins [26]. The expression of three proteins of interest was evaluated in serum (i) of control patients and lung cancer patients before and after second cycle of chemotherapy (transferrin, fibrinogen α chain), as well as (ii) lung cancer patients before chemotherapy in relation to SCC and ADC (vitronectin). The Western blot was performed as described by Repetto et al. [27] with some modifications. Equal amounts of protein (10 μg) were fractionated on 12% Criterion™ TGX Stain-Free™ Protein Gels (Bio-Rad, Hercules, CA, USA). After electrophoresis, gels were activated on a Chemidoc according to manufacturer instructions (Bio-Rad), then transferred to PVDF membranes using Mini Trans-Biol Cell (Bio-Rad) at 60 V for 90 min. After transfer, a stain-free image of PVDF membranes for total protein normalization was obtained before membranes were rinsed briefly in distilled water and blocked with 5% bovine serum albumin (Sigma-Aldrich, St. Louis, MO, USA), then incubated with primary polyclonal antibodies (Abcam, Cambridge, UK) against transferrin (1:10000), fibrinogen (1:5000), vitronectin (1:1000) overnight at 4˚C. After rinsing the membrane to remove unbound primary antibodies, it was exposed to goat anti-rabbit antibodies (1:5000; Sigma) linked to alkaline phosphatase. Products were visualized by incubation in a solution of alkaline phosphate buffer with an addition of NBT (Sigma) and BCIP (Sigma) in the dark. Antibody-bound proteins were detected by enhanced chemiluminescence using the Chemidoc Imaging System (Bio-Rad). All band intensities were measured with Image Lab Software Version 5.2 (Bio-Rad, Hercules). The image of the gel acquired before its transfer was used as control for equal protein loading among samples. The volume density of each target protein band was normalized to its respective total protein content, whereas total protein band was normalized to the total protein loaded into each lane using stain-free technology with data expressed in arbitrary units.

Functional analysis
Ingenuity pathway analysis (IPA; IngenuityR Pathway Analysis, IPAR, Qiagen, Redwood City, CA, USA) software was used to investigate the functional and canonical pathways that were enriched in the differentially expressed proteins (http://www.ingenuity.com). Fisher's exact test and Benjamini-Hochberg multiple testing corrections were used to calculate statistical significance (p < 0.05).

Statistical analysis
Statistical analysis of changes in protein abundance was performed using the Biological Variance Module of DeCyder Differential In Gel Analysis version 5.02 software (GE Healthcare) on eight biological replicates (individual patients). Direct comparisons of spot volumes were made between the Cy3-or Cy5-labelled samples and the Cy2-labelled pool standard for each gel. The Cy3/Cy2 and Cy5/Cy2 ratio was used to calculate average changes in abundance. Data are expressed as log standardised abundances to ensure a normal distribution of the data. One-way ANOVA, t-test and the average ratio test were performed; changes in protein spot abundance were considered statistically significant at p < 0.05. For the MS PMF and MS/MS ion search, statistically significant (p � 0.05) matches by MASCOT were regarded as correct hits.

Depletion of albumin and IgG from blood plasma
Depletion of Alb and IgG from blood plasma decreased the amount of protein by 73%, from 3.02 ± 0.27 mg applied to the spin trap column to 0.83 ± 0.09 mg recovered after depletion (n = 24). Electropherograms indicated that the Alb and IgG fractions were not visible by CBB staining in the depleted samples (Fig 1).

2D-DIGE analysis of differentially expressed proteins in the blood plasma of control patients and lung cancer patients before chemotherapy
Comparison of all patients. Out of 32 differentially abundant spots, we identified 24 differentially expressed proteins or proteoforms (eight spots could not be identified) in the blood plasma of control patients compared with all NSCLC patients before chemotherapy (Table 3 and Fig 2). Control plasma was characterised by a higher abundance of complement C3, coagulation factor XII, fibrinogen β chain, prothrombin isoform 2, gelsolin isoform e, proapolipoprotein, inter-α-globulin inhibitor H4, α-2-HS-glycoprotein, α-2-macroglobulin isoforms a and b, and protein SP40,40. On the other hand, the plasma of lung cancer patients before 2D-DIGE analysis of blood plasma of lung cancer patients after chemotherapy  chemotherapy was characterised by a higher abundance of fibrinogen α chain, zinc-α-glycoprotein precursor, five proteoforms of haptoglobin, and orsomucoid 1. Although the samples included both cancer stages I and III proteomic changes were similar regardless of stage. We have provided examples with magnified regions of all gels for samples at stage I and stage III lung cancer with differential protein expression, as well as line charts of the selected spots (charts generated from DeCyder software) in S1 Table. These data confirm similar changes in the blood proteome regardless of cancer stage. Comparative analysis of ADC and SCC patients. We re-analysed our results taking the specific lung cancer diagnosis into consideration. Eight spots discriminating the proteome of SCC patients from that of ADC patients were identified ( Table 4). The plasma of SCC patients was characterised by a higher abundance of vitronectin, coagulation factor XIII, plasminogen, and gelsolin. On the other hand, the plasma of ADC patients was characterised by a higher abundance of transferrin, immunoglobin heavy chain constant region mu, and leucine-rich alpha-2-glycoprotein precursor.

2-DIGE analysis of differentially expressed proteins in the blood plasma of control patients and lung cancer patients after second cycle of chemotherapy
Comparison of all patients. We identified 41 differentially expressed proteins or proteoforms in the blood plasma of control patients compared with NSCLC patients after the second cycle of chemotherapy (Table 3). However, 13 proteins or proteoforms were the same as those detected before chemotherapy. The control plasma contained higher abundances of proapolipoprotein, coagulation factor XIII, clusterin, fibrinogen α chain, hemoglobin beta chain, interα-trypsin inhibitor, α-2-macroglobulin isoform b, protein SP40,40, four proteoforms of transferrin, and three proteoforms of serotransferrin X1. On the other hand, the plasma of lung cancer patients after chemotherapy was characterised by a higher abundance of two forms of apolipoprotein A-IV precursor, four forms of fibrinogen α chain, five forms of hemopexin, haptoglobin, α-2-macroglobulin isoform b, immunoglobulin heavy chain constant α 1 membrane bound isoform 1, and leucine-rich α-2 glycoprotein precursor. Comparative analysis of ADC and SCC patients. We identified four spots discriminating the proteome of SCC patients from that of ADC patients (Table 4). Three proteins discriminating SCC from ADC after a second cycle of chemotherapy were confirmed, including coagulation factor XIII and gelsolin. Transferrin and haptoglobin hp2 abundance was higher in the blood plasma of ADC patients than in that of SCC patients.

2-DIGE analysis of proteins differentially expressed in the blood plasma of lung cancer patients before and after second cycle of chemotherapy
Comparison of all patients. The average numbers of protein spots (mean ± SD) and CV (%) in the control group, in lung cancer patients before chemotherapy, and in lung cancer patients after the second cycle of chemotherapy were 1375 ± 136 (9.9%), 1341 ± 156 (11.6%), and 1270 ± 100 (7.9%), respectively. We identified seven differentially expressed proteins or proteoforms in the blood plasma of NSCLC patients before chemotherapy compared with after the second cycle of chemotherapy (Table 3). Plasma before chemotherapy was characterised by a higher abundance of complement C3 preproprotein, hemoglobin β chain variant S-Wake, two variants of transferrin, and two variants of serotransferrins isoform X1. After a second cycle of chemotherapy, blood plasma contained a higher abundance of hemopexin precursor.
Comparative analysis of ADC and SCC patients. Two proteins were differently expressed in ADC patients before chemotherapy and after the second cycle of chemotherapy; A negative average ratio denotes a higher abundance of proteins in patients before chemotherapy, whereas a positive average ratio denotes a higher abundance of proteins in cancer patients after second cycle of chemotherapy. hemopexin was more abundant before chemotherapy and fibrinogen gamma was more abundant after chemotherapy (Table 5). For SCC patients, five detected spots were more abundant after chemotherapy, including the haemoglobin beta chain, complement C3, transferrin, and two variants of serotransferrins. Using Western blot we confirmed further corroborate the decrease in transferrin in serum of lung cancer patients after a second cycle of chemotherapy ( Fig 3A); an increase in content of fibrinogen α chain in lung patients before and after second cycle of chemotherapy ( Fig 3B) and decrease in the content of vitronectin in serum of lung cancer patients before chemotherapy in relation to SCC and ADC (Fig 3C). The images of the gels after SDS-PAGE electrophoresis and membranes after transfer and before incubation with primary antibodies are shown in S1 Fig. Changes in protein abundance across individual patients. S2 Table summarises the observed changes in expression of selected proteins (493, 887, 1019, 1572) across gels with blood plasma from lung cancer patients before the first cycle and after the second cycle of chemotherapy. Generally, proteins, including transferrin (spot no. 493) changed in the same way in samples before and after the second cycle of chemotherapy. However, outliers could be found for each of the presented spots originating from different samples, for example 55263 Cy5 and 55260 Cy5 for spot no. 493, 55269 Cy5 for spot no. 887, 55262 Cy3 for spot no. 1019, and 55262 Cy3 and 55263 Cy5 for spot no. 1572; this could reflect patient-specific protein expression patterns.

Ingenuity pathway analysis
A summary of the IPA for differentially expressed proteins in blood from control and lung cancer patients before chemotherapy is provided in Table 6. The most significant enriched canonical pathways included "acute phase response signaling", "FXR/RXR and LXR/RXR activation", and "coagulation system". Figs 4-7 depict differentially expressed proteins mapped to the most significant enriched canonical pathways. The molecular and cellular function lists included "cell-to-cell signalling and interaction", lipid metabolism", "molecular transport", and "free radical scavenging". The physiological system development and function lists included "haematological system development and function" and "immune response". Table 6. Functional analysis (IPA) of differentially abundant proteins between the blood plasma of control patients and lung cancer patients before chemotherapy.

Ingenuity Canonical Pathways Molecules
Acute Phase Response Signaling  HP, APOA1, C3, ORM1, ITIH4, AHSG, FGB, FGA, A2M, F2   LXR/RXR and FXR/RXR Activation  APOA1, C3, APOA4, ORM1, ITIH4,  2D-DIGE analysis of blood plasma of lung cancer patients after chemotherapy A summary of the IPA pathway analysis for blood proteins enriched in control and lung cancer patients after second cycle of chemotherapy is provided in Table 7. The top canonical pathways associated with the identified blood proteins included "FXR/RXR and LXR/RXR activation", "acute phase response signaling" and"clathrin-mediated endocytosis signaling" (Figs 4-6 and 8), while the molecular and cellular function lists included "cell-to-cell signaling and interaction"," lipid metabolism", "molecular transport" and "free radical scavenging". The physiological system development and function lists included "haematological system development and function" and "immune response".

Discussion
In this study, the 2D-DIGE separation was performed on depleted blood from control patients and lung cancer patients before the first and after the second cycle of chemotherapy. Analyses of gels revealed significant changes in the type and abundance of proteins and/or their proteoforms between control patients and lung cancer patients, both before and after chemotherapy. Significant changes in the type and abundance of proteins and/or their proteoforms between ADC and SCC patients were also observed.
The depletion procedure used in this study was successful, because no albumin nor immunoglobin was detected on the gels, with the exception of immunoglobin heavy chain constant α 1 membrane bound isoform 1, which is recognised as a tumour-related cell membrane protein [28]. Moreover, other proteins that tend to be removed together with albumin (called the 2D-DIGE analysis of blood plasma of lung cancer patients after chemotherapy albuminome) such as apolipoproteins, clusterin, complement inhibitor, clusterin, haptoglobin, hemopexin, leucine-rich α-2 glycoprotein, and transferrin [29] were not lost during purification, because they were identified in the depleted serum (Table 3). However, our results demonstrated clear changes in the abundance of classical or highly abundant blood plasma proteins, as defined by Strohkamp et al. [19]. These proteins include transferrin, 2-macroglobulin, haptoglobin, leucine-rich α-2 glycoprotein, fibrinogen, apolipoprotein A-IV, and clusterin. Given the large range (powers of 10) in the circulating concentrations of various proteins in blood plasma [29], the detection of these proteins was expected due to their high concentrations within the detection limits of 2D-DIGE.
Differences in the numbers of identified proteins between different studies employing 2D-DIGE of blood plasma proteins likely reflect differences in methodological approaches and the composition of the control and cancer patient groups. Differences in methodologies mainly concern the extraction of blood proteins for 2D-DIGE. For example, Wen et al. [30] only used the glycoprotein fraction of serum, which was obtained with the use of ConA affinity columns. Most studies removed high-abundance proteins from serum to varying degrees. For example, Wen et al. [30] removed five abundant proteins, including albumin, IgA, IgG, transferrin, and HP. Okano et al. [22] and Dowling et al. [31] removed six of the most abundant proteins (albumin, transferrin, haptoglobin, alpha-1-antitrypsin, IgA, and IgG). On the other hand, Rodríguez-Piñeiro et al. [32] removed 20 of the most abundant proteins (albumin, transferrin, α1-acid glycoprotein, complement C1q, IgG, fibrinogen, ceruloplasmin, complement C3, IgA, α2-macroglobulin, apolipoprotein A-1, complement C4, IgM, α1-antitrypsin, apolipoprotein A-II, plasminogen, IgD, haptoglobin, apolipoprotein B, and prealbumin). Clearly, significant differences in depletion methodology do exist and can be a significant factor in the comparative interpretation of the results of 2D-DIGE studies of the proteome of lung cancer patients. This calls for the development of standardised procedures for the preparation of serum samples for 2D-DIGE analysis.
Nevertheless, the results of our study are quite consistent with previous reports indicating the aberrant expression of plasma proteins in lung cancer [33]. To our knowledge, six studies of blood plasma from cancer patients focusing on a comparative proteomics approach using 2D-DIGE have been published, including Okano et al. [22,34], Dowling et al. [31], Hoagland et al. [35], Rodríguez-Piñeiro et al. [32] and Wen et al. [30]. Several proteins were identified 2D-DIGE analysis of blood plasma of lung cancer patients after chemotherapy across all studies (alpha-2-HS-glycoprotein or haptoglobin); however, proteins exclusive to a single study were also reported (alpha-2-macroglobulin isoform A and B, coagulation factor XIII, fibrinogen alpha chain and beta chain, IG heavy chain α 1 membrane bound isoform, orsomucoid 1, and zinc alpha 2 glycoprotein). It is worth mentioning that, out of the five potential plasma cancer biomarkers identified using a very different methodology (profiling of plasma proteome with monoclonal antibody libraries) [36], four proteins (alpha-1-antichymotrypsin, haptoglobin, complement C9, and leucine-rich α-2 glycoprotein precursor) were also identified using 2D-DIGE studies. Therefore, electrophoretic and immunological methods for the detection of changes in the blood plasma proteins of lung cancer patients produce similar results.
The results of our study and previous studies employing 2D-DIGE [30][31][32]34,35] clearly indicate that 2D-DIGE can be an effective method to identify proteins related to inflammation, especially acute phase proteins (APPs). These proteins include haptoglobin and its various forms, complement component C3, clusterin, and serum amyloid A [37][38]. Moreover, changes in other serum proteins are clearly indicated in this study and similar studies. Dowling et al. [37] used ELISA analysis developed specifically for five blood proteins to define abundance trends for different cancers. In this study, 2D-DIGE analysis indicated 41 proteins or their proteoforms had differential abundance (Tables 3 and 4) and it is reasonable to assume that these proteins should be targeted for future individual measurements to develop more powerful tools for cancer profiling.
It has long been recognised that there is a link between cancer and inflammation. Inflammation is known to be both a cause and a consequence of cancer [39]. Therefore, a chronic inflammatory-like state is regarded as a hallmark of cancer and is associated with cancer development and disease progression [40][41], including lung cancer [42][43]. For a long time, the potential role of APPs has been underestimated and attributed to only representing cancer 2D-DIGE analysis of blood plasma of lung cancer patients after chemotherapy epiphenomena [37]. However, recent progress in proteomics studies strongly suggests that the variable expression of APPs can be used to profile the distinct types, subtypes, and even stages of cancer. Dowling et al. [37] was able to indicate different abundance trends for APPs in different cancers, including lung cancer; a similar approach was successfully employed by Wang et al. [44]. Together, these results strongly suggest the potential of APPs for cancer fingerprinting. It is also important to note that changes in APPs in response to cancer can be clearly demonstrated despite the presence of patients with non-cancerous diseases (causing inflammatory conditions) in control groups, which was indicated in this study and in the results reported by Rodríguez-Piñeiro et al. [32]. This strongly supports the specificity of cancer profiling using APPs. Therefore, 2D-DIGE represents a powerful tool to explore the signature of APPs in the blood plasma of lung cancer patients. It should be stressed that, although it is believed that APPs originate from the liver rather than from tumour cells, it is also possible that APPs can be directly produced by tumour tissue [44]. Therefore, further comparative studies are warranted to determine the origin of APPs in blood in order to better understand the relationship between APPs and cancer.
The results of our study clearly demonstrate that plasma proteins that are differentially expressed in lung cancer patients can differ in ways more nuanced than protein concentration. For example, both in this study and in that of Dowling et al. [31], several proteoforms of haptoglobin were identified (five in this study and four in Dowling et al. [31]). This suggests that changes in blood plasma proteins in response to lung cancer can be related not only to changes in their abundance, but also to post-translational modifications (PTMs). More broadly speaking, the large number of PTMs may relate to the quantitative and qualitative discrepancies between genomic, transcriptomics and their protein counterparts [45][46][47]. This suggestion is also supported by the recent indication of the importance of quantitative and qualitative discrepancies between genomic/transcriptomic alterations and their protein counterparts, mostly related to t. To date, the majority of the serum/plasma proteomics studies have focused on the measurement of the abundance of total proteins [48]; however, the research focus is now shifting towards studying the relationships between PTMs and cancer. These PTMs include glycosylation (including glycan-modified derivatives of haptoglobin) [35], fucosylation [49], phosphorylation, acetylation, arginine methylation, and lysine methylation [48][49][50][51], and several PTMs of histone proteins [52]. Our study provides a list of several forms of proteins that are altered in lung cancer patients, including proapolipoprotein, apolipoprotein AIV, clusterin, gelsolin, fibrinogen, haptoglobin (see above), hemopexin, transferrin, and serotransferin. Proteoforms can also contribute to mechanisms besides protein synthesis that are responsible both for the increases and decreases in protein abundance observed in this study and by other authors [30][31][32]34,35]. Additional studies are required to determine the exact nature of PTMs of these proteins and their usefulness for cancer profiling.
To our knowledge, we have identified, for the first time, proteins and their proteoforms that change in abundance before and after a second cycle of chemotherapeutic treatment for lung cancer. Most changes observed (4 out of 7) were in transferrin and serotransferrin. These changes likely reflect disturbances in iron turnover (iron deficiency) after chemotherapyinduced anaemia [53]. Because transferrin is an iron-binding protein, changes in its abundance in response to chemotherapy likely reflect disturbances in iron turnover. This suggestion is supported by recent findings indicating changes in transferrin levels due to chemotherapy [54][55]. The presence of transferrin proteoforms, demonstrated in this study, strongly suggests that PTMs contribute to the mechanisms of chemotherapy-induced changes in the blood plasma proteome. The identification of changes in haemoglobin abundance after chemotherapy can also be explained by chemotherapy-induced anaemia, because changes in haemoglobin abundance have also been reported [54][55]. Furthermore, changes in haemoglobin scavenger proteins, such as hemopexin, as reported in this study, are likely to be a part of the above-mentioned changes. The last protein identified in this study was complement C3 which others have documented as in the early host response to chemotherapy [56].
In this study, the number of differentially abundant proteins was higher in blood plasma obtained from lung cancer patients after a second cycle of chemotherapy compared to controls than it was in blood plasma obtained before chemotherapy compared to controls. This may be partially explained by the effects of chemotherapy itself on the blood proteome. On the other hand, the increase in several proteins likely reflects the progression of cancer development. For example, complement component 3, which can be a biomarker for chemotherapy (see above), has also been indicated as a prognostic factor for NSCLC [57][58]. For this reason, additional investigation is necessary to define the specific functions of plasma biomarkers, both for chemotherapy and cancer progression.
To our knowledge, this is the first study to directly compare the blood plasma proteome of ADC and SCC lung cancer patients using the 2D-DIGE approach [13]. Although the number of patients in this study was very restricted (four patients per cancer), our results clearly suggest that proteomics studies that profile ADC and SCC using a larger patient sample size are highly justified, because different sets of proteins can be specifically attributed to either ADC or SCC. It is especially interesting to identify vitronectin as a potential marker for SCC because this protein was identified in SCC patients before chemotherapy. This protein has been recently indicated as a potent migration-enhancing factor of cancer cells chaperoned by fibrinogen [59]. Therefore, additional studies are warranted to test the usefulness of vitronectin as a potential marker of SCC.
Our results clearly indicate the heterogeneity in profiles of particular proteins or their proteoforms in individual patients. This agrees with established knowledge that the performance of individual markers is variable due to their low sensitivity and specificity [1]. Therefore, it is unlikely that a single biomarker for any particular form of cancer can be identified; rather, a multi-marker approach is recommended [10][11].
In summary, our study has extended the list of potential lung cancer biomarkers. Our results emphasise the potential role of inflammatory proteins as biomarkers of lung cancer. Chemotherapy is accompanied by changes in proteins involved in anaemia. SCC can be distinguished from ADC using proteomics profiling, with a special emphasis on vitronectin. The presence of numerous proteoforms for several biomarkers warrants an investigation of the relationship between PTMs and cancer.
Supporting information S1 Table. The