2.1. Patient Selection and Ethical Aspects

Blood samples from 180 patients with CRC and 70 healthy volunteers were obtained. Samples were collected between 2012 and 2016, prior to surgery, and aliquots were stored at −80 °C until processing. Moreover, as the duration of cryopreservation could potentially affect the protein profile, as it was previously evidenced by us [13], cases and controls were also rigorously matched for the length of cryopreservation period. The CRC serum samples from patients were prepared from blood specimens collected before the start of any anticancer treatment. The samples were collected in the Basque Biobank for Research following a strict and standardized protocol to ensure the appropriate processing and storage that preserves a correct protein profile. Samples were obtained from peripheral blood. Tubes were centrifuged at 2500 rpm for 20 min at room temperature and placed in 500 μL aliquots.

The assessment of metastatic and non-metastatic patients was based on the “Protocol for the Examination of Resection Specimens From Patients With Primary Carcinoma of the Colon and Rectum” of the College of American Pathologists (2020) [14].

All the experiments carried out in this study comply with the current Spanish and European Union legal regulations. Samples and data from patients included in this study were provided by the Basque Biobank for Research—OEHUN (http://www.biobancovasco.org/). All patients at Basurto University Hospital were informed about the potential use for research of their biological samples and accepted this eventuality by signing a specific document approved by the Ethical and Scientific Committees of the Basque Country Public Health System (Osakidetza) (CEIC 11-51).

2.2. Solid Phase Extraction

Functionalised super-paramagnetic beads (DynabeadsVR RPC 18 Life Technologies Dynal, Irvine, CA, USA) with a C18 alkyl-modified surface were used to prepare a low molecular weight (LMW) protein fraction from serum. Serum sample preparation was performed as described in detail in a previous paper [15]. Some modifications were applied to the DynabeadsVR RPC 18 manufacturer’s protocol. Briefly, 40 μL of the bead suspension were washed once with water and thrice with 100 μL of 200 mM NaCl, 0.1% TFA. The beads were re-suspended in 20 μL of water, mixed with 50 μL of serum and incubated at room temperature for 5 min. After incubation, the tube was placed in a magnetic concentrator (Dynal MPCVR, Invitrogen Dynal, Oslo, Norway), and the supernatant was discarded. The bead-peptide complex was washed thrice with 300 μL of 0.1% TFA in water, and the bound peptides were eluted by incubation at room temperature for 2 min with 12 μL of a 1:1 acetonitrile/water solution containing 3.5 pmol/μL of a synthetic internal standard peptide (MW 1419.76). This standard peptide, which was later used to normalize the data, was directly spiked in the eluting solution in order to avoid any interference during the binding reaction between analytes and the magnetic beads.

2.3. API-MALDI/TOF MS Analysis

The target plates for the API-MALDI/TOF analysis were prepared following the standard procedure previously cited [4]. Briefly, one volume of the eluted sample was mixed with two volumes of premade matrix solution (6.2 mg of CHCA in 1 mL of 36% methanol, 56% acetonitrile and 8% water). Next, 1 μL of this mixture was spotted in quadruplicate onto the MALDI target plate and allowed to dry at room temperature in a vacuum desiccator. API-MALDI/TOF analysis was performed in positive mode on a 6210 Time of Flight LC/MS (Agilent, Santa Clara, CA, USA) coupled with an atmospheric pressure PDF-MALDI Ion Source (Agilent) equipped with a 337 nm nitrogen laser. The following voltages were applied: fragmentor 300 V, skimmer 60 V, OCT RF 300 V. The acquisition laser power was set at 35% of maximum (peak power 75 kW, pulse energy 300 mJ). Data acquisition was automated using Mass Hunter software (Agilent) and programmed to accumulate 600 shots per spectrum. The irradiation program was automated using the spiral motion control of the PDF-MALDI ion source. The instrument was externally calibrated using Tuning Mix (Agilent). The nominal resolution of the instrument was 20.000 (17.000 observed), and the nominal mass accuracy (with internal calibration) was <2 ppm.

2.4. Nano-LC Mass Spectrometry

An aliquot of the DynabeadsVR RPC 18 extracted peptides was analysed by nano-HPLC-MS/MS using an Ultimate 3000 nano-HPLC system managed by the CHROMELEON Chromatography Data System (CDS) software (Thermo Scientific, Oslo, Norway) connected to a Hybrid Quadrupole-Orbitrap mass spectrometer (Q Exactive, Thermo Scientific).

The peptides were re-suspended immediately before analysis and were first loaded onto a trapping column (Acclaim PepMap C18, length 20 mm, diameter 0.1 mm, particle size 5 μm, pore size 100 Å, Thermo Fisher Scientific, Oslo, Norway) using the loading solvent (5–95% acetonitrile/water + 0.05% trifluoroacetic acid) at a flow rate of 5 μL/min for 5 min. The trapping column was then switched in-line with the separation column, and the peptides were eluted with increasing organic solvent at a flow rate of 300 nL/min. Mobile phase A was 0.1% formic acid in water; mobile phase B was 95–5% acetonitrile/water + 0.08% formic acid. The separations were carried out at 35 °C using an Easy spray column (PepMap RSLC C18, length 150 mm, diameter 0.075 mm, particle size 3 µm, pore size 100 Å, Thermo Scientific) applying a linear gradient from 4% to 95% of eluent B in 55 min.

All the analyses were carried out in the positive ion mode in a mass range between 395 and 2000 m/z. Single MS survey scans were performed in the Orbitrap using a maximal ion injection time of 100 ms. The resolution was set to 70,000, and the automatic gain control was set to 3 × 106 ions. The experiments were performed in data-dependent acquisition mode with alternating MS and MS/MS experiments. The minimum MS signal to obtain MS/MS was set to 500 ions, with the most prominent ion signal selected for MS/MS using an isolation window of 2 Da. The m/z values of signals already selected for MS/MS were put on an exclusion list for 5 s using dynamic exclusion. In all cases, one micro-scan was recorded. Collision-induced dissociation (CID) was done with a target value of 5000 ions, a maximal ion injection time of 50 ms, normalized collision energy of 35%. A maximum of 10 MS/MS experiments/MS scan were performed.

Raw MS files were processed with the Thermo Scientific Proteome Discoverer software version 1.4. Peak list files were obtained by the SEQUEST search engine against the Homo Sapiens database containing both forward and reversed protein sequences. The resulting peptide hits were filtered for a maximum 1% FDR (false discovery rate) using the percolator tool. The peptide mass deviation was set to 10 ppm, and a minimum of six amino acids to identify peptides were required. The database search parameters were: mass tolerance precursor 20 ppm, mass tolerance fragment CID 0.8 Da with dynamic modification of deamidation (N, Q), oxidation (M). For all searches, the option trypsin with two missed cleavages was selected.

2.5. API-MALDI/TOF MS Data Format Conversion

Raw data generated by the Qualitative Analysis software (Agilent Technologies, Santa Clara, CA, USA) were converted from the proprietary format (.d) to the mzML standard format by using MSconvert, a tool of the software suite ProteoWizard [16]. Four spectra, corresponding to the quadruplicated spots, were then determined for each sample by averaging 15 consecutive scans from the original data by means of the MALDIquant R package [17].

2.6. Sample Selection

All serum samples were tested for their integrity by means of SeraDeg [11], a publicly available web tool aimed to assess the quality of sera through a comparative analysis of the contents, in the related MS spectra, of fibrinopeptide A (fpA) fragments, a molecule which is known to be easily degraded under physico-chemical stress conditions. For each sample, SeraDeg provides quality scores (SeraDeg quality score—SDQS) that allow researchers to assess the quality of samples and discard those samples that do not have an adequate quality.

This score is mainly based on the overall abundance of fpA fragments, which is compared to the overall abundance of the same molecules in a reference spectrum derived from a sample of assessed quality and the ratio between the percentages of high-mass (m/z 1350.6 and 1465.63) and low-mass (m/z 905.46 and 1077.53) fpA fragments.

The spectra created during the previous step by using MALDIquant were exported in text format (two columns: one for the m/z value, one for the relative abundance) and included in a compressed zip archive in order to analyse them with SeraDeg. The following thresholds were applied for the selection: overall abundance of fpA fragments greater than 33.33% of the overall abundance in the reference spectrum, percentage of high-mass fpA fragments greater than 66.67% of the percentage of low-mass fpA fragments.

According to the results of this analysis, only a subset of the samples that were initially acquired was finally used in the following statistical analysis.

2.7. Spectra Pre-Processing

Lists of peaks were determined for each spectrum by applying a two-step procedure. All peaks were first determined by selecting local maxima (i.e., values in the given spectrum higher than both the previous and the following values). The final list of peaks was then determined by applying a median filter (half window size of 20 and signal-to-noise ratio of 2.5).

The lists of peaks were then analysed by means of Geena 2, a publicly available web tool for pre-processing MALDI/TOF spectra that, among other functions, is able to:

• Normalize both abundances and adjust m/z values against a reference signal;

• Sum up isotopic peaks of the same molecule;

• Average technical replicas from each sample, thus producing a single spectrum that is representative for the sample;

• Align representative spectra and produce a table of aligned signals with corresponding abundances.

Both the average and alignment steps above allow discarding those signals that are included only in a few spectra, so that sporadic signals can be removed from the output. For the average step, we adopted a high threshold (at least three spectra out of the four replicas must include the signal for its inclusion in the alignment). For the alignment step, instead, we had to define a low threshold in order to prevent that signals present only in one of the groups under analysis were discarded. The smaller group led this choice. We set this threshold to five: only signals found in at least five representative spectra were included in the final alignment.

2.8. Statistical Analysis

The output generated by Geena 2 was then analysed by means of SAMR, in order to identify signals significantly different in the groups, i.e., healthy volunteers, patients affected by metastatic CRC and patients affected by non-metastatic CRC. The multiple regression model [18] was applied to peptides eventually selected by the SAM analysis to take into account the age and gender differences between the two groups of patients and controls. The parameters estimated by this model can be interpreted as the difference in the mean values of a peptide between patients and controls. Given the small size of our sample, we chose to obtain a robust estimate of the standard errors of the model parameters by using a non-parametric bootstrap.

3.1. API-MALDI/TOF MS Analysis

To determine the low-molecular-weight (LMW) protein profile spectra, the whole sample set was analysed by API-MALDI/TOF mass spectrometry performed on the LMW fraction of sera. The good quality of our MS spectra can be appreciated in Figure 1.

Open in a separate window

Figure 1

Serum sample spectra. The figure includes two spectra: healthy control (A) and metastatic colorectal cancer (CRC) patient (B). The intervals around the fragment at m/z 1561.725 have been zoomed (i.s.: internal standard).

In order to avoid procedural biases, samples were randomly distributed during processing. All spectra were acquired in quadruplicate in a mass range from 800 to 3000 m/z.

3.2. Selection of Samples

After the acquisition, the API-MALDI/TOF spectra were tested with SeraDeg in order to avoid possible biases due to altered preservation conditions. As previously reported, this operation selected a set of samples of adequate quality and preservation including 172 pathological samples, of which 92 patients had been diagnosed with metastatic CRC, 80 patients had been diagnosed with non-metastatic CRC, and 19 samples were from healthy volunteers (Table 1).

3.3. Selection of Candidate Signals

The sample set selected by SeraDeg was used as a discovery cohort to identify possible candidate peptides able to distinguish controls from patients with reference to metastatic spreading. After the pre-processing steps carried out by means of Geena 2 [20], a total of 122 peaks were detected. Among these, 19 peaks were found to be related to fibrinopeptide A or B and were not considered in the following analyses. Indeed, fpA fragments were used, as previously stated, by SeraDeg for the selection of less degraded samples and consequently taking them into account could erroneously drive the final outcome. The number of useful signals was finally reduced to 103. The MS Excel files including alignment data generated by Geena 2 as prepared for the SAM analysis for the comparison between controls and non-metastatic samples and between controls and metastatic samples are included as Supplementary Material Tables S1 and S2.

The significance of the SAM analysis was assessed by using the q-value that is similar to the well-known p-value adapted to the situation where a large number of comparisons is carried out, which is our case since we evaluated 103 peptides. The significance threshold for the q-value is 5%. When we compared the three groups of samples, the SAM analysis highlighted only three signals (m/z 1561.72, 2021.08 and 2399.17) having a q-value of 5% or less either in the comparison between controls and non-metastatic samples or between controls and metastatic samples (see Table 2). No significant signals were found in the comparison between samples from non-metastatic and metastatic patients.

Indeed, a signal at m/z 1051.71 was also highlighted by the SAM analysis as a significant signal (q-value < 5%) in both comparisons, with a fold change of 2.5 for controls versus non metastatic samples and of 2.7 for controls versus metastatic samples. However, this signal is known to be derived from a polymer contaminant from either polypropylene or polyethylene plastic tubes [21]. Consequently, it has been excluded from any further consideration.

Only one of the other signals, the one at m/z 1561.72, was significantly higher in non-metastatic CRC patients when compared with healthy controls (fold change 16.4). When the healthy controls were compared with the metastatic CRC patients, beyond the 1561.72 signal (whose fold change increased to 17.8), the other two signals (at m/z 2021.08 and 2399.17, with fold change 2.9 and 6.9, respectively) were also found significantly increased. These results are also represented in the form of dot plots in Figure 2 and Figure 3.

Open in a separate window

Figure 2

Dot plots representing abundance values for controls and non-metastatic patients. From top to bottom, plots refer to signals m/z 1561.72, 2021.08 and 2399.17. Dots on the left (black) represent controls, dots on the right (red) non-metastatic patients. The horizontal lines show the average values. Dots are randomly spread along the abscise axis for visibility. Scales are adapted to values.

Open in a separate window

Figure 3

Dot plots representing abundance values for controls and metastatic patients. From top to bottom, plots refer to signals m/z 1561.72, 2021.08 and 2399.17. Dots on the left (black) represent controls, dots on the right (red) metastatic patients. The horizontal lines show the average values. Dots are randomly spread along the abscise axis for visibility. Scales are adapted to values.

In Figure 2, the comparison between controls and non-metastatic patient is shown, while the comparison of controls and metastatic patients is shown in Figure 3. For each of the three significant signals, m/z 1561.72, 2021.08 and 2399.17, the abundance values of the signal in controls’ and patients’ spectra are shown separately.

3.4. Identification of the Selected Serum Peptides by Nano-HPLC-ESI-MS/MS

In order to identify the origin of the selected candidate signals, a tandem mass spectrometry experiment was performed analysing a CRC serum sample that has one of the highest abundances of these molecules. Following the submission of the data obtained by the LC separation and MS/MS analysis to the SEQUEST search engine [22] against the Uniprot Homo Sapiens database, the amino acid sequences of peptides m/z 1561.72 and m/z 2021.08 were, respectively, determined, with high confidence, as belonging to prothrombin (Uniprot ID: P00734, sequence: “TATSEYQTFFNPR”) (Figure S1) and to the fragment C3f of complement system C3 (Uniprot ID: P01024, sequence: “SSKITHRIHWESASLLR”) (Figure S2), while the signal 2399.17 was identified as a double α-CHCA matrix adduct (with molecular weight 189.04 Da) to the peptide at m/z 2021.08 (C3f). A signal at m/z 2210.13 corresponding to a single α-CHCA matrix adduct to the C3f peptide at m/z 2021.08 was also identified, but it was not considered significant by the SAM analysis.

We want to thank Erasmus+ Programme for contributing to Beitia’s mobility, allowing her to carry out experiments at the San Martino Hospital Laboratories.