An optimized MALDI MSI protocol for spatial detection of tryptic peptides in fresh frozen prostate tissue

Abstract MALDI MS imaging (MSI) is a powerful analytical tool for spatial peptide detection in heterogeneous tissues. Proper sample preparation is crucial to achieve high quality, reproducible measurements. Here we developed an optimized protocol for spatially resolved proteolytic peptide detection with MALDI time‐of‐flight MSI of fresh frozen prostate tissue sections. The parameters tested included four different tissue washes, four methods of protein denaturation, four methods of trypsin digestion (different trypsin densities, sprayers, and incubation times), and five matrix deposition methods (different sprayers, settings, and matrix concentrations). Evaluation criteria were the number of detected and excluded peaks, percentage of high mass peaks, signal‐to‐noise ratio, spatial localization, and average intensities of identified peptides, all of which were integrated into a weighted quality evaluation scoring system. Based on these scores, the optimized protocol included an ice‐cold EtOH+H2O wash, a 5 min heating step at 95°C, tryptic digestion incubated for 17h at 37°C and CHCA matrix deposited at a final amount of 1.8 μg/mm2. Including a heat‐induced protein denaturation step after tissue wash is a new methodological approach that could be useful also for other tissue types. This optimized protocol for spatial peptide detection using MALDI MSI facilitates future biomarker discovery in prostate cancer and may be useful in studies of other tissue types.

heating step at 95 • C, tryptic digestion incubated for 17h at 37 • C and CHCA matrix deposited at a final amount of 1.8 μg/mm 2 . Including a heat-induced protein denaturation step after tissue wash is a new methodological approach that could be useful also for other tissue types. This optimized protocol for spatial peptide detection using MALDI MSI facilitates future biomarker discovery in prostate cancer and may be useful in studies of other tissue types.

K E Y W O R D S
fresh frozen, MALDI mass spectrometry imaging, peptides, prostate tissue, proteomics INTRODUCTION Proteins carryout major processes in living organisms and aberrant protein regulation is a key player in disease progression [1]. Large-scale proteomics studies are therefore important in biomarker research.
Most forms of cancer are heterogeneous with tumor tissue containing different cell types and cancer cells present. Such tumor tissue contains spatial information that will be lost in most bulk proteomic methods.
Prostate cancer tissue is inherently heterogeneous, and only a few studies have successfully identified differentially expressed intact protein levels using MALDI MSI comparing cancer and benign fresh frozen prostate tissue [8][9][10]. Proteins identified as discriminating tumor from benign prostate tissue include mitogen-activated protein kinase/extracellular signal-regulated kinase 2 and overexpression of biliverdin reductase B [10,19]. In a recent study on fresh frozen prostate cancer tissue, MALDI MSI was used to detect different levels of tryptic peptides on cancerous tissues in comparison to benign tissue.
The authors reported two proteins (ribonuclease T2 and a heat shock protein) as potential biomarkers for aggressive prostate cancer tissue [5]. Additionally, other studies have found differential tryptic peptides patterns in formalin-fixed paraffin-embedded (FFPE) prostate cancer tissue [6,7,20]. The optimal protocol for detection of intact proteins or peptides with MALDI MSI depends on tissue characteristics and sample type (e.g. fresh frozen or FFPE tissue), the desired mass resolution, m/z interval, and sensitivity of the MSI instrument.
A critical part of MALDI MSI peptide detection is the use of an efficient, robust and reproducible sample preparation procedure optimized to the target tissue investigated [21,22]. Small deviations from an optimized sample preparation protocol can induce variations in mass intensity and spatial localization of analytes affecting the validity of the results. High-quality tissue collection and preservation are essential to limit analyte degradation and maintain the integrity of the tissue [22,23]. Commonly, fresh frozen tissue has been considered the most suitable material for analyses in MALDI MSI [24,25]. Fresh frozen tissue requires careful handling and storage at low temperatures (-80 • C) to avoid protein degradation and analyte delocalization [24][25][26]. An alternative to fresh frozen tissue is FFPE tissue, frequently implemented in clinical pathology and available in biobanks worldwide [6,7,20]. FFPE tissue preserves tissue morphology and provides easy long-term storage at room temperature (RT). However, removal of paraffin embedding and reversing the formalin-induced crosslinks (antigen retrieval) adds handling time to the sample preparation procedure and exposes the tissue for harsher mechanical and chemical treatment [24,27]. By using fresh frozen tissue, we limit additional tissue processing, thereby limiting technical variation in peptide detection in MSI experiments.
The main steps of proteolytic peptide detection with MALDI MSI on fresh frozen tissue samples typically include cryosectioning, washing the tissue sections to remove interfering compounds, on-tissue enzymatic digestion, matrix deposition and MALDI MSI data acquisition.
Most of these steps, if not all, need to be optimized for every tissue type analyzed. Several methodological variants of sample preparation for peptide detection with MALDI MSI have been described in literature, but there is no clear consensus on an optimal protocol for fresh frozen tissues in general [28].
The aim of this study was to establish an optimal protocol for reproducible analyses of tryptic peptides using high-speed MALDI MSI on fresh frozen prostate tissue samples for a subsequent high quality clinical study. We performed a series of experiments, following a linear experimental design, where new tests were adjusted based on previous results to achieve an optimized working protocol. Quantity and quality of peptide detection were evaluated by several criteria: number of peaks detected on tissue, number of excluded peaks (false positives), signal delocalization due to analyte diffusion/migration, spectral quality estimation by signal-to-noise ratio (S/N), and signal intensities of selected peptides.

Materials
All chemicals and organic solvents were purchased through Sigma-  Figure S1) and in total 125 sections were included in the study.
After cryosectioning, the slides were stored at −80 • C until analysis (maximum 2 months).

MALDI MSI Sample preparation workflow
The protocol workflow from tissue extraction and preparation until MSI analyses is presented in Figure 1. We followed a linear experimental design, where the selected protocol parameters to test were based on previous lab results and planned sequentially during the optimization period. At least three replicates were carried out per method. In total, 25 unique methods were tested and distributed accordingly.

Optimization of tissue wash
Prostate tissue sections on ITO slides were thawed at room temperature in a vacuum desiccator before washing with either Carnoy's,

Heating step
Various heating strategies were applied to test for potential effects on the following trypsin digestion. The heating step was performed after the washing step and four different experimental procedures were tested: 1) slides placed in a humidity chamber with a saturated K 2 SO 4 solution for 10 min at 70 • C [7]; 2) slides placed in a staining jar filled with saturated K 2 SO 4 (covering the bottom of the jar, not in contact with the tissue) and heated at 95 • C for 5 min under water-vapor; 3) using heat-induced antigen retrieval (AR) in a 2100 Retriever (Aptum Biologics Ltd., UK) or a kitchen pressure cooker with 10 mM citric acid (pH = 6.0) as AR buffer for 30 min at 121 • C, and 4) no heating step was applied. For procedure 3), the AR buffer was rinsed off with ddH 2 O (2 × 1 min) and the slides were dried in a vacuum-desiccator.

Trypsin digestion
Lyophilized trypsin was reconstituted immediately prior to usage in cold H 2 O to give a final concentration of either 0.02 or 0.1 μg/μL. To validate trypsin activity, 1 μL of Cytochrome C (CytC, 1 mg/mL) was spotted on each slide [30]. Trypsin was applied using a SunCollect sprayer (SunChrom, Friedrichsdorf, Germany) or an HTX M5 Sprayer (HTX-Technologies, Chapel Hill, North Carolina, USA), abbreviated "HTX sprayer" throughout the paper, using customized and optimized spray-

Matrix solution and deposition
Three different matrix solutions of α-Cyano-4-hydroxycinnamic acid (CHCA) were compared for the analysis of peptides: 5 mg/mL, 7 mg/mL or 10 mg/mL, all of which were dissolved in 50% (v/v) acetonitrile (ACN) incl. 0.2% TFA. CHCA matrix was sprayed with the two same sprayers mentioned above (SunCollect (SunChrome) and HTX TM-Sprayers (HTX Technologies)). Collectively for all the tests performed on the HTX sprayer (HTX Technologies), N 2 flow rate was always kept at 10 psi, the spray pattern was set to "criss-cross" and nozzle height was appointed to 40 mm by default (technical details are presented in Table 1). Settings were chosen to ensure homogenous layers of matrix and small, evenly sized crystals. Matrix density (W, μg/mm 2 ) is presented in Table 1 and was calculated by Supplementary Equation S1.

Data Processing
The CytC region was used as verification of trypsin digestion.
If CytC tryptic peptide peaks, such as m/z 1168.6±0.6 Da (peptide sequence TGPNLHGLFGR), were present, further analysis was performed. Each average spectrum from the defined on-tissue regions of interest (ROIs) was initially assessed by visual subjective inspection in flexImaging v.5.0 and peaks of interest were noted for later analyses.
The total ion current (TIC) normalized average spectra were baseline corrected and matrix spectra were subtracted from on-tissue spectra in mMass v.5.5.0 [31]. Peak picking was performed with a minimum S/N threshold of 3.0, a minimum relative intensity threshold of 0.1, a peak picking height of 100%, and isotopic peaks were removed.
Acquisition data were imported and processed in SCiLS Lab (v.2019b SCiLS). Manual visual evaluation of every peak in the spectra was performed in SCiLS to ensure that deisotoping was achieved correctly, and that duplicate peaks and peaks with abnormal shapes were excluded, creating a postprocessed peak list for further analysis. A manual count of peaks in the postprocessed spectra were addressed as "number of peaks." The forthcoming results in this paper is based on this postprocessed peak list. The "number of excluded peaks" was the sum of peaks subtracted from the preprocessed peak list as they either clearly belonged to the matrix, were isotopic peaks and/or technical noise.

Peptide identification with MS/MS
For peptide identification we used two methodological approaches. LC-MS/MS on peptide extracts was used for peptide sequencing.
Three fresh frozen serial sections from the same prostate tissue sample were treated with the ice-cold EtOH+H 2 O wash, heated at 95 • C for 5 min and sprayed with trypsin (T2 protocol in Supplementary Table   S1) and incubated overnight at 37.5 • C. Peptides were extracted from the top of two of the tissue sections with 2% ACN 0.1% acetic acid and 5% ACN 0.1% acetic acid solutions as described by Drake et al. [35].
Tryptic peptides solutions from these two serial tissue sections were The "target value" was set to 20 000 and dynamic exclusion was activated and set to 0.4 min. The quadrupole isolation width was set to 2 Th for m/z < 700 and 3 Th for m/z > 800. The third serial section was covered by CHCA matrix (protocol M1, Table 1) imaged with MALDI-TOF MSI and peak picking was perfomed as described above. These peaks were matched with MALDI-Orbitrap data to obtain high accuracy masses.

Quality evaluation
The evaluation measures used to determine the optimal method were:

Statistics
Established quality parameters were checked for normal distribution and means were compared by using two-tailed independent Student's t-test among the different testing steps using the total QE-score with a significance level of p≤0.05 in SPSS (IBM ® SPSS® Statistics, version

27) and Microsoft® Excel® for Microsoft 365 MSO (version 2008).
For non-normally distributed samples, nonparametric Mann-Whitney tests were performed in SPSS.

RESULTS AND DISCUSSION
We have developed an optimized protocol for spatial detection of proteolytic peptides in fresh frozen prostate tissue using MALDI-TOF MSI where a representative spectrum with selected peptide masses is shown in Figure 2. Peptide detection on fresh frozen tissue has generally been considered challenging within the imaging community [38] in comparison to MSI measurement of metabolites or lipids as those protocols require fewer sample preparation steps. A selection of previously published MALDI MSI protocols analyzing proteomics on fresh frozen tissue was used as a starting point and iteratively adapted by trial and error and further compared to detect spatial peptide distribution [2,9,12,13,15,21,39]. The key optimization steps tested were washing, heating, tryptic digestion, and matrix application (Supplementary Table S2

Tissue preparation (step 1)
Cryosectioning of fresh frozen prostate tissue from tissue cores of 10 μm thickness provided a more clean-cut tissue surface than 4 μm sections, as the tissue sections of 4 μm had a more crocked or irregular shape which subsequently affected the quality of the MSI images (Supplementary Figure S4). This could be explained by the texture and structure of the prostate tissue, the freezing or by the quality of the sectioning, as thicker sections are technically easier to cut.
However, if possible without the expense of section quality, thinner sections can improve analyte intensities in MALDI MSI experiments [40]. Thickness of 4 μm is more frequently used with FFPE tissue [41], while 10 μm thickness is more common for fresh frozen tissue in MSI experiments [42]. The majority of the experiments (21 of 25 protocols) were performed with a thickness of 10 μm due to higher sectioning quality observed during testing.

3.1.2
Optimizing tissue wash for improved spatial peptide localization (step 2) A selection of four washing procedures were tested and resulting MALDI images are presented in Figure 3. We also initially tested other  [5,6,39], but these were excluded from our study due to detrimental effects on the final HE images.
Overall, methods using EtOH+H 2 O wash gave on average 39% higher QE scores than methods using Carnoy's wash, which additionally resulted in alterations of the tissue morphology ( Figure 3A). Therefore, we further focused on the optimization potential of EtOH+H 2 O wash, initially by reducing the time of the H 2 O steps and later by alternatively using ice-cold EtOH and H 2 O with the purpose of minimizing delocalization. We observed improved spatial localization using the  Figure S5). However, on average, all three EtOH+H 2 O washes led to comparable localization and total QE scores when only including samples undergoing heat treatment in Step 3.

3.1.3
Brief heat treatment improved signal intensity (step 3) Of the four different heating procedures tested, the 5 min at 95 • C protein denaturation step where the slides were quickly brought to 95 • C under water-vapor had a significantly higher QE-score in comparison to protocols with no heat treatment (p ≤ 0.001; Figure 4A), 10 min at 70 • C in a humid chamber (p ≤ 0.001; Figure 4A) and in comparison to antigen retrieval (AR; p ≤ 0.05; Figure 4A). Of note, the sections run with AR showed a higher variability in the localization score between sections than 5 min at 95 • C or 10 min at 70 • C ( Figure 4C) and typically lacked mass in the higher mass range (> m/z 2000; Figure 4B), resulting in lower reproducibility. The protein denaturation step of 10 min at 70 • C scored higher (p ≤ 0.05) than no-heating, and AR resulted in significantly (p ≤ 0.001) higher scores than no heating and slightly higher scores than 10 min at 70 • C protocol ( Figure 4A).
Completely leaving out the heating step from the sample preparation procedure, gave the lowest localization scores ( Figure 4C) and typically lacked heavier masses (> m/z 2000 ; Figure 4B). Our results clearly show that including a heating step was beneficial for detection of peptide signals in general and by using a higher temperature of 95 • C for 5 min, increased the QE scores. Although the AR protocol provided data of higher quality than not using heating at all or heating for 10 minutes at 70 • C, the observed higher variability in localization and lack of higher masses, makes this method less reliable. This could be explained by the harsh conditions of submerging the tissue sections in a boiling acidic solution at 121 • C and the much longer incubation time of approximately 30 minutes. Rapid heating (95 • C for 5 min), also called heat-stabilization, is a common treatment of fresh tissue to inactivate proteins [43]. Heating the tissue results in irreversible denaturation of proteins through unfolding which prevents autolytic degradation and facilitates the access of trypsin for a more efficient digestion. The same effect was also observed by Zheng and DeMarco, where heating plasma samples at 99 • C for 5 minutes prior to trypsin digestion resulted in higher peptide signals from certain proteins while ensuring higher peptide stability compared to using no heating [44]. In our study, ice-cold tissue wash combined with a subsequent heating step at high humidity (method 6, Supplementary Table S3) gave the highest QE scores for localization and peptide signal. Although comparing the different washing steps could not demonstrate a clear benefit for icecold EtOH wash, we suggest that proteins might be fixed at their inherent position during the ice-cold wash and then being unfolded in-place through the rapid heating step, making them more accessible for enzymatic cleavage and thereby producing a higher number of peptides at their specific location.
As the protein denaturation step (heating step) was included later in the timeline, not all washing procedures were tested with this step. Therefore, we cannot exclude the possibility that the other washing procedures could be more effective when also including a heating step for peptide detection on fresh frozen prostate tissue sections.

3.1.4
Low trypsin density and overnight incubation produced high-quality measurements (step [4][5] There was no remarkable difference between applying slightly higher or lower trypsin density (1.3 ng/mm 2 vs. 6.7 ng/mm 2 ) judging from the QE scores of samples undergoing the heat treatment of 95 • C for  Figure S6). As using less trypsin is more costefficient, we preferred the trypsin application method T2 resulting in trypsin density of 1.3 ng/mm 2 . While others have suggested increased S/N with increasing trypsin concentration for FFPE kidney, heart and aorta tissue samples [14], this was not the case in our study. A higher trypsin concentration may also increase the degree of autolytic peptides from the trypsin, which could give rise to more unwanted signal interference [45].
When comparing all samples with short digestion (2 h at 50 • C) to all samples with overnight digestion (17 h at 37 • C), the total QE score for overnight digestion was on average significantly higher (22%, p ≤ 0.001, Figure 5C). Further, short digestion led to a higher number of noisy background peaks with similar intensities and a typical distribution of 1 Da spacing ( Figure 5A). Short digestion offered highly localized signals, though failed to detect peptides on tissue, especially above m/z 2000, which resulted in spectra of lower quality ( Figure 5A). MSI spectra of sections with overnight digestion showed a significantly higher percentage of peaks belonging to the upper mass range of the MSI spectrum (on average 215% higher score, p ≤ 0.001; Figure 5D).

3.1.5
Matrix solution and application (step 6) Developing the optimal matrix application protocol is crucial to ensure analyte extraction at their spatial localization. Due to the many param-

Peptide identifications
Of the masses identified from LC-MS/MS, 2 723 tryptic peptides from 838 proteins, a total of 24 pepetides from 12 proteins were also detected with MALDI MSI ( A maximum mass error of 5 ppm between detected m/z value from high-resolution MALDI-Orbitrap and theoretical m/z was used. Oxidation of methionine is indicated as "M(Ox)" in the peptide sequence. 1 Peptides of actin (m/z 945.552) and collagen α-1(I) (m/z 945.439) likely overlaps and is detected as one peak in the MALDI-TOF MSI data. 2 Peptides of collagen α-1(I) (m/z 868.428) and collagen α-2(I) (m/z 868.464) likely overlaps and is detected as one peak in the MALDI-TOF MSI data. 3 This mass may also be peptide sequence AAQDRDQIYR arising from transformer-2 protein homolog Β (P62995) Recent findings have addressed different types of collagens to be active components of reactive stroma in prostate cancer which contributes to disease progression [46,47]. Actin was another protein identified through detection of several peptides. As actin is a key component of muscle fibers and prostate stroma contains a high proportion of smooth muscle cells, this was a reasonable finding. Actin is also suggested as a prognostic marker in prostate cancer [48] and is along with collagens an interesting protein for further investigation as a clinical marker for prostate cancer. For the remaining 9 proteins identified, we could detect and assign one peptide each. Nevertheless, presence in prostate tissue is reasonable for all of them and functional roles in cancer have been reported previously [49][50][51][52][53][54]. The MS/MS results demonstrate that peptides of biological and clinical interest could be detected with our optimized sample preparation protocol.

Applicability of method
Although the described optimizations of sample preparation for peptide detection are performed on fresh frozen prostate tissue, the process of developing this protocol could be relevant for peptide detection in other tissue types or organs. From our experience, some degree of protocol optimization is always necessary and this study provides a guide on the critical steps to optimize. We show that the heating step at high humidity (as opposed to submerged in solution) was a necessity for improved MSI peptide detection and localization. Further, tissue wash was the second most crucial step to optimize for MALDI MSI peptide detection, followed by trypsin application and matrix deposition. The optimal washing procedure is also what we anticipate that will vary the most across different tissue types due to different molecular compositions and functional structure as has been pointed out before in intact protein analysis [21,39,41]. For instance, for tissues such as breast and liver, containing a high concentration of fat, stronger organic solvents may be required in order to wash away fatty compounds to improve peptide detection.

CONCLUDING REMARKS
This study presents an optimal sample preparation protocol for spatial peptide detection on fresh frozen prostate tissue using MALDI MSI.
Most importantly, we have shown that a 5 min heating step at 95 • C, directly performed after tissue wash, enhances the signal intensity on tissue and increases the quality of the MSI spectra. This protocol successfully managed to increase the number of detected peptides on tissue and provide their inherent spatial distribution minimizing delocalization. We have demonstrated that proper sample preparation is crucial to achieve high-quality MALDI MSI measurements ensuring reproducibility, which will facilitate future experiments aiming for the discovery of novel biomarkers for aggressive prostate cancer. Additionally, we provide an optimized protocol that highlights the critical performing steps in tryptic peptide analysis that could be useful for other protocol set-ups within the MALDI imaging community and also for general LC-MS peptide analysis.