Reduced Hemoglobin Signal and Improved Detection of Endogenous Proteins in Blood-Rich Tissues for MALDI Mass Spectrometry Imaging

Tissue preparation and MALDI protein imaging were performed as summarized in Scheme 1 on mouse liver, spleen, and human endometrium samples. In our optimized protocol, introduction of the lysis buffer step after alcohol fixation (70, 90, 95% EtOH) was chosen to ensure complete precipitation of proteins and prevent delocalization during the aqueous lysis buffer step. Further, additional washing steps in water, various concentrations of ethanol, and Carnoy’s fluid were necessary to ensure complete removal of lipid and salts for optimal protein detection prior to imaging by MALDI MS. To investigate the effect of lysis buffer washing on hemoglobin signal, 3 replicate sets of 5 serial sections of mouse spleen were imaged by MALDI MS after 0 (control), 2, 5, 10, and 15 minutes of washing in the lysis buffer ( Figure 1A and 1B ). The trend in average signal intensity decrease for hemoglobin α and β after various washing times was determined from the average signal of 50 pixels from 3 adjacent sections of mouse spleen tissue in a similar region of red pulp ( Figure 2A and 2B ). In mouse spleen, hemoglobin α ( m/z 15005) was observed to decrease in signal intensity by an average of 17±5% (n=3) after 2 minutes, 25±8% (n=3) after 5 minutes, 34±8% (n=3) after 10 minutes, and 37±9% (n=3) after 15 minutes. For hemoglobin β ( m/z 15705), a decrease in signal by 45±10% (n=3) was observed after washing for 2 minutes, 52±13% (n=3) after 5 minutes, 59±18% (n=3) after 10 minutes, and 61±8% (n=3) after 15 minutes. For hemoglobin α and β, while no significant difference was observed between the signal intensity of the control versus 2 minutes (p>0.05) and control versus 5 minutes (p>0.05), increasing the washing time to 10 minutes resulted in a significant decrease in signal intensity relative to the control sample (p0.05) ( Figure 2A and 2B ). Next, the effect of washing in lysis buffer was investigated in mouse liver tissue. The average signal intensity was determined from the average mass spectra of 50 pixels from 3 replicate sections in a similar region of tissue surrounding major areas of blood vessels ( Figure S1 and S2 ). A similar trend in signal intensity decrease was observed in mouse liver tissue compared to mouse spleen tissue. Thus, the results from both tissue types indicate that a washing time of 10 minutes is optimal to achieve a statistical decrease of hemoglobin signal while minimizing the amount of time required for sample preparation as no further decrease was observed when increasing the washing time from 10 to 15 minutes. A comparison of average mass spectra for spleen and liver before and after an optimized time of 10 minutes of washing in the lysis buffer is depicted in Figure 3A and 3B . For mouse spleen and liver, the data from a small region of 50 pixels was extracted and averaged in the red pulp region and outside of major regions of blood vessels, respectively, due to tissue heterogeneity. In these mass spectra, while direct hemolysis and a decrease in signal intensity of hemoglobin subunits was observed, importantly, washing with the lysis buffer did not hinder direct analysis of a variety of other endogenous proteins. For example, enhancement of COX7C signal intensity was observed in spleen ( m/z 5443) and liver ( m/z 5446) after 10 minutes of lysis by 51±13% (p<0.05) and 46±22% (p<0.05), respectively ( Figure 3A and 3B , Figure S3 and S4 ). Other examples of proteins in mouse spleen and liver with enhanced signal or retained intensity are summarized in ion images and box plots for selected ions as depicted in Figures 1A , S1 , S3 , and S4 . Thus, the optimized protocol was observed to effectively reduce signal from hemoglobin subunits after only 10 minutes of washing in lysis buffer, while retaining or enhancing detection of a variety of other endogenous proteins. Further, the number of protein signals was investigated comparing control and 10-minute washed samples for the 3 tissue types studied. In the mouse spleen tissue, we observed an increase in number of protein signals from 243±22 to 387±68 in the average molecular profiles for control and washed samples, respectively (intensity threshold = 0.2, n = 3 replicates, p = 0.02). Similarly, in human endometrium, we observed an increase in number of protein signals in the average molecular profiles from 297±29 to 401±17 for control and washed samples, respectively (intensity threshold = 1, n = 3 replicates, p = 0.006). However, for mouse liver tissue we observed a non-statistical decrease in the number of protein signals, where 451±82 and 261±32 protein signals were obtained from the control and washed samples, respectively (intensity threshold = 1, n = 3 replicates, p = 0.06). These results suggest that the number of protein signals can potentially increase after using the lysis washing protocol as observed in spleen and endometrium tissue. However, as the change in number of protein signals appears to be variable between tissue types and a non-statistical decrease was observed in the liver tissue, further studies are needed to investigate the effect of tissue composition and type on number of protein signals detected. In order to confirm that protein delocalization does not occur as a result of the aqueous buffer washing step, we performed MALDI MS imaging at a higher spatial resolution (50 μm) on a control and 10-minute washed sample of mouse spleen tissues ( Figure S5 ). Comparison of the ion images for the control and washed sample depict clear and distinct borders between red and white pulp histological regions of the tissue, strongly indicating that no observable delocalization is expected due to the lysis washing protocol. Lastly, to verify the release of hemoglobin from the tissue section into buffer solution, the lysis buffer was collected after washing a mouse spleen tissue section for 10 minutes for subsequent analysis. Briefly, after desalting and protein isolation on a C18 resin, the sample was co-crystalized with matrix, spotted on a MALDI target plate, and the molecular profile was obtained ( Figure S6 ). As expected, we observed high relative abundances of hemoglobin α and β proteins in the mass spectrum of the collected buffer, suggesting that the release of these proteins from the tissue section into the buffer occurred during the lysis washing step. Although a few other proteins were also detected in the mass spectrum obtained from the buffer solution, which could be from red blood cells, the high relative abundance of hemoglobin proteins suggests that the lysis washing step is largely selective and effective in the removal of hemoglobin proteins. Protein imaging of endometrial tissue can be challenging as the rich blood content and high ionization efficiency of hemoglobin subunits can result in significant ion suppression of other proteins. Thus, we explore the use of the optimized lysis buffer protocol to reduce hemoglobin detection from endometrium tissue. As distinct histological regions are not observable in endometrium tissue, average mass spectra over the entire tissue of adjacent sections before and after 10 minutes of washing in lysis buffer were compared ( Figure 3C ). After treatment with lysis buffer for 10 minutes, both hemoglobin α ( m/z 15123) and β ( m/z 15864) subunits were observed to decrease in signal intensity by 15±7% (p<0.05) and 27±5% (p<0.05), respectively ( Figure 4A – B , 5 ). Concomitantly, the relative abundance from other proteins such as histones H4 ( m/z 11305), H2B ( m/z 13788), and H2A ( m/z 14002) were observed to be retained or enhanced by 59±15% (p0.05), and 41±12% (p<0.05), respectively ( Figure 4C – E ). Other examples of proteins observed to increase in signal intensity include proteins COX7C ( m/z 5449, 67±15% increase, p<0.05), serine protease inhibitor kazal-type 3 (SPINK3, m/z 6120, 61±15% increase, p<0.05), and unidentified proteins at m/z 5043 (26±4% increase, p<0.05) and m/z 5935 (43±7% increase, p<0.05) ( Figure 4F – I , 5 ). As depicted in Figure 4J , hematoxylin and eosin (H&E) staining can be performed on the tissue samples after MALDI MS imaging of samples washed with lysis buffer indicating the developed protocol is histologically compatible. We also evaluated the effect of the hemolysis protocol on tissue histology by analyzing hematoxylin and eosin stained tissue sections using light microscopy before and after washing with the lysis buffer for 10 minutes. A reduction in red pigment and disruption in the extracellular matrix of blood containing regions after lysis buffer application could be qualitatively observed, suggesting successful washing of hemoglobin from tissue sections ( Figure 6 ). For example, in the spleen tissue, a reduction in red pigment and increase in purple pigment and was observed in stained sample after application of the lysis buffer for 10 minutes ( Figure 6A ). In liver tissue, reduction in redness can be visualized most clearly in regions of blood vessels ( Figure 6B ). Further, minimal disruption was observed in the epithelial and glandular cells of the tissue types studied suggesting high specificity of the lysis buffer for targeting red blood cells in tissue sections. For example, in Figure 6C , disruption in the extracellular matrix of blood can be visualized by the disordered histology outside of regions of endometrial glands. Overall, use of the lysis buffer for tissue section preparation appears to effectively target red blood cells, without disrupting other histological features.

The use of an ammonium chloride buffer directly on tissue sections to facilitate hemolysis and removal of erythrocytes prior to MALDI MS imaging was investigated. This protocol offers a viable approach to improve protein detection by MS imaging in tissues containing high blood content by reducing the susceptibility of endogenous proteins to ion suppression from hemoglobin. Treatment with a lysis buffer for 10 minutes was found to be a simple and effective method to reduce signal intensity of hemoglobin α and β subunits from tissue sections while retaining or enhancing signal from other proteins. We foresee that this protocol can further be applied for sample preparation to study molecular markers in a variety of diseases and tissue samples where substantial ion suppression from hemoglobin in molecular profiles is anticipated due to high blood content. Further, this tissue preparation could be applied to improve molecular imaging and detection in various cancerous tissues where significant ion suppression from hemoglobin may arise from extensive tumor vasculature.

Acetonitrile, acetic acid, chloroform, ethanol, trifluoracetic acid, and 2-hydroxy-5-methoxybenzoic acid/5-methoxysalicylic acid were purchased from Fisher Scientific (Waltham, MA). Mouse spleen and liver tissues were purchased from BioIVT (Westbury, NY). Human endometrial tissue was collected from patients undergoing endometriosis surgeries by Dr. Michael T. Breen at the Dell Medical School under approved IRB protocols from the University of Texas IRB and the Seton Family of Hospitals IRB. Unstained tissue sections were visualized on an Epson Perfection V600 Photo Scanner (Epson America, Inc.) at 4800 dpi resolution. Hematoxylin and eosin (H&E) stained tissue sections were visualized on a NanoZoomer-SQ digital slide scanner at 40× magnification (Hamamatsu). Mouse liver, mouse spleen, and human endometrium were sectioned at 12 μm on a CryoStar NX50 cryostat (Thermo Fisher Scientific) and mounted onto indium tin oxide (ITO) slides (Delta Technologies). Slides were placed in a desiccator to dry for 15 minutes prior to washing. Tissue sections were submerged in ethanol (70, 90, 95%) for 30 seconds each for to precipitate proteins and prevent delocalization during the aqueous lysis step. A hydrophobic barrier was drawn around each tissue section using a hydrophobic PAP pen (GeneTex), then an ammonium-chloride potassium (ACK) lysis buffer (Gibco) was aliquoted within the barrier for 2, 5, 10, or 15 minutes. After lysing, the tissue sections were washed twice in water for 30 seconds each, followed by a wash in ethanol (70%, 100%, 30 seconds each), Carnoy’s fluid (60% ethanol, 30% chloroform, 10% glacial acetic acid, 2 minutes), ethanol (100%, 30 seconds), water (30 seconds), and ethanol (100%, 30 seconds). Optimization of the washing time was performed on mouse tissues by collecting 5 serial sections on an ITO glass slide and adjusting the washing time of each tissue in buffer to include 0 (control), 2, 5, 10, and 15 min timepoints. Replicate experiments (n=3) were performed using the same tissue block to determine the average and percent decrease in signal intensity among replicate slides. To analyze protein contents released into the lysis buffer during the washing step, the lysis buffer was collected from a mouse spleen tissue section after fixation in ethanol solutions (70, 90, 95%) and washing in buffer for 10 minutes. Desalting of the buffer and isolation of potential protein contents were performed using a ZipTip with a C18 resin, then eluted and collected for analysis. Briefly, the resin was prepared by rinsing with 3 10 μL aliquots of 50% ACN and three 10 μL aliquots of 0.1% TFA. The collected sample of buffer was acidified to contain 0.1% TFA. The protein contents in the buffer were bound and concentrated by performing 20 passes of buffer solution through the C18 resin. After protein isolation, the resin was rinsed 5 times in 0.1% TFA to ensure complete removal of salts prior to protein elution in 50% ACN. The isolated protein contents were co-crystalized with 2-hydroxy-5-methoxybenzoic acid/5-methoxysalicylic acid (super-DHB, 40 mg/mL, 90/10/0.1 acetonitrile/water/trifluoroacetic acid) and spotted on a MALDI target plate for analysis. For protein imaging experiments, super-DHB (40 mg/mL, 90/10/0.1 acetonitrile/water/trifluoroacetic acid) was applied using an HTX M5 Sprayer (HTX Technologies) over a series of 12 passes, at a flow rate of 0.1 mL/min, nozzle velocity of 1200 mm/min, a nozzle height of 40 mm, track spacing of 2 mm in a crisscross pattern, and nozzle temperature of 75°C. Protein extraction was enhanced through rehydration in a sealed chamber containing 50% acetic acid vapor for 5 minutes at 37°C. All MALDI imaging data were acquired on a Bruker RapifleX MALDI TOF/TOF mass spectrometer (Bruker Daltonics) operated in linear positive ion mode from m/z 2000–24000. MALDI images were collected at a spatial resolution of 100 μm using a random walk raster and 2000 laser shots per pixel. Mass resolution was optimized for m/z 12000. Mass spectral information recorded on FlexImaging 5.0 (Bruker Daltonics) was imported into SCiLS Lab MVS 2020b Pro software ( http://scils.de/ ; Bremen, Germany) after acquisition for preprocessing and visualization. A baseline subtraction (convolution algorithm), normalization (root mean squared algorithm), and weak spatial denoising were applied to all data acquired prior to generating average representative mass spectra over measurement regions. Putative protein identification was performed based on comparison of measured m/z to data from MSiMass List ( maldi-msi.org ). For statistical analysis comparing the means of two sample sets, a t-test was used with a significance threshold set at 0.05. The number of protein signals detected was determined by exporting the average molecular profiles obtained from SCiLS as .csv files and performing peak picking using the MALDIquant package in the R programming language (version 4.1.2) with a fixed intensity threshold for each tissue type.

Matrix-assisted laser desorption/ionization (MALDI) mass spectrometry (MS) imaging is a powerful tool to investigate the correlation between spatially resolved molecular information with histological evaluation in thin tissue sections [ 1 – 8 ]. MALDI MS imaging enables the direct analysis of a variety of endogenous molecular species such as small metabolites, lipids, peptides, and proteins and has been employed to identify molecular markers to distinguish between normal and cancerous tissues as well as investigating disease progression and prognosis [ 9 – 15 ]. Molecular profiles acquired by MALDI MS can provide powerful insights for clinical studies, yet, the ionization process in MALDI has been described as inherently susceptible to ion suppression effects [ 16 – 18 ]. The presence of highly abundant molecules or molecules with higher ionization efficiencies compared to others can yield ions that dominate molecular profiles, thereby hindering detection of suppressed species and the depth of molecular information obtained [ 19 ]. In complex samples such as tissue sections, the presence of endogenous lipids and biological salts have been observed to affect the efficiency of protein desorption and ionization [ 20 , 21 ]. As such, numerous washing procedures have been developed to remove lipids and salts directly from tissue sections and enable detection of proteins directly from biological tissue sections [ 22 , 23 ]. For example, in a study by Lemaire et al ., rinsing tissue sections in organic solvents such as chloroform and xylenes were reported to improve MALDI MS signal sensitivity and increase the number of detected peptides/proteins by 34% and 44%, respectively [ 20 ]. Similarly, in another study by Garza et al ., treatment of tissue sections in organic solvents were observed to effectively remove dominating triacylglycerol and phosphatidylcholine lipids and enable the detection of multiply charged protein ions [ 24 ]. However, even after lipid and salt removal, protein molecular profiles can be easily dominated by hemoglobin alpha (Hb α) and hemoglobin beta (Hb β) proteins due to the overwhelming abundance of blood present in some tissue types and high ionization efficiency observed for hemoglobin, thus limiting the detection of other endogenous proteins [ 25 – 29 ]. Therefore, procedures which enable the selective removal of red blood cells and hemoglobin proteins from tissues is an intriguing area of exploration for improvement in direct analysis mass spectrometry techniques. A commonly used approach for the removal of red blood cells in blood and tissue homogenate samples involves the use of hemolyzing agents, such as an ammonium chloride solution [ 30 ]. When suspended in an isotonic solution of NH 4 Cl, erythrocytes undergo rapid hemolysis as free NH 3 diffuses through the erythrocytes’ cellular membrane, followed by an increase in intracellular concentration of OH − . Intracellular CO 2 then reacts with the OH − , forming HCO 3 − as a product. The intracellular HCO 3 − is then exchanged with extracellular Cl − across the plasma membrane mediated through the Cl − /HCO 3 − band 3 anion transport protein. The net result is an influx of NH 4 Cl inside the red blood cells, causing cellular swelling and lysis of the cell membrane [ 31 ]. Typical use of ammonium chloride buffers involves hemolysis in samples such as whole blood and lymphoid tissue homogenates during sample preparation involving DNA/RNA isolation and flow cytometry [ 32 – 34 ]. This buffer has also been applied in tissue engineering, such as during the isolation of pure populations of stem cells [ 35 – 37 ]. Although hemolyzing agents have been widely applied to whole blood and tissue, their application for direct use on tissue sections have not yet been described. Here, we describe a protocol to selectively reduce hemoglobin proteins in blood-rich tissue sections using an ammonium-chloride hemolyzing agent for MALDI MS imaging. Efficacy of this method and the effect of washing time was investigated in mouse spleen and liver tissue sections. Further, application of this protocol was demonstrated for MALDI-MS imaging of proteins in blood-rich human endometrial tissues. Use of the lysis buffer revealed an effective decrease of hemoglobin signal detected in the mass spectra obtained directly from tissue sections and enhancement of other proteins such as histones H4 and H2A, and COX7C. We show that this washing procedure can be easily integrated into protein MALDI MS imaging workflows with limited additional time added to the sample preparation protocol and provides an effective approach for removal of ion suppressing hemoglobin proteins.