In the search and characterization of disease biomarkers for use in liquid biopsy applications, proximal fluids like blood and urine are commonly used. Proximal fluids are found adjacent to a given tissue or organ and represent a repertoire of secreted proteins and shed cells reflective of the physiological state of that tissue. For prostate cancer and other genitourinary diseases, proximal fluids are represented by seminal plasma and expressed-prostatic secretion in urine (EPSu) (Drake and Kislinger, 2014; Nawaz et al., 2014). EPSu, also termed post-digital rectal exam (DRE) urine, represents the fluid being secreted by the prostate following a digital rectal prostate massage, which in turn can be collected in voided urine post-exam (Drake et al., 2009; Drake and Kislinger, 2014). The prostate gland secretes many proteins and other biomolecules in a prostatic fluid that combines with seminal fluid and sperm from the seminal vesicles during ejaculation. Many of these prostatic proteins are glycoproteins, like prostate specific antigen (PSA), that perform functions to activate sperm and suppress the vaginal immune micro-environment. Our group has previously characterized the proteomic composition of EPSu and prostatic secretions, identifying hundreds of different prostate-derived glycoproteins (Drake et al., 2009; Drake et al., 2010; Kim et al., 2012; Principe et al., 2012). Development and evaluation of extensive targeted proteomic assays to these proteins in EPSu are in progress for use in prostate cancer diagnosis (Kim et al., 2016; Otto et al., 2020). The prostatic fluids, EPSu and urine are also rich in extracellular vesicles (EV), which are a source for many ongoing non-coding RNA and related oligonucleotide-targeted diagnostic assays for prostate cancer and multiple diseases (Van Gils et al., 2007; Laxman et al., 2008; Linxweiler and Junker, 2020). EV obtained from urine and EPSu continues to be a highly active area for diagnostic assay development (Wang et al., 2020; Erdbrügger et al., 2021).

The majority of proteins in EPSu, urine, and associated EVs are glycosylated (Drake et al., 2009; Drake et al., 2010; Kim et al., 2012; Principe et al., 2012), either on asparagine residues, termed N-linked glycosylation, or on serine or threonine residues, termed O-glycosylation. Changes in glycosylation have been well documented in prostate cancer associated tissues, cells and biofluids (Drake et al., 2015; Scott and Munkley, 2019; Tkac et al., 2019). Extensive glycoproteomic approaches, i.e., characterization of the glycan structures at the peptide sites of modification, have been reported for urine and EPSu glycoprotein targets (Leymarie et al., 2013; Saraswat et al., 2015; Brown et al., 2020). One of the most highly characterized glycoproteins is PSA (White et al., 2009; Leymarie et al., 2013; Nyalwidhe et al., 2013; Jia et al., 2017; Kammeijer et al., 2018; Hatakeyama et al., 2021), due to its known role in prostate cancer diagnosis and relatively simple glycosylation pattern of having a single N-linked glycosylation site. Although changes in glycosylation of PSA and many other prostatic glycoproteins have diagnostic potential, assays to efficiently characterize N-glycans in urine can be lengthy and require multiple processing steps, precluding large scale clinical utility (Song et al., 2019; Hanzawa et al., 2021). Lectin arrays have been effectively used to profile glycan motifs in large cohorts of clinical urine samples (Kawakita et al., 2021; Mise et al., 2021), however this approach cannot determine full glycan compositions or distinguish N-linked or O-linked origins. Based on an adaptation of a recently published workflow for rapid characterization of serum and plasma N-glycans (Blaschke et al., 2020), we report herein a more efficient slide-based approach combined with MALDI imaging mass spectrometry (IMS) workflows to detect total N-glycan profiles of urine, EPSu and prostatic fluid samples. Comparative results with N-glycans detected and histologically mapped in prostate cancer tissues by MALDI IMS are also included.

To complement previous proteomic studies of proximal prostatic fluids obtained in the urology clinic as related to prostate cancers (Drake et al., 2009; Drake et al., 2010; Kim et al., 2012; Principe et al., 2012) a series of different EPSu, EPSd and EPSev samples were used to develop a MALDI-based N-glycan profiling method. The goal was to have a workflow that required minimal sample processing and could be completed in a 6–8 h timeline, in contrast to current glycomic analysis workflows for urine and prostatic fluids that require multiple processing, derivatization and purification steps prior to analysis. A previous slide-based approach used for serum and plasma N-glycan profiling (Blaschke et al., 2020) was the starting point, and a workflow summarized in Figure 1 was developed for EPSu. A key feature is use of an amine reactive slide chemistry that covalently binds target glycoproteins, facilitating washing steps to remove lipid and salts prior to spraying of a molecular coating of PNGase F PRIME to release N-glycans. An additional concentration and buffer exchange step was added for EPSu and EPSd, using a 10,000 MW cut-off spin cartridge to concentrate and allow buffer exchange of the sample prior to addition to the amine-reactive slide. This step results in a three to four fold increase in concentration of glycoproteins in the biofluid. An SDS-polyacrylamide gel image showing protein loading examples for the EPSu, EPSd and EPSev samples are provided in Supplementary Figure S1.

Initial experiments focused on optimizing detection of N-glycans in EPSu and EPSd sample pairs obtained from the same donors, as well as existing EPSev pool samples. Similar to what was previously determined for serum or plasma preparation on the amine-reactive hydrogel slides, the key for optimal N-glycan detection was inclusion of the Carnoy’s solution wash after spotting, which serves to remove lipids and salts, as well as denature the bound glycoproteins facilitating access for PNGase F. Thus far, a total of 35 EPSu, 10 EPSd and 8 EPSev samples have been analyzed with the workflow shown in Figure 1. Cumulatively, the resulting N-glycans detected in each sample type are summarized in Supplementary Table S1, and structural class groupings are shown in Figure 2. Broadly, the EPSu samples had the most N-glycan species detected (n = 182) versus EPSd (n = 135). These numbers include multiple versions of the same N-glycan compositions for sialylated and sulfates species, which can vary in mass due to varying numbers of sodium ions associating with the charged groups. These glycoforms were included in Supplementary Table S1, but were generally detected at lower intensity values.

There was a range of protein concentrations across the samples in each sample type examined. While this created differences in the total intensity of the N-glycan profile and number of N-glycans that could be detected, these differences were accounted for by only comparing relative intensities, i.e. an individual N-glycan’s intensity relative to the total intensity of the N-glycans in that sample that were also seen in all sample types being compared. Representative samples with the most N-glycans detected were selected and compared from each sample type to display the breadth the N-glycan profiles.

The intensity of the N-glycan classes varied across the prostatic fluid samples (Figure 2A). EPSd and EPSev N-glycan classes had similar intensities, except for the higher amount of high-mannose N-glycans in EPSev. For all samples, the majority of the N-glycans were biantennary and/or fucosylated. EPSu had approximately 15 and 20% less sialylation than the EPSd and EPSev, respectively, but had an increased amount of tetraantennary and sulfated N-glycans. Many of these findings were replicated when grouping the N-glycans detected in each sample based on composition (Figure 2B). About half of the N-glycans in EPSu were fucosylated with no sialic acid compared to approximately 30% in the EPSd and EPSev samples. The EPSd and EPSev samples had higher levels of N-glycans with sialic acids and no fucose. When examining the intensity of individual N-glycans in the prostatic fluid samples, the most abundant N-glycans were typically biantennary with two galactoses (Figure 2C). In concordance with the N-glycan class comparison, EPSev had more high abundance high-mannose N-glycans than the other samples, and m/z 1419.4755 (Hex6HexNAc2 + 1Na) was the second most abundant N-glycan. The sulfated N-glycan m/z 2056.6156 (Hex5HexNAc4NeuAc1 + 1SO4 + 2Na) had a relative intensity of 6.3% in the EPSu, compared to 1.2% in EPSd and less than 1% in EPSev.

In order to compare mass spectra and absolute intensity values of N-glycan peaks, an EPSu and EPSd sample with similar protein levels from the same cohort were prepped together and imaged in the same run (Figure 3). Substantial differences were seen in most of the high-intensity N-glycan peaks, except for m/z 1809.6393 (Hex5dHex1HexNAc4 + 1Na), m/z 2465.8669 (Hex6dHex1HexNAc5NeuAc1 + 1Na), and m/z 2800.9263 (Hex6dHex1HexNAc5NeuAc2 + 3Na). The increased levels of sialylated N-glycans in the EPSd sample and increased fucosylated species in the EPSu sample is also evident.

The advantage of developing the assay using EPSu samples are the inherently higher protein concentrations present from the prostatic fluid mixture in these samples relative to normal urine. Therefore, for comparison, control urine samples from pools of four healthy male and four healthy female donors were processed and analyzed using this workflow. This required 4 ml of starting fluid for concentration and desalting using larger filtration tubes, but was otherwise the same workflow as described for EPSu/d samples. Overall, N-glycan class intensities were similar for both samples (Figure 4A). There were high levels of biantennary, fucosylated, and sialylated N-glycans and low levels of hybrid N-glycans. The rest of the N-glycan classes had approximately 10% intensity. The biggest difference between the male and female samples was the slightly lower level of fucosylation in the female sample. The distribution of N-glycans based on composition displayed very similar profiles (Figure 4B). The biggest difference was a 3.2% increase in high-mannose N-glycans in the female sample. Accordingly, m/z 1419.4755 (Hex6HexNAc2 + 1Na) was the sixth most abundant N-glycan in males and the second most abundant in females (Figure 4C). Similar to the EPSu samples, the most abundant N-glycans are primarily biantennary with two galactoses.

As the glycoprotein constituents of the EPSu and EPSd samples are secreted by prostate glands, a comparative N-glycan comparison of prostate tissues was done using previously reported N-glycan MALDI IMS approaches (Drake et al., 2017, 2018b; McDowell et al., 2021). In the example shown in Figure 5A, a pair of non-tumor and tumor tissues from the same donor were evaluated. Use of the non-tumor tissue from the same donor is done to demonstrate the specificity of tumor specific N-glycans detected in the tumor containing tissue [a Gleason grade 8 (4 + 4), stage pT3b]. The tumor region is quite distinct, localized to the bottom left corner of the tissue. Additional H&E images of both tissues with increasing magnification are provided in Supplementary Figure S2. In Figures 5B,C are shown representative N-glycan classes and their tissue distributions, from a total of 73 N-glycans detected. These structures include tumor-associated paucimannose (Hex3HecNAc2), high mannose (Hex5-Hex9HexNAc2), and branched fucosylated species (Hex7HexNAc6Fuc2) (Figure 5C), consistent with previous reports (Drake et al., 2018b). Two of the most abundant sialylated biantennary N-glycans detected are stroma-associated, as shown in the overlay image with a tumor N-glycan (in red), Hex5HexNAc4NeuAc1 in blue and Hex5HexNAc4Fuc1NeuAc1 in green (Figure 5B, and individually in Figure 5C). A segmentation analysis of the 73 N-glycans is shown in Figure 5D, illustrating how different N-glycan classes are associated with different histopathology features. This representative tissue was selected for another feature, as it was noted that there was a distinct intra-lumen glandular N-glycan signature that could be detected. As shown in Figure 6A, the two sialylated biantennary N-glycans (Hex5HexNAc4NeuAc1 in blue and Hex5HexNAc4Fuc1NeuAc1 in green) provide a stromal scaffold image. This was used to detect which N-glycan species were present in the lumen regions, as illustrated in Figure 6B for Hex7HexNAc6 in red, and in Figure 6C for a highlighted gland region. Doing this, 38 N-glycans were detected in the lumen of glands. Many of these same N-glycans are also tumor associated, but the histopathology differences are significant between tumor and lumen.

Using the lists of 73 tissue N-glycans and the subset of 38 lumen N-glycans, these were compared to the N-glycans detected in EPSu and EPSd samples. As shown in the Venn diagram in Figure 7, 44 N-glycans were present in the tissues and both EPS samples (Figure 7A), and 18 N-glycans were shared in the lumen and EPS samples (Figure 7B). Structurally, these shared N-glycans were the high mannose and the most abundant branched fucosylated N-glycans.

The goal of our study was to develop and optimize a more efficient and rapid method to evaluate N-linked glycosylation by MALDI-MS in urine and EPSu samples. The workflow is similar to what was developed for serum and plasma (Blaschke et al., 2020), utilizing an amine reactive slide to capture glycoproteins on a solid surface. An additional 1 h step of sample processing by molecular weight filtration and centrifugation for urine was included in order to concentrate and solvent exchange. Spotting only 1 ul, the urine glycoprotein spots are sprayed with a molecular coating of PNGase F, followed by analysis on a MALDI-FTICR or MALDI-QTOF mass spectrometer. The method is applicable to urine, EPSu and EPSd samples. An EPSev prep was spotted directly onto the amine reactive slides without any pre-processing. Routinely, 100 or more N-glycans can be detected depending on the sample type.

The N-glycan profiling of the prostatic fluid samples displayed more high-mannose N-glycans in the EPSev compared to the EPSu and EPSd, reflected in the intensities of the N-glycan classes and the most abundant individual N-glycans (Figure 1). Relatively high intensities of high-mannose N-glycans have been noted in non-EPS urinary exosomes as well (Zou et al., 2017). The similarity of the EPSd and EPSev N-glycan profiles suggests that a significant portion of the EPS glycoproteins may be within exosomes. In conjunction with these findings, direct comparison of mass spectra from EPSu and EPSd with similar protein levels displayed distinct N-glycan profiles for EPSu and EPSd (Figure 2). While male and female urine N-glycan profiles and individual N-glycan abundances displayed few differences, there was a slightly higher amount of fucosylation in the male sample. This trend has previously been identified in male and female plasma samples (Dotz and Wuhrer, 2019). Similar levels of sulfation were detected in the urine controls and EPSu, which was higher than the EPSd and EPSev, indicating that the sulfated N-glycans in EPSu may be originating from urine glycoproteins.

A comprehensive MS analysis of the N-glycans present in adult and pediatric urine samples has been published (Li et al., 2020). In this study, PNGase F released N-glycans were isolated, labeled with aminobenzoic acid, and methylamidated prior to analysis by LC-MS/MS. The authors reported 116 N-glycan compositions could be detected, and a subset of 46 N-glycans that were reproducibly detected and further quantified (Li et al., 2020). This study provides an optimal benchmark for our current study to evaluate how many of the 46 N-glycans could be detected by our solid-phase and single day MALDI MS workflow. Comparing the MALDI data from the male and female urine standards, the majority of N-glycans are detected, especially the biantennary, high mannose and fucosylated species. The MALDI-based assay is not as effective for detection of larger tri- and tetra-sialylated N-glycans, and for N-glycans with molecular masses above 3,700 m/z.

One observation noted from our data was that the intensity levels of N-glycans detected were variable, even when protein concentrations were comparable. There was also not a large variation in the types of N-glycan structures detected in the samples, especially for the most abundant species. These variations could reflect significantly different protein levels of individual glycoprotein species, or variations in the amount of the most abundant protein in human urine, uromodulin. Uromodulin, also called Tamm−Horsfall protein, has eight N-glycosylation sites and has been comprehensively studied for N-glycosylation content and composition (Li et al., 2021). Interestingly in this study, N-glycan analysis comparisons of a uromodulin depleted urine versus non-depleted sample indicated little difference in the N-glycan compositions detected in both samples (Li et al., 2021). Our study only represents a few sample numbers, so no conclusions can be reached until statistically relevant sample numbers are analyzed. The rapidity and efficiency of the slide-based MALDI assay will facilitate evaluation of larger cohorts. Possibly pairing the N-glycan data with the quantitative MRM proteomic assay developed for EPSu proteins could address both the changes in N-glycan levels and protein concentrations in the same sample (Kim et al., 2016; Otto et al., 2020). It also does not preclude direct targeting of specific glycoproteins present for N-glycan content analysis, as is done so often for PSA. We have already reported a solid-phase antibody array approach to N-glycan profile individual serum glycoproteins captured by their specific antibodies (Black et al., 2019a; 2019b). A similar approach targeting the abundant urine and EV glycoproteins like PSA is in progress.

The N-glycans obtained from a representative FFPE prostate cancer tissue were also included to illustrate the N-glycans that can be detected in the lumen of the prostate glands, representing prostatic fluid still present during tissue fixation. Based on hundreds of prostate cancer tissues analyzed by N-glycan IMS, to be reported separately, detection of luminal N-glycans in these tissues is highly variable. Fluid remnants can be seen in the H&E stains, but not all gland lumens contain this, and can be completely absent in many tissues. Presumably this variation reflects how much fluid was present at the time of prostatectomy, but could also reflect differences in FFPE tissue preparation and processing. While this is a variable that may lack clinical diagnostic significance, the selected tissue in Figure 5 highlights the tumor-associated and secreted N-glycans (n = 38) in tissue that can potentially be detected in the EPSu, EPSd and EPSev samples. These are primarily paucimannose, high mannose and branched multi-fucosylated N-glycans.

In summary, we present an initial N-glycan profiling workflow applicable to urine and prostatic fluids. It also represents a scalable framework, depending on the application and intent of the assay. There is already demonstrable sensitivity for detecting N-glycans efficiently in low concentrations of starting material, hence analysis of larger clinical cohorts is feasible. If time of preparation is less of a concern, an increased amount of starting fluid can be concentrated to increase protein concentrations prior to spotting. Additionally, the use of the concentration and solvent/buffer exchange step in the workflow can be adapted to other biofluids with protein concentrations similar to urine, i.e., saliva, cerebrospinal fluid, bronchial lavage. The method can also be used with other enzymes besides PNGase F, like endoglycosidase F3 specific to core fucoses (West et al., 2020), or other glycosidases specific to other N-glycan structural classes.