Spatial localization of β-unsaturated aldehyde markers in murine diabetic kidney tissue by mass spectrometry imaging

Diabetic nephropathy (DN) is the leading cause of end-stage renal disease. Limitations in current diagnosis and screening methods have sparked a search for more specific and conclusive biomarkers. Hyperglycemic conditions generate a plethora of harmful molecules in circulation and within tissues. Oxidative stress generates reactive α-dicarbonyls and β-unsaturated hydroxyhexenals, which react with proteins to form advanced glycation end products. Mass spectrometry imaging (MSI) enables the detection and spatial localization of molecules in biological tissue sections. Here, for the first time, the localization and semiquantitative analysis of “reactive aldehydes” (RAs) 4-hydroxyhexenal (4-HHE), 4-hydroxynonenal (4-HNE), and 4-oxo-2-nonenal (4-ONE) in the kidney tissues of a diabetic mouse model is presented. Ionization efficiency was enhanced through on-tissue chemical derivatization (OTCD) using Girard’s reagent T (GT), forming positively charged hydrazone derivatives. MSI analysis was performed using matrix-assisted laser desorption ionization (MALDI) coupled with Fourier-transform ion cyclotron resonance mass spectrometry (FT-ICR). RA levels were elevated in diabetic kidney tissues compared to lean controls and localized throughout the kidney sections at a spatial resolution of 100 µm. This was confirmed by liquid extraction surface analysis–MSI (LESA-MSI) and liquid chromatography–mass spectrometry (LC–MS). This method identified β-unsaturated aldehydes as “potential” biomarkers of DN and demonstrated the capability of OTCD-MSI for detection and localization of poorly ionizable molecules by adapting existing chemical derivatization methods. Untargeted exploratory distribution analysis of some precursor lipids was also assessed using MALDI-FT-ICR-MSI. Graphical abstract Supplementary Information The online version contains supplementary material available at 10.1007/s00216-022-04229-7.


LESA-MSI optimized conditions
LESA MSI was performed using a Triversa Nanomate (Advion Biosystems, Inc, Ithaca, NY, USA) coupled to a Thermo Scientific™ TSQ Vantage Triple Stage Quadrupole Mass Spectrometer (Thermo Fisher Scientific, Bremen, GmbH & CO. KG). Spectra were acquired using selected reaction monitoring (SRM) in positive ion mode.
Aspiration was set to 0.9 µl solvent (methanol: water: formic acid 70:30:0.1% v/v). A 0.7 µl volume was dispensed on to the sample surface at a height of 0.4 mm and delayed for 5 s maintaining the liquid micro junction before 1.1 µl was re-aspirated at a height of 0.2 mm and delivered to the nanochip (5 µm internal diameter). Sample delivery time was set to 1 min; nitrogen gas pressure was 0.5 psi, 1.5 kV applied in positive ion mode. The system was cooled to 10°C. Images of kidney sections mounted on slides (600 dpi) were obtained using a scanner (Epson Perfection V330, software version 3.9.2.5 EN, Seiko Epson Corporation, Nagano, Japan) and viewed in LESA Points (Version 1.

Preliminary (proof-of-concept) RA Experiment
An initial small-scale experiment was performed for screening of potential OTCD reagents, Girard's reagent T (GT) and Dansyl hydrazine (DH), for detection of reactive aldehydes and to investigate signal intensity in diabetic kidney tissues.
Four slides were prepared in the configuration laid out in Fig. S2(a), using one diabetic mouse sacrificed at the third time point (16 weeks) and one lean control.
No further preparation was required for LESA-MSI. However, CHCA matrix was applied for MALDI-FT-ICR-MSI.

Untargeted Lipid Analysis
Three slides were prepared in the configuration in Fig. S2

OTCD-LESA-MSI Analysis
A second, ambient ionization method, LESA-MSI coupled to a triple quadrupole analyzer was used in this study to validate the findings of the MALDI-FT-ICR-MSI analysis. Though the spatial resolution in this case was 1 mm, some localization information was obtained and the trend in signal intensities between groups was matched to MALDI-FT-ICR-MSI results.
During LESA-MSI method development, 70:30% v/v methanol: water was selected as the most effective solvent (Fig. S5). As such, this was used in all further LESA-MSI analysis.
In the preliminary study, following OTCD with GT, LESA-MSI distribution patterns displayed an increase of signal intensity of RAs in diabetic kidney sections compared to a control (non-diabetic/lean) section (Fig. S6). Three replicate slides were analyzed to support the larger MALDI-FT-ICR-MSI experiment (Fig. S7a). For all RAs tested, the signal intensity in diabetic tissues was greater than control sections. TP1 sections displayed the highest intensity for RAs on all slides, except for 4-ONE on slide L. Mean signal intensities further demonstrate RA detection at higher concentrations at TP1 (Fig. S7b). 4-HHE was mainly scattered throughout the section, while 4-ONE and 4-HNE were mostly localized in the cortex at TP1 and TP2 but shifted to medulla at TP3.

Untargeted Lipid Analysis
Untargeted lipid analysis was conducted on kidney sections to investigate the differences between lipid profiles of diabetic and control tissues. MSiReader (Open source Release 1.02) [40] identified discriminatory lipid peaks between control and diabetic tissues which were matched against a positive ion database with < 5ppm mass accuracy. Flex Imaging was then used to visualize the matched lipid m/z intervals. Positive ion-matched peaks are listed in Table S3.
Signal for all these lipids was increased in diabetic tissues as observed in the MS images, except at m/z 732.55378 which was higher in the control tissue. At TP1, ions at m/z 703.57485 and 732.55378 were spatially distributed at a higher intensity towards the cortex, while ions of m/z 723.49352 and 798.54081 were at a higher intensity towards the medulla. At TP2 and TP3, all ions intensities were spatially distributed at a higher intensity in the medulla, except for m/z 732.55378 which is higher in the cortex. Lipid species were further confirmed by tandem MS/MS analysis using Lipid Blast version Full-Release-3, https://doi.org/10.1038/nmeth.2551 (Fiehn Lab, CA, US). The discrimination between control and diabetic tissues demonstrates a change in lipid profile in DN and the possible relationship between lipid peroxidation and RA generation.

Selectivity
The selectivity was determined by co-elution assessment analysis of wash, standard 0 (matrix blanks, 0ng/ml) and (LLOQ 0.5 ng/ml and 1.0. ng/g). Neither blank nor solvent washes showed peaks corresponding to RAs.

Matrix effect.
RAs free serum and tissue was used to matrix match.

Accuracy and precision
The accuracy and precision were determined at 1.0 (LQC) and 25 (HQC) ng/ml for serum and 10 (LQC) and 50 ng/g (HQC) for tissue. Sample size was n=6 across two separate analytical runs. Intra-day accuracy ranged between 89.43-99.5% and precision did not exceed 3.2% for all RAs in both serum and tissue. Inter-day accuracy ranged between 93.1%-102.3% and precision did not exceed 6.71%.
Intra analyst precision passed at all 3 analysts with lowest 88.5% accuracy and CV% with the highest result of 3.25%. Inter-analyst accuracy passed at 89.7% and CV% passed at 4.8%.

Carry over
No signals were observed for RAs in a solvent injection followed by HQCs injection showing no significant carry over.

Stability
Autosampler stability was determined for all RAs after sample preparation. Samples were stored at 4˚C for 48 hours after analysis. All results met acceptance criteria after 48 hours for both all RAs with levels against T=0 for all RAs above 85%.