The introduction of new mass-spectrometry-based techniques to neuroscience has created a wide array of new opportunities within this discipline 20,21. Nonetheless, many challenges related to areas such as safety, biocompatibility, and practicability must still be overcome. The present article explores chemical biopsy as a new tool that can be used in neurochemistry applications. Previous ex vivo and in vivo investigations of animal brains have demonstrated that some of this technology’s unique features may be able to fill the gaps present in existing methods 10,11. In addition, prior findings have shown that chemical biopsy is safe for use in laboratory animal studies, and enables the non-depletive analysis of labile endogenous substances, as well as metabolism quenching due to its ability to restrict the access of enzymes to absorbed substances 10,13,15,22.
This paper documents the first application of chemical biopsy for the analysis of human brain tissue in vivo, which was achieved by introducing a few adjustments to previously reported protocols. Firstly, it was necessary to sterilize the SPME fibers before inserting them into the human participants’ brains. To this end, ethylene oxide was used in this study, but other research has shown that steam sterilization can be used as well 23.
Another modification to previous SPME protocols made in this research was the compromise that was struck between analytical needs and practicality in a clinical setting. Specifically, the selected sampling time must be long enough to ensure adequate sensitivity and data quality, but short enough to be acceptable for the neurosurgeon and the patient. The length of the extractive phase coating on the probe must also be sufficient to attain reasonable recovery of the metabolites, but not too long to fit within the selected location. A literature review revealed that a sampling time of 15–30 min is usually used in metabolomics studies 24. However, such an extraction time would be impractical for the current study; since the extractions were performed on patients prior to the actual biopsy procedure while they were still conscious, the extraction time used in this research could be longer than a few minutes in duration. Therefore, 4 min was chosen as an optimum extraction time. Huq et al.’s foundational work on measuring free drug concentrations in solid tissue demonstrated that equilibrium time is established more quickly in complex matrices than in simple solutions like phosphate buffered saline (PBS) or agarose gel. This phenomenon is due to the fact that analytes are bound by tissue components (e.g., proteins), which in turn act as a local reservoir for the molecules to be extracted 25. As a result, the substances distributed in the cellular and intercellular spaces are able to be rapidly transported to the sorbent. Based on this finding, it is easy to conclude that even a short extraction time (e.g., 5 min) might be sufficient to achieve satisfactory recovery when extraction is performed from a complex matrix 25.
Unlike SPME, gold standard method in neuroscience – MD requires a relatively long sampling time, as it takes over 30 min to collect a sufficient amount of dialysate to enable the determination of target compounds. Moreover some experiments require even longer sampling times in order to obtain satisfactory recovery 26. Furthermore, a special cannula must be installed in the patient’s skull 3–4 days prior to the experiment in order to allow the MD probe to be introduced to the brain; thus, dates of MD experiments must be planned in advance. Attempts to modify this technology led to the development of push-pull probes (PPP), which enable faster analysis, but lower sensitivity and accuracy 26. Compared to MD and SPME, PPPs and electrochemical methods guarantee better temporal resolution, although push-pull probes are still prone to capillary clogging 26,27. Additionally, while electrochemical methods are able to return results in a few seconds, they only enable the analysis of selected substances, which may considerably reduce the scope of research 26,28.
As previously mentioned, extraction time and coating length must be customized in order to perform extractions from brain structures using SPME. The grey matter is a thin layer on the surface of the brain ranging between 1.0 and 4.5 mm in thickness (average thickness is approximately 2.5 mm) depending on the region 29. Despite the fact that detailed computed tomography (CT) and magnetic resonance tomography (MRI) fusion images are obtained prior to the biopsy, it is not possible to estimate the depth of the grey matter in the sampled area based on these images. In the present study, probes with a coating length of 3 mm were selected in order to enable the extraction of metabolites from grey matter with maximum precision and acceptable sensitivity. However, to further increase sensitivity without compromising extraction time, a modification was introduced to the design of the sampling device. Prior research has shown that it is possible to increase analyte extraction by increasing the surface area of the extraction phase (coating) 14, which is usually achieved by changing the device’s geometry to a thin-film format. The most commonly used thin-film variant in bioanalysis is the “blade” or Coated Blade Spray format, as they can be inserted directly into the tissue 22. Unfortunately, such devices are too invasive for use in brain studies. Therefore, an alternative strategy was employed in the present study, whereby recovery was enhanced by sampling each brain structure with two fibers simultaneously (i.e., the device is comprised of four fibers). After sampling, the two fibers used to sample each structure were desorbed together, thus increasing the amount of analytes extracted within the set 4 min (Figs. 1, 2). This modified approach enabled the detection of a wide range of analytes (Figs. SA1-SA4). Nonetheless, it is important to note that the range of extracted metabolites was still limited, and, consistent with the research conducted by Cudjoe et al. and Boyaci et al., for instance neurotransmitters were not observed in untargeted analyses 10,11. However, it is possible to extract neurotransmitters with SPME by optimizing the analytical protocol for that specific group of substances 10,15.
Zhang et al. 30 employed a similar spatial sampling approach to assess the concentration of pollutants in fish tissue samples, using segmented SPME fibers to simultaneously sample adipose and muscle tissues. After sampling, Zhang et al. cut off the coated parts of the fiber, and desorbed them in separate vials. Recently, another approach to spatial resolution analysis with SPME was presented by Lendor et al., who analyzed the fluoxetine profile in rat brains using SPME fibers for extraction, and desorption electrospray ionization (DESI) coupled to mass spectrometry for target analyte detection 31. Enhanced spatial resolution was achieved by moving the DESI probe along the fiber that had been inserted into the studied tissue.
The analytical methods most commonly used in in vivo neurochemical studies generally do not provide spatial resolution of brain structures that is as detailed as the method presented by Lendor et al. 31 With MD probes-even those with a diameter similar to SPME fibers (about 200µm)- it is impossible to obtain comparable results due to the method’s different governing principles. Thus, sampling must be carried out in a different way, which usually results in slower spatial resolution or time delay (i.e., sampling the grey and white matter sequentially rather than simultaneously) 26,32. Moreover, if sampling from two locations is planned, two cannulas outfitted with MD probes must be used, which will increase the invasiveness of the procedure and impact spatial resolution. Additionally, the MD probe needs to be connected to the device that pumps the perfusate during the sampling procedure. This means that a two single-channel-syringe pumps or one two-channel-syringe pump would be required in order to analyze two brain locations, which creates a need for additional space in the surgery room to accommodate the necessary extra equipment.
Distinct differences in the metabolomic profiles of the two studied brain matters were expected, but PCA did not show clear separation (Fig. SA5). Similarly, lipidomic analysis did not result in any differentiation between the analysed structures (Fig. 4). It was observed that the samples taken from patient B12 were located very close to each other (almost overlapping), which may lead to the assumption that, in this case, sampling was performed only from one type of matter (Fig. 4). As mentioned before, it can be quite difficult to locate the precise border between white and grey matters, as they vary in thickness across brain regions 29. Moreover, in in vivo human studies conducted in clinical environments, it is not possible to sample brain tissue purely for scientific purposes and to create a standardized study group. The patients from whom samples were collected in this study constituted a heterogeneous group in terms of age, sex, type, and location of brain lesion (Table SA1), which could affect the repeatability of the obtained results. Various factors, such as age, sex, medical conditions, sampling location in the cerebral cortex, and treatment applied before neurosurgical procedures, all influence the brains’ metabolome profile, and they all contributed to the relatively high variation observed in the average peak areas (Table SA2-SA5). For these reasons, biochemical in vivo brain studies are almost exclusively performed on laboratory animals, wherein entire neurosurgical procedures are designed for the research at hand. The current study was conducted on a small cohort, which makes it difficult to draw solid conclusions. Indeed, our results only verify the concept and identify drawbacks that should be addressed in future studies seeking to analyze a larger number of patients.
Given the small cohort and great variability of sampled areas and patients in the current study, the chemometric and statistical analysis of the identified metabolites predictably did not reveal significant differences in the metabolome profiles of the white and grey matter (Fig. SA5, Table SA2., and Table SA3). However, the analysis did reveal that most of the pathways in which non-hydrophobic species were engaged were related to amino acid metabolism. On the other hand, the best matches for metabolic pathways were assigned to lipids (Fig. 3). This result is in line with previous reports identifying lipids species as the predominant component in brain tissue composition 6,15,29,33. With regards to cerebral lipids, it is worth mentioning that MRI, a clinically approved method of differentiation of brain structures and detecting lesions, is mainly based on differences in lipid content (cholesterol and phospholipids) 34. Prior findings have also shown that white matter contains about 60% more lipids and 10–15% less proteins than grey matter 34. Similar observations regarding the prominent role of lipids in brain cells have been made using other mass-spectrometry-based methods, including REIMS and DESI 20. Cudjoe et al., concluded that SPME could be a complementary technology to MD, as it covers a broader hydrophobic range of metabolites (e.g., lipids), while MD is suitable for highly polar species (e.g., amino acids) 10. Both methods were compared by the performance of untargeted metabolomic profiling as well as targeted analysis of neurotransmitters (dopamine and serotonin) obtained from brain of living rats. Recent work by Boyaci et al. 11 has also confirmed Cudjoe et al.’s 10 observations regarding the complementarity of SPME and MD. Solid-phase microextraction makes it possible to tailor the protocol to the metabolites of interest, as it allows researchers to employ more selective coating and desorption parameters, and it can be coupled with appropriately optimized analytical instrumentation. In the present study, the results of the metabolomic analysis indicated a need for more in-depth lipidomic analysis. In order to better assess the lipid composition of the examined tissues, two types of sorbent were tested. First, fibers coated with C18 were analyzed, as they have been used in lipidomic studies previously, and they have a high affinity towards hydrophobic species 12,35,36. The second type of fiber that was selected for analysis was HLB-sorbent-coated fibers, as previous studies have demonstrated their high level of performance 36,37. The results of our comparison showed that the peak areas of detected features were larger in the samples collected with the C18 fibers, which indicated that this sorbent offered slightly better recoveries. The areas of only a few compounds showed a statistical significant difference (about 11% in RPLC-HRMS and 4% in HILIC-HRMS analysis) (Table SB1 and Table SB2). Based on this finding, the lipidomic analysis was carried out using the C18 fiber. This is consistent with the work of Boyaci et al. 11, who used the C18 sorbent for their lipidomics analysis, and a mixed-mode extractive phase for their metabolomics analysis.
Despite the lack of statistically significant separation between the two types of brain matter determined by PCA (Fig. 4), the results nevertheless showed the characteristic patterns of lipids in those structures. For instance, the results showed that phospholipids were present in greater level in the white matter, which could be due to the higher number of glia cells in this structure (Table 1). It has been proven that astrocytes and oligodendrocytes function as a reservoir for energy substrates (e.g., lipids for neurons), which means that there is intensive crosstalk and metabolite transfer between glia and neurons 38,39. These metabolites are probably more readily available to an SPME probe. It was also observed that the average peak area for TGs was bigger in the grey matter compared to their peak areas in the white matter (Table 2, Table SA4). Since the grey matter is located on the brain surface, many nutritional substances in their free form are first transported there. The high CV value for TGs in the present study can be partially explained by variations in the individual patients’ intake of nutrients, but it may also be due to the contamination of the fiber coatings by cerebrospinal fluid or blood present on the brain surface during the procedure, as both contain a wide range of substances (e.g., TGs) 40,41. Moreover, biological processes related to underlying disorders (e.g., obesity) may also influence the amount of TGs detected in individual patients 40. The detailed lipidome analysis results revealed that mainly phospholipids were identified in brain tissue (Table S.A.4., Table S.A.5.). This finding was unsurprising, as phospholipids are the main components of membranes and are engaged in variety of metabolic processes 42. The level of PC(36:1) was higher in white matter than in grey matter (the white-to-grey matter ratio was 1.3 for H+ adduct and 1.6 for Na+ adduct) while the opposite was observed for PC(34:1) (i.e., the white-to-grey matter ratio was 0.8) (Table S.A.5). Martínez-Gardeazabal et al. 42 used matrix-assisted laser desorption/ionization imaging mass spectrometry (MALDI-IMS) to map the lipidome in rat brains, and found that some lipids are more abundant in white matter than in grey matter. The authors detected a wide range of lipids, and spatial distributions of PC(36:1) and PC(34:1) that corresponded to our results. In addition, Martínez-Gardeazabal et al. 42 observed a higher intensity of LPC(16:0) compared to LPC(18:0), which is also in agreement with our results. In contrast, Jarmush et al. 43 used DESI to conduct a comparative analysis of neoplastic tissue, white matter, and grey matter ex vivo. This approach allowed the visualization and spatial resolution of the phospholipids and sphingolipids in a studied tissue 43,44. Some of the lipid species detected in the samples analyzed by Jarmush et al. 43 (e.g., PC(32:1), PC(34:1), PC(36:2), PE(36:2p), PE(38:5p), PE(40:6p)), were also detected in the present study, however there were differences in the formed adducts. Jarmusch et al. 43 reported that the white and grey matter had different lipid profiles, but they did not discuss this difference from a biological perspective. Among the mentioned phospholipids, two were identified as discriminative for differentiating white and grey structures 43,44. Specifically, they found that the presence of PC(34:1) in grey matter was more intense than in white matter, while the opposite was true for PE(36:2p). These findings were confirmed by the results of the current study (Table S.A.4., S.A.5.). MALDI-IMS and DESI enable the mapping of lipid species, but they cannot be applied for in vivo studies in clinical environments because of the complex setup of the hardware. Moreover, they provide less accurate identification compared to LC-MS/MS. In contrast, REIMS enables the fast detection of lipids in vivo, but not the accurate identification of individual species 21,45. This diagnostic method is based on comparing the profile of detected signals against the existing database. In addition, it provides results in real time, which makes it a great tool for assessing neoplasma margins during surgical procedures. However, the above-mentioned methods are not suitable for assessing structures/lesions located under the surface of the studied tissue, unless the area is exposed (e.g., during the removal of the lesion). This indicates that SPME can complement the portfolio of the existing MS-based approaches in neurological research.
This work documents the first use of SPME for the study of human brain tissue in vivo. Our analysis revealed some separation between the brain structures, with larger amounts of lipids being detected in the white matter than in the grey matter. Regardless, more detailed studies using larger study groups are required in order to draw solid conclusions regarding the biochemical profile of human brain tissue.
The research described herein demonstrates that it is possible to achieve spatially resolved chemical characterization of a living human brain in a fast and low invasive manner, thus not disturbing other medical procedures (e.g., biopsy). The sampling device can be customized into a personalized diagnostic tool, and the proposed protocol can be applied in numerous clinical settings when supported by imaging data. In addition, this research proves that the proposed method complements those already used in neuroscience, for example, microdialysis, push-pull analysis, and electrochemical methods. Furthermore, the findings of this research also show that chemical biopsies performed with SPME probes enable more in-depth chemical analysis of brain tissue compared to other clinically available methods (e.g., MRI, biopsy). Not only is SPME capable of extracting lipids and more polar metabolites, but its lipid-characterization capabilities are also unique compared to other in vivo sampling devices. The results presented in this paper indicate that the proposed tool could be potentially used as a medical device, as it causes minimum damage and features operational parameters that are compatible with surgical procedures.
Finally, it should be emphasized that more portable reading devices-rather than mass spectrometers using optical spectroscopic techniques, such as fluorescence or Raman could be used in targeted determinations in order to create a technology that is more compatible with bedside or operational theatre applications 46.