In the last decade, MSI techniques have been extensively used to study the distribution of different elements, drug compounds and endogenous biomolecules such as lipids, peptides, and proteins in various tissues (Becker et al.
2010; Shariatgorji et al.
2014; Mathur et al.
2009). The aim of the current study was to test whether commercial MALDI-MSI and LA-ICP-MSI setups would allow the identification of cortical areas in human brain sections based on element and biomolecule distribution patterns, using V1 as a region that is particularly well characterized by microanatomical and functional means (Hubel and Wiesel
1977; Hinds et al.
2009; Zilles et al. 2009; Palomero-Gallagher and Zilles
2017).
Protein distribution patterns
In situ MS/MS analysis identified 71 peptides from 13 proteins. Many additional peptide signatures were observed, but not identified due to their low concentration, which precluded acquisition of high-quality MS/MS fragmentation spectra in situ. Although peptide identifications can be increased by parallel processing of adjacent tissue sections, followed by peptide extraction and analysis by nano-LC–MS/MS (Heijs et al.
2015), we did not follow this approach because the limited dataset already contained distinct protein/peptide patterns that were reproducibly observed in each biological and technical replicate. However, it should be noted that this does not hold true at the level of normalized intensity due to a variety of biological (differing age, gender and unknown case history; Table
1) and technical factors (different post-mortem delays, variability in tryptic digest efficiency).
Nevertheless, the peptide patterns showed a well-defined laminar resolution. Peptides derived from MBP were found in the WM and in layers IVb and I, as previously described (Horton and Hocking
1997). Layer IVb is an unequivocal characteristic of V1 that is absent in V2, MBP peptides therefore clearly demarcated V1. In contrast, neuromodulin (GAP43, Van Lookeren Campagne et al.
1989; Leu et al.
2010; Holahan
2017) and microtubule-associated protein tau demarcated layer IVb and, therefore, also V1 due to their lowest laminar concentration in this sublayer. Synapsin, a protein linked to synaptic transmission, was mainly enriched in supragranular layers, barely detected in layer IV, and then slightly increased again in the infragranular layers. This distribution was in accordance with previous studies reporting that thalamocortical terminals arriving in layer IV, which provide the driving input to V1 from the lateral geniculate nucleus, do not use synapsin (Owe et al.
2013). Furthermore, similar distribution patterns were observed for myristoylated alanine-rich C kinase substrate (MARCKS) and neuromodulin. These two presynaptic proteins, which are both involved in regulating the dynamics of the actin cytoskeleton at the synaptic membrane (Laux et al.
2000) and are considered of diagnostic value for neurodegenerative diseases (Remnestal et al.
2016), showed a distinct absence in layer IV and, therefore, differentiated V1 from V2. Both MARCKS mRNA and neuromodulin mRNA were co-expressed in monkey V1 (Higo et al.
2002,
2004). However, intense signals in layers IVb, V and VI were only observed for MARCKS (Higo et al.
2002). In contrast, we observed both proteins in layers V and VI (Supplementary Figs. S8, S11). Neurofilament light protein was predominantly observed in the GM, sparing supragranular layers, in agreement with a mesh-like distribution in layer IVa as described by immunohistochemistry (Preuss et al.
1999). The glial fibrillary acidic protein (GFAP), a protein mainly expressed in astrocytes, was mainly confined to layer I although immunohistochemical studies have shown that GFAP-positive astrocytes accumulate more in supragranular layers than in infragranular layers and WM, and that highest GFAP expression occurs in those cells forming the glia limitans (Eilam et al.
2016). It is possible that the concentration of GFAP in layers II/III is so much lower compared to layer I (or astrocytes forming the glia limitans) that it did not reach the detection limit of our method.
Overall, MALDI-MSI data showed good agreement with the distribution patterns of proteins previously studied in V1. Moreover, we identified peptide signals derived from brain acidic soluble protein (BASP-1), actin cytoplasmic protein and stathmin, all of which were previously not described in the primary visual cortex. BASP-1, a presynaptic protein involved in several cellular processes e.g. during brain development in rodents (Kropotova et al.
2013), was most abundant in supragranular layers including layer IV and is missing in layer IVb, which also made it possible to differentiate V1 from V2. Stathmin, a protein involved in regulating cytoskeletal dynamics and adult neurogenesis (Kedracka-Krok et al.
2016; Martel et al.
2016), was detected in supragranular layers and showed lower intensities in infragranular layers while sparing layer IVb completely.
Bauernfeind et al. (
2015) recently studied protein distribution in different cortical areas, including cingulate cortex, motor cortex or primary visual cortex, and observed similar patterns of protein expression among supra- and infragranular layers of neocortex that were consistent with the cytoarchitectonic features independent of the region. Remarkably, no distinct signature of V1 was reported, whereas we observed distinct differences between V1 and V2 due to differential protein accumulation particularly in layer IVb. This may be explained by the different sets of proteins detected by the different technical approaches employed: Bauernfeind et al. (
2015) analyzed distinct matrix spots with 200 µm diameter and detected intact proteins in a mass range between 2 and 40 kDa using an MALDI-TOF mass spectrometer, whereas we detected peptides after in situ protein digest and coating of the entire sections with a thin layer of matrix and thereby indirectly observed also proteins with a higher molecular weight, with little overlap between the two datasets. Finally, in comparison with immunohistochemical approaches where antibody cross-reactivity is hard to exclude, our in situ MS/MS analysis provided direct evidence for the sequence of the visualized peptides.
Metal ion homeostasis is severely affected in a variety of neurodegenerative diseases, including Alzheimer's disease, Parkinson's disease (Bourassa and Miller
2012) and traumatic brain injury (Portbury et al.
2016). Specifically affected cortical areas can be identified by LA-ICP-MSI, which measures concentration of metals and other elements in a spatially resolved manner (Sussulini et al.
2017). Such findings make it necessary to study the distribution of metals in a more systematic way in different regions of brains of healthy controls. Here we mapped the element distributions in area V1. Discrete patterns for selected elements were found. For example, copper was distributed across the cortical cross-section, and appeared enriched in layer IV (Fig.
3). This pattern was similar to that described for Cu in the human insular cortex (Dobrowolska et al.
2008) and in agreement with previous reports indicating that Cu was more abundant in GM than WM in nonhuman primate brain (Bonilla et al.
1984; Ramos et al.
2014; Knauer et al.
2017). Iron was observed in blood vessels, as expected due to its well-known association with hemoglobin, but also along layer IV and infragranular layers. Key proteins of iron homeostasis are also involved in Cu regulation (Mueller et al.
2009), which may explain the similar distributions of Fe and Cu in layer IV and infragranular layers.
Patterns of lipid distribution
Lipids represent the largest component of the brain (Li et al.
2017). Not only the amount, but also the number of cerebral lipid species increased along the phylogenetic line from mice over apes to man, as well as with age within a single species (Bozek et al.
2015; Li et al.
2017). This suggests that lipids might offer a key to improve our understanding of the connectome and higher cognitive functions that evolved in primates (Bozek et al.
2015; Li et al.
2017). For example, the composition of phospholipids (PLs) with varying acyl chain length and the number of unsaturated C=C double bonds define the characteristics and functional efficacy of neural membranes (Van Meer et al.
2008). Apart from their structural role, polyunsaturated lipids are further precursors for important second messengers such as arachidonic (C20:4), eicosapentaenoic (C20:5), docosapentaenoic (C22:5) and docosahexaenoic (C22:6) acids (Guichardant et al.
2011) which are partly known to be involved in neuronal signaling (Gantz and Bean
2017). This is in agreement with the recently proposed hypothesis of “small molecule co-transmission” (Nusbaum et al.
2017), which suggests that membrane compounds regulate neurotransmitter signaling independently and in conjunction transmitter receptors. Furthermore, lipid rafts have been identified as important mediators of mGlu1 receptor-mediated signaling (Roh et al.
2014). It is, therefore, of great interest to obtain detailed information on the distribution of lipids in the brain, and to correlate their occurrence with the role of these areas in certain functional networks. Mass spectrometry-based methods such as MALDI-MSI, DESI-MS, SIMS, nanoparticlelaser desorption ionization or 40 keV argon cluster SIMS (Skraskova et al.
2015; Mohammadi et al.
2016; Bodzon-Kulakowska et al.
2017) are uniquely suited to obtain such data as it is generally not possible to use techniques such as fluorescence tags and fluorescence microscopy to define the location of the relatively small and dynamic lipids.
In this study, we revealed unique regional distribution patterns of more than 120 lipid species based on high-accuracy mass measurements. Specific lipid species were found in either GM or WM, enriched in distinct cortical layers or sublayers. Among all observed lipids, we found 20 species with a relevant distribution demarcating the border between V1 and V2, mostly based on a distinct sublayer-specific localization. For instance, PC O-34:0 was highly enriched in subcortical WM, whereas PC-40:6 was neither observed in WM nor in supragranular GM. This may indicate specific function(s) confined to layer IVc, a cortical sublayer of V1 that also showed specific protein expression related to ocular dominance columns in primate experiments (Ataman et al.
2016).
PCs are the most abundant lipids in the occipital cortex, at a concentration of 19 µmol/g tissue (Abbott et al.
2013). We identified 23 different PC species, each with distinct distribution patterns. Several PCs were specifically observed in layer IVc, including PC_40:6, PC_40:7 and PC_38:6 (Figs.
5,
6; Table
2, Supplementary Fig. S15). These PCs contain the long-chain polyunsaturated docosahexaenoic acid (DHA) as FA, which is known to play an important role within the visual system, in neurotransmission at synapses and during brain development (Sugiura et al.
2009; Sugiura and Setou
2009). Furthermore, these lipids are critical for the maturation of visual functions (Uauy et al.
2001). Interestingly, fibers arriving with the Radiatio optica from the lateral geniculate nucleus via layer IVb terminate with synaptic contacts mainly at layer IVc (Casagrande and Xu
2004). Specific accumulation of polyunsaturated PCs in this region may indicate an important role within the cell membranes contributing to synaptic transmission. Consistent with that hypothesis, a loss of PC_40:6 and PC_40:7 has been observed in the parieto-occipital cortex of Parkinson patients suffering from GM atrophy and visual hallucinations (Cheng et al.
2011). Moreover, saturated PC’s such as PC_30:0, PC_32:0, PC_33:0 or PC_34:0 were found in the GM and specifically enriched in supragranular layers. The latter contains high numbers of neurons and dendrites and also higher amounts of palmitic acid (16:0) than WM (Skinner et al.
1993; Sugiura and Setou
2009; Veloso et al.
2011a,
b; Martinez-Gardeazabal et al.
2017). In contrast, PCs containing 18:0, 18:1 or 18:2 as FAs are located in WM, for example, PC_33:2, PC_36:2, PC_36:1 or PC_38:2. This is in agreement with reports that showed 18:1 FA accumulation in myelin sheets (Kishimoto et al.
1969; Veloso et al.
2011a,
b).
Similar to PCs, PLs containing palmitic acid, such as PA_32:1 or PA_34:1, were enriched in supragranular layers. Likewise, a PA with polyunsaturated FA, PA_40:7, accumulated specifically in layer IVc. PS_40:6, which is the most abundant PS in brain (Hicks et al.
2006), and PS_42:9 were observed along the GM, but not in layer IVc. The vast majority of the sphingolipids that we have detected are distributed along the WM and blood vessels exceptions are, for example, SM_d33:1 and SM_d38:1. The former is specifically accumulating in the supragranular layers, whereas the latter is distributed along the GM except in layer IVb. It has been reported that SM_d38:1 is decreased in Alzheimer disease in the prefrontal cortex (Chan et al.
2012). Glucosylceramide (GlcCer), which are present in higher concentration in adults compared to infant brain (Li et al.
2017), were mainly present in the WM. Also all other lipid classes, e.g. glycerolipids, showed specific distribution patterns with enhanced or reduced amounts confined to specific cortical layers. However, the function(s) of these lipids still need to be established.