Decreased DHA‐containing phospholipids in the neocortex of dementia with Lewy bodies are associated with soluble Aβ₄₂, phosphorylated α‐synuclein, and synaptopathology

Abstract

Docosahexaenoic acid (DHA) is an essential omega‐3 polyunsaturated fatty acid implicated in cognitive functions by promoting synaptic protein expression. While alterations of specific DHA‐containing phospholipids have been described in the neocortex of patients with Alzheimer's disease (AD), the status of these lipids in dementia with Lewy bodies (DLB), known to manifest aggregated α‐synuclein‐containing Lewy bodies together with variable amyloid pathology, is unclear. In this study, post‐mortem samples from the parietal cortex of 25 DLB patients and 17 age‐matched controls were processed for phospholipidomics analyses using a liquid chromatography–tandem mass spectrometry (LC–MS/MS) platform. After controlling for false discovery rate, six out of the 46 identified putative DHA‐phospholipid species were significantly decreased in DLB, with only one showing increase. Altered putative DHA‐phospholipid species were subsequently validated with further LC–MS/MS measurements. Of the DHA‐containing phospholipid (DCP) species showing decreases, five negatively correlated with soluble beta‐amyloid (Aβ42) levels, whilst three also correlated with phosphorylated α‐synuclein (all p < 0.05). Furthermore, five of these phospholipid species correlated with deficits of presynaptic Rab3A, postsynaptic neurogranin, or both (all p < 0.05). Finally, we found altered immunoreactivities of brain lysolipid DHA transporter, MFSD2A, and the fatty acid binding protein FABP5 in DLB parietal cortex. In summary, we report alterations of specific DCP species in DLB, as well as their associations with markers of neuropathological burden and synaptopathology. These results support the potential role of DHA perturbations in DLB as well as therapeutic targets.

Alterations of DHA‐containing phospholipid species in DLB neocortex. (A) Heat map of the 46 putative DHA‐containing phospholipid species detected. (B) Bar graphs of the seven DHA‐containing phospholipid species found to be significantly altered in DLB.

1. INTRODUCTION

Neurodegenerative dementias are characterized by abnormal aggregation of peptides in the cortex, progressive synaptopathology, and neuronal loss, culminating in brain atrophy and clinical symptoms in patients [1, 2, 3, 4]. Following Alzheimer's disease (AD), dementia with Lewy bodies (DLB) has been recognized to be the second most commonest neurodegenerative dementias, accounting for 10%–20% of all cases [5, 6]. AD is characterized neuropathologically by amyloid plaques and neurofibrillary tangles, which are composed of aggregated, fibrillar 42‐amino‐acid Aβ (Aβ42) and abnormally hyperphosphorylated tau protein respectively. Notably, accumulating studies suggested soluble Aβ to be more neurotoxic than amyloid plaques and induce synaptic dysfunction [7, 8]. DLB is characterized by cortical Lewy bodies (LBs) containing aggregated α‐synuclein [9]. α‐synuclein in LBs undergoes various modifications including ubiquitination and phosphorylation [10]. Of these modifications, phosphorylation at serine 129 of α‐synuclein is suggested to enhance the aggregation of α‐synuclein and promote formation of α‐synuclein filaments and oligomers, which also confers neurotoxicity [10, 11, 12]. The extent of α‐synuclein phosphorylation in LBs is still under examination [13]. Besides LBs, DLB is known to manifest varying degrees of AD pathology [14, 15, 16], consisting of intercellular neuritic plaques containing insoluble fibrillar 42‐amino‐acid Aβ (Aβ42) which are derived from stepwise aggregation of soluble Aβ42 monomers and oligomers. Notably, previous studies have found soluble Aβ to induce synaptic dysfunction as well as greater neurotoxicity than insoluble amyloid aggregates [7, 8]. Besides amyloid plaques, another pathological feature found in both AD and LBD is intracellular neurofibrillary tangles consisting of abnormally phosphorylated tau proteins.

Given the high lipid composition of the brain, investigations of lipid alterations have been informative in neurodegenerative diseases. One of the best‐studied lipid species is docosahexaenoic acid (DHA, 22:6 n‐3) [17, 18], the most common essential omega‐3 polyunsaturated fatty acid (PUFA) in the human brain. DHA is known to promote synaptic protein expression, thereby supporting its involvement in learning and memory [17, 18, 19, 20]. Indeed, rodent models of DHA deficiency have been shown to result in learning deficits [21, 22, 23]. Consequently, DHA deficits are postulated to underlie cognitive impairments in dementia patients, and increasing evidence also supports the involvement of DHA perturbations in AD pathogenesis [18, 24]. However, early postmortem studies comparing total DHA levels between AD and control cortex reported no significant alterations at the level of phospholipid subclasses [25, 26, 27, 28, 29], each of which contains several distinct molecular species depending on the composition of the two fatty acid chains. In recent years, the development of highly sensitive mass spectrometric platforms has facilitated the discovery of complex molecular changes in the brain lipidome in disease states. For instance, recent studies revealed that distinct lipid species, including DHA‐containing phospholipids (DCP), were selectively depleted in AD, vascular dementia (VaD), and mixed AD/VaD dementia brains, and were also associated with neuritic plaques burden and disease duration [30, 31]. These findings suggest that alterations in distinct DCP species may play a role in the AD process. By comparison, there is currently a dearth of studies on DCP in DLB brain. Furthermore, potential associations between DCP species and alpha‐synuclein pathology, synaptopathology as well as concomitant AD pathology also remain to be explored.

In this study, we used a tandem mass spectrometry‐based lipidomics platform to comprehensively measure DCP species in the parietal cortex (Brodmann area 40) of a neuropathologically well‐characterized cohort of DLB patients and aged controls, then correlated DCP alterations with markers of DLB and AD neuropathological burden together with immunoreactivities of pre‐ and post‐synaptic proteins. Lastly, we aimed to elucidate potential mechanisms underlying observed DCP changes by measuring the levels of Major Facilitator Superfamily Domain 2A (MFSD2A), the primary transporter for DHA uptake into the brain [32], and fatty acid binding protein 5 (FABP5), implicated in regulating brain lipid homeostasis [33].

2. MATERIALS AND METHODS

2.1. Participants, cognitive and neuropathological assessments, and brain tissues

Study participants were initially recruited into longitudinal dementia studies in the UK and Norway with detailed neuropathological follow‐up [16], and the selection of brain tissues for this study was based on tissue availability, together with clinicopathological consensus diagnoses of DLB, including the DLB Consortium's “one‐year rule” [34] and the Movement Disorders Society criteria [35] as previously described [36]. At death, informed consent was sought from next‐of‐kin before removal of brains, which were collected via University Hospital Stavanger, Norway, and the United Kingdom Brains for Dementia Research network (https://www.brainsfordementiaresearch.org.uk/) encompassing the Thomas Willis Oxford Brain Collection, the London Neurodegenerative Diseases Brain Bank, and Newcastle University. The collection and study of brain tissues have received Institutional Review Board approval in both the UK (08/H1010/4) and Singapore institutions (NUS 12‐062E). Brains of 25 DLB patients were divided into hemispheres, with one formalin‐fixed for neuropathological assessments, which included the National Institute on Aging‐Alzheimer's Association guidelines with phases of Aβ deposition, Braak stages, and CERAD scores [37, 38, 39], as well as the Newcastle/McKeith criteria for LB disease [34]. The other hemisphere was dissected to obtain 1 cm³ blocks from the parietal lobe (Brodmann Area, BA40) followed by fresh freezing and storage at −80°C. Brains from 17 age‐matched controls were also included in this study. Control subjects were neurologically and cognitively normal, had only age‐associated neuropathological changes, and no history of psychiatric diseases.

2.2. Brain tissue processing and lipid extraction

All chemicals and reagents used were from Sigma‐Aldrich (St Louis, MO, USA) and of reagent grade unless otherwise specified. Blocks of gray matter from BA40 were thawed on ice, then dissected free of meninges and blood vessels before homogenization in 50 mM Tris buffer (pH 7.4) with added Complete ULTRA™ protease inhibitor tablets and PhosSTOP™ phosphatase inhibitor cocktail (Roche life Science, Penzberg, Germany). Prior to lipid extraction, a pooled quality control (QC) sample was prepared by pooling of 4 μL of 66 brain samples into an Eppendorf tube. Samples were then randomized using Microsoft Excel. An aliquot of the QC sample was included for every 11 test samples, with the randomized sequence followed throughout the experiment. For lipid extraction, 180 μL of chilled chloroform/methanol (1:2 v/v) containing internal standards that include DMPA, DMPE, DMPS, DMPC, LPC 20:0, LPE 14:0, LPA 17:0, DMPG and PI25:0 (Avanti Polar Lipid, Alabaster, Alabama, USA) were added to each 10 μL of sample. Samples were then incubated with agitation on a thermomixer at 700 rpm, 4°C, for 1 h in the dark. At the end of the incubation, 60 μL of chilled chloroform and 50 μL of chilled Milli‐Q® water (Merck Millipore, Billerica, MA, USA) were added and vortexed. Samples were then centrifuged at 10,000 rpm for 7 min, and the lower organic phase was extracted. Re‐extraction of the remaining aqueous phase was carried out by adding 500 μL of chilled chloroform, and extracted organic fractions were pooled, then centrifuged in a SpeedVac™ vacuum concentrator (Thermo Fisher Scientific) until dry. Lipid extract was re‐suspended in 100 μL of chilled chloroform/methanol (1:2 v/v), and stored at −80°C.

2.3. Mass spectrometry analyses

Samples were thawed for measurements of phospholipids by a liquid chromatography–tandem mass spectrometry (LC–MS/MS) system (Agilent 6460 Triple Quadrupole MS equipped with an Agilent 1290 Infinity LC). The QC sample injections were used to monitor reproducibility of injections: Coefficient of variation (CoV) was calculated for each lipid measured. Lipids with CoV > 25% were excluded from further analysis. Similarly, lipids with potential carryover (signal in blank injections >10% of average signal in QC samples) were not retained for analysis. Finally, linearity of response was assessed by duplicate injections of a dilution series of the QC samples. The column used in these measurements was a Phenomenex Kinetex HILIC 100 Å; column length, 150 mm; column internal diameter, 2.1 mm; particle size, 1.8 μm. Gradient elution was performed with solvents A (95% acetonitrile/ 5% 25 mM ammonium formate, pH 4.6) and B (50% acetonitrile/ 50% 25 mM ammonium formate, pH 4.6), with a gradient range from 1% to 75% solvent B in 6 min, 75 to 90% solvent B in 1 min, and 90 to 1% solvent B in 0.1 min followed by 1% solvent B for 3 min (total run time of 10.1 min). Levels of individual phospholipid species, including those belonging to the phosphotidylcholine (PC), phosphotidylethanolamine (PE), phosphotidylserine (PS), phosphatidylinositol (PI), phosphatidylglycerol (PG), lysophosphotidylcholine (LPC), and lysophosphotidylethanolamine (LPE) subclasses, were measured using multiple reaction monitoring (MRM). Quantitative data were extracted using Agilent MassHunter Quantitative Analysis (QQQ) software, and data were manually curated to ensure correct peak integration. Area under the curve for the extracted ion chromatogram peaks for each MRM transition and lipid species were normalized to the internal standard of the relevant class, and subsequently normalized to the protein concentration to derive the unit of nmol/mg protein. To mitigate the potential confounding effect of interindividual variation in the total level of phospholipids, we expressed the level of each lipid species as mean ± s.e.m. of the percentage of total level of phospholipids measured [40]. The LC–MS/MS approach used here yields lipid species identification at the level of fatty acid sum composition, with putative species of equivalent sum compositions reported together, separating by “/” (e.g., PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7). Therefore, phospholipid species with a fatty acid sum composition containing at least 38 carbons and 6 double bonds were initially considered putatively DHA‐containing [32], with the exception of PE 36:1/PE O‐38:8/PE P‐38:7 and PI 40:8/PI 39:1/PI O‐40:1/PI P‐40:0. For validation, significantly altered species underwent further MS/MS analyses to confirm the presence of DHA (see below).

2.4. Confirming the presence of DHA in the significantly altered phospholipid species

As mentioned earlier, phospholipid species with a fatty acid sum composition containing at least 38 carbons and 6 double bonds were initially considered putatively DHA‐containing [32]. To confirm the presence of DHA in these putatively DHA‐containing lipid species, further LC–MS/MS analysis was conducted on species that were significantly altered in DLB. The pooled QC sample was analyzed by LC–MS/MS using both HILIC and reverse phase (RP) separation, and a Vanquish LC system coupled to a Q‐Exactive plus mass spectrometer (Thermo Scientific). For HILIC separation, the same column and gradient as described above were used. For RP separation, the column used was an Eclipse Plus C18, column length, 50 mm; column internal diameter, 2.1 mm; particle size, 1.8 μm. RP gradient elution was performed with solvents A (40% acetonitrile/ 60% water with 10 mM ammonium formate) and B (10% acetonitrile/ 90% isopropanol with 10 mM ammonium formate), with a gradient range from 15% to 50% solvent B in 2.5 min, 50%–57% solvent B in 0.1 min, 57%–70% in 6.4 min, 70%–93% in 0.1 min, 93%–96% in 1.9 min, 96%–100% in 0.1 min, kept at 100% for 2.7 min, 100%–15% in 0.1 min, kept at 15% for 1.1 min (total run time of 15 min). Mass spectrometric parameters were as follows: MS Resolution 70,000; MS/MS resolution, 15,000S; MS can range, from 250 to 1250 m/z; positive and negative electrospray ionization, MS/MS inclusion list of species of interest for targeted fragmentation, isolation window 1.0 m/z. Data were interpreted manually using Xcalibur and FreeStyle software (Thermo Scientific). DHA‐containing lipid species were confirmed by accurate mass (Δppm <5 ppm), retention time, and presence of relevant DHA product ion in negative ionization.

2.5. Brain soluble Aβ_42, phosphorylated tau, neurogranin and Rab3A measurements using ELISA

Measurements for soluble Aβ₄₂, phosphorylated tau, neurogranin (as a postsynaptic marker), and Rab3A (as a presynaptic marker) by enzyme‐linked immunoassays (ELISAs) were as previously reported [1, 36].

2.6. Processing of brain tissues for pSer129 α‐synuclein measurement

As pSer129 alpha‐syn is mainly localized in LB‐containing insoluble fractions, dissected brain pieces were dispersed in a glass tissue grinder at 50 mg tissue wet weight/mL in ice‐cold mild extraction buffer (250 mM sucrose, 20 mM HEPES pH 7.4, 10 mM KCl, 1.5 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, with Complete Mini™ protease inhibitor and PhosSTOP™ phosphatase inhibitor tablets from Roche Diagnostics, USA added at recommended dilutions). Dispersed lysates were passed through a 26G needle 10 times using 1 mL syringes. After centrifugation at 800 x g for 5 min at 4°C, resulting pellets were re‐suspended in 1% sarkosyl Tris‐buffer, and centrifuged at 160,000 x g for 50 min at 4°C. The resulting pellets were further suspended in 5M guanidine buffer, shaking for 3 h at 25°C, then further centrifuged at 100,000 x g for 30 min at 4°C. This final supernatant was designated as the insoluble fractions [41]. Protein concentrations were measured with Coomassie Plus™ (Bradford) Assay Kit (Thermo Scientific). Before immunoblotting, proteins in the insoluble fractions were precipitated using ethanol, dissolved in Laemmli Sample buffer (Bio‐Rad), and heated up to 95°C for 5 min. The insoluble fractions were used for the measurement of monomeric pSer129 α‐synuclein and total α‐synuclein.

2.7. Immunoblotting of brain pSer129 α‐synuclein, total α‐synuclein, MFSD2A and FABP5

Samples were loaded on a 12% SDS‐polyacrylamide gel and transferred to a nitrocellulose membrane using the iBlot 7 min Blotting Dry‐transfer system (Invitrogen). Membranes were blocked with 5% BSA (in 20 mM Tris‐buffered saline, pH 7.6 with 0.1%Tween 20 (TBS‐T)) for 1 h at room temperature and then incubated with the respective primary antibodies overnight at 4°C at specified concentrations (see Table S1: for detailed information on the antibodies used). After incubation with primary antibody, membranes were washed with PBS‐T for 3 x 10 min and incubated with appropriate horseradish peroxidase (HRP)‐conjugated secondary antibody (Jackson ImmunoResearch) for 1 h at room temperature. Immunoreactivities were visualized with Luminata Crescendo Western HRP Substrate (Merck Millipore, Darmstadt, Germany) and quantified with Alliance 4.7 imaging software (UVItec, Cambridge, UK). For total α‐synuclein, immunoblots for pSer129 α‐synuclein were stripped and re‐probed with anti‐α‐synuclein (total) (Santa Cruz, Dallas, Texas, USA). Equal loadings of protein were calculated by protein assay and validated by Ponceau S staining.

2.8. Statistical analyses

All analyses were performed using the SPSS software (version 25, IBM Inc.). Comparisons of demographic and disease variables between DLB and controls were compared by Mann–Whitney U tests, while comparisons of lipidomics data were subject to false‐discovery rate correction for multiple testing with a cut‐off of 0.1 [42]. Correlations amongst lipid levels and other variables were performed using Spearman's rank correlation. For all analyses, p < 0.05 was considered statistically significant.

3. RESULTS

3.1. Demographics of study participants

Controls (N = 17) were well matched with DLB patients (N = 25) in age at death, brain pH, sex distribution, and post‐mortem delay (Table 1). For Braak staging, controls expectedly showed low scores (all subjects with Braak 0‐II), while DLB showed a range of Braak stages from lowest (Braak 0‐II) to the highest (Braak V‐VI, see Table 1). This observation of variable Braak stages within DLB is in agreement with previous studies [43, 44].

TABLE 1.

Demographics and clinical data.

	Control	DLB
Maximum number of cases (N)	17	25
Age at death (y)	79.7 ± 2	81.5 ± 1
Gender (M/F)	10/7	14/11
PMI (h)	41.0 ± 5	41.0 ± 6
Brain pH	6.48 ± 0.07	6.29 ± 0.08
Braak staging (n)
0–II	17	6
III–IV	0	9
V–VI	0	10

Note: Data are expressed as mean ± s.e.m. DLB, dementia with Lewy bodies. None of the variables such as age, PMI, and brain pH were significantly different between the groups.

3.2. Alterations of specific DHA‐containing phospholipid species in DLB neocortex

Of the 214 phospholipid species from seven phospholipid subclasses (phosphatidylcholines, PC; phosphatidylserines, PS; phosphatidylinositols, PI; phosphatidylethanolamines, PE; phosphatidylglycerols, PG; lyso‐PC, LPC and lyso‐PE, LPE that were detectable [Table S2]), 46 species were considered putatively DHA‐containing (i.e., fatty acid sum composition containing at least 38 carbons and 6 double bonds; also called DCP, Figure 1A). Next, expressing the level of each phospholipid species relative to the total level of phospholipids measured (% of total phospholipids), we found that majority of the DHA‐containing species belonged to the PE and PS subclasses, making up approximately 6% and 10% of total phospholipids, respectively (Figure 1B). There was no significant change in DHA at the level of phospholipid subclass in DLB (Figure 1B), in agreement with previous studies on other neurodegenerative dementias like AD [25, 26, 27, 29].

FIGURE 1.

(A) Flowchart summary of lipidomics results. A total number of 214 phospholipid species were measured. Among these measured species, 46 were considered putatively DHA‐containing (fatty acid sum composition containing at least 38 carbons and 6 double bonds). Of these 46 species, only seven were significantly altered in DLB. The presence of DHA in these seven species was confirmed by a subsequent LC MS/MS run. (B) The DHA levels in each phospholipid subclass in control and DLB expressed as mean ± s.e.m. of the percentage of the total level of phospholipids.

We next examined if, instead of phospholipid subclass, there were specific DCP species that were altered in DLB. Figure 2A shows the heat map of the 46 putative DCP species detected, while Figure 2B shows the fold‐change values (DLB vs. controls) of the seven species found to be significantly altered in DLB, amongst which six (PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7, PC 40:7, PC 40:8, PC 40:9, PE 38:8/PE 37:1/PE O‐38:1/PE P‐38:0 and PS 42:6) were decreased, whilst one (PE 37:6/ PE O‐38:6/ PE P‐38:5) was increased (Figure 2B).

FIGURE 2.

Alterations of DHA‐containing phospholipid species in DLB neocortex. (A) Heat map of the 46 putative DHA‐containing phospholipid species detected. (B) Bar graphs of the seven DHA‐containing phospholipid species found to be significantly altered in DLB. Y‐axis denotes fold change of each phospholipid species in DLB compared with controls. A fold change value of <1 implies that the phospholipid species is decreased in DLB compared with controls.

Of note, DCP species of the PC subclass showed the broadest reductions in DLB, with four of the 10 detected species, namely PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7, PC 40:7, PC 40:8, and PC 40:9, showing a significant decrease (Figure 2A, B). In contrast, all but one of the PS species showed trends toward increase (Figure 2A). Only one lipid species in PS, namely PS 42:6, showed a significant decrease (Figure 2A, B). For PE, one species (PE 37:6/PE O‐38:6/PE P‐38:5) was increased while another (PE 38:8/PE 37:1/PE O‐38:1/PE P‐38:0) was decreased, with similar nonsignificant trends toward both increases and decreases in the other DCP in the same subclass (Figure 2A, B). No significant difference was found in the levels of the DCP species within PI, PG, LPC, and LPE subclasses (Figure 2A).

3.3. Validation of significantly altered DCP species

Using further tandem MS/MS analyses, we were able to confirm presence of DHA (22:6) in five of the seven significantly altered species. PC 40:7 (PC 18:1/22:6) had DHA as the predominant or sole phospholipid species present. PE 37:6/PE O‐38:6/PE P‐38:5, PS 42:6, PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7 and PC 40:8 had DHA present, although it was not the most abundant phospholipid present. We were unable to confirm the presence of DHA in PE 38:8/PE 37:1/PE O‐38:1/PE P‐38:0 and PC 40:9, due to low abundance which hindered MS/MS analyses.

3.4. Correlations between altered DHA‐containing phospholipid species and neuropathological parameters

We next measured potential correlations between the significantly altered DCP species and the neuropathological indices. Soluble Aβ42 and monomeric pSer129 α‐synuclein were significantly increased in the brains of DLB subjects compared with controls (Figure 3 and Figure S1, p < 0.001 for both). pSer396 tau also showed a nonsignificant trend towards higher immunoreactivities in DLB subjects (p = 0.09). There was no significant difference between DLB and controls in all other tested parameters such as soluble Aβ40, total α‐synuclein, and total tau (Figure 3, all p > 0.05). As such, subsequent analyses focused on correlations between the levels of each significantly altered DCP species with soluble Aβ₄₂ or pSer129 α‐synuclein (Table 2).

FIGURE 3.

Soluble Aβ, tau, and pSer129 α‐synuclein in DLB neocortex. Bar graphs are mean ± s.e.m. soluble Aβ₄₂, Aβ₄₀, pSer396 tau, and total tau concentrations in pg/mg brain protein, together with monomeric pSer129 α‐synuclein and monomeric total α‐synuclein immunoreactivities in arbitrary units, in 17 controls and 25 DLB. *Significantly different from Control (Mann–Whitney test, p < 0.001).

TABLE 2.

Correlations between significantly decreased DHA‐containing phospholipid species with soluble Aβ42 and pSer129 α‐synuclein monomers.

	Neuropathological burden
DHA‐containing phospholipid species that are significantly DECREASED in DLB	Soluble Aβ42	pSer129 α‐synuclein
PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7	−0.384 * (p = 0.012)	−0.384 * (p = 0.012)
PC 40:7	−0.524 * (p < 0.001)	−0.582 * (p < 0.001)
PC 40:8	−0.557 * (p < 0.001)	−0.475 * (p = 0.001)
PC 40:9	−0.304 * (p = 0.050)	−0.288 (p = 0.065)
PS 42:6	−0.263 (p = 0.092)	−0.190 (p = 0.229)
PE 38:8/PE 37:1/PE O‐38:1/PE P‐38:0	−0.343 * (p = 0.026)	−0.298 (p = 0.056)

	Neuropathological burden
DHA‐containing phospholipid species that are significantly INCREASED in DLB	Soluble Aβ42	pSer129 α‐synuclein
PE 37:6/PE O‐38:6/PE P‐38:5	0.298 (p = 0.056)	0.431 * (p = 0.004)

Note: Rho values of Spearman correlations between each DHA‐containing phospholipid species with soluble Aβ42 and pSer129 α‐synuclein monomers in the combined cohort of control and DLB subjects. Bold and asterisk indicate a significant Spearman correlation.

Interestingly, decreases of PC species negatively correlated with soluble Aβ₄₂ levels (rho values ranging from −0.304 to −0.557; all p ≤ 0.012 except PC40:9, with p = 0.050). Similar correlations were found between PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7, PC 40:7, PC 40:8, and pSer126 α‐synuclein (rho values ranging from −0.384 to −0.582; all p ≤ 0.012). Furthermore, PE 38:8/PE 37:1/PE O‐38:1/PE P‐38:0 also correlated negatively with soluble Aβ₄₂ levels (rho = −0.343; p = 0.026). Lastly, PE 37:6/PE O‐38:6/PE P‐38:5, which was significantly increased in DLB, correlated positively with pSer126 α‐synuclein (rho = 0.431; p = 0.004).

3.5. Correlations between altered DHA‐containing phospholipid species and synaptic proteins

Given that DHA promotes synaptic protein expression, we postulate that reduced DCP species might be associated with synaptic protein deficits. Thus, we examined correlations between each significantly decreased lipid species with selected pre‐ (Rab3A) and postsynaptic (neurogranin) proteins, both of which have previously been shown to be reduced in DLB [1] (also see Figure 4). In this study, five out of six reduced DCP species correlated with Rab3A and/or neurogranin reduction (Rab3A: rho values ranging from 0.367 to 0.561; all p ≤ 0.025; Neurogranin: rho values ranging from 0.415 to 0.576; all p ≤ 0.011; Table 3). By contrast, PE 37:6/PE O‐38:6/PE P‐38:5 correlated negatively with Rab3A and neurogranin (rho = −0.390; p = 0.017 and rho = −0.464; p = 0.004, respectively).

FIGURE 4.

Synaptic proteins Rab3A and neurogranin in DLB neocortex. Bar graphs are mean ± s.e.m. Rab3A and neurogranin concentrations in pg/mL, in 15 controls and 22 DLB. Measurements were not performed for two controls and three DLB cases due to limited tissue availability. *Significantly different from control (Mann–Whitney test, p < 0.05).

TABLE 3.

Correlations between significantly decreased DHA‐containing phospholipid species with synaptic proteins.

	Synaptic markers
DHA‐containing phospholipid species that are significantly DECREASED in DLB	Rab3A	Neurogranin
PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7	0.367 * (p = 0.025)	0.576 * (p < 0.001)
PC 40:7	0.416* (p = 0.010)	0.440 * (p = 0.006)
PC 40:8	0.561 * (p < 0.001)	0.415 * (p = 0.011)
PC 40:9	0.297 (p = 0.074)	0.463 * (p = 0.004)
PS 42:6	0.417 * (p = 0.010)	0.454 * (p = 0.005)
PE 38:8/PE 37:1/PE O‐38:1/PE P‐38:0	0.302 (p = 0.070)	0.295 (p = 0.076)

	Synaptic markers
DHA‐containing phospholipid species that are significantly INCREASED in DLB	Rab3A	Neurogranin
PE 37:6/PE O‐38:6/PE P‐38:5	−0.390 * (p = 0.017)	−0.464 * (p = 0.004)

Note: Rho values of Spearman correlations between each DHA‐containing phospholipid species with Rab3A and neurogranin in the combined cohort of control and DLB subjects. Bold and asterisk indicate a significant Spearman correlation.

3.6. Alterations in MFSD2A and FABP5 levels in DLB neocortex

In order to investigate potential mechanisms underlying the observed DCP changes, we measured the lysolipid transporter MFSD2A responsible for DHA uptake into the brain [32, 33], and fatty acid‐binding protein FABP5 putatively implicated in the regulation of lipid transport and homeostasis [32, 33]. We found that MFSD2A and FABP5 were differentially altered in DLB cortex, with decreased immunoreactivities for MFSD2A contrasting with increased FABP5 (Figure 5). Whilst none of the lipid species correlated significantly with MFSD2A, PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7, PC 40:7, PC 40:9, and PE 38:8/PE 37:1/PE O‐38:1/PE P‐38:0 correlated negatively with FABP5 (rho values ranging from −0.308 to −0.439; all p < 0.05, see Table 4), whilst PE 37:6/ PE O‐38:6/ PE P‐38:5 showed a positive correlation (rho = 0.507; p = 0.001).

FIGURE 5.

DHA transporter MFSD2A and lipid‐binding protein FABP5 in DLB neocortex. Bar graphs of MFSD2A and FABP5 immunoreactivities in mean ± s.e.m. arbitrary units and normalized to GAPDH and β‐actin respectively, in 17 controls and 25 DLB. *Significantly different from controls (Mann–Whitney test, p > 0.05).

TABLE 4.

Correlations between significantly decreased DHA‐containing phospholipid species with DHA transporters.

	DHA transporters
DHA‐containing phospholipid species that are significantly DECREASED in DLB	FABP5	MFSD2A
PC 38:1/PC 39:8/PC O‐40:8/PC P‐40:7	−0.429 * (p = 0.005)	0.297 (p = 0.056)
PC 40:7	−0.439 * (p = 0.004)	0.228 (p = 0.147)
PC 40:8	−0.287 (p = 0.066)	0.174 (p = 0.270)
PC 40:9	−0.319 * (p = 0.039)	0.288 (p = 0.064)
PS 42:6	−0.240 (p = 0.125)	0.010 (p = 0.948)
PE 38:8/PE 37:1/PE O‐38:1/PE P‐38:0	−0.308 * (p = 0.047)	0.119 (p = 0.445)

	DHA transporters
DHA‐containing phospholipid species that are significantly INCREASED in DLB	FABP5	MFSD2A
PE 37:6/PE O‐38:6/PE P‐38:5	0.507 * (p = 0.001)	−0.148 (p = 0.351)

Note: Rho values of Spearman correlations between each DHA‐containing phospholipid species with Rab3A and neurogranin in the combined cohort of control and DLB subjects. Bold and asterisk indicate a significant Spearman correlation.

4. DISCUSSION

4.1. Summary of findings.

The present study is the first to assess changes in DCP species (predominantly found in 38:6 and 40:6 species in the individual phospholipid subclasses [32]) in DLB and to associate these changes with neuropathological parameters and synaptic markers. In our study, the DLB brains displayed concomitant, variable AD neuropathological burden, consistent with previous studies [14, 15, 16]. Our results revealed that seven DCP species were significantly altered in DLB compared with cognitively normal controls, amongst which six species were decreased in DLB. Of these six species, five of them correlated negatively with soluble Aβ₄₂, while three species correlated negatively with pSer129 α‐synuclein. That the majority of the DCP species also correlated positively with the synaptic proteins Rab3A and neurogranin implies a potential role of these species in promoting synaptic protein levels. Overall, our results warranted further examination of the roles that deficits of DCP species play in DLB pathogenesis, and as potential therapeutic targets or biomarkers of DLB.

Early postmortem studies comparing total DHA levels within individual phospholipid subclasses between AD and controls did not show significant changes [25, 26, 27, 29], in agreement with present findings (see Figure 1B). However, more recent studies measuring lipids at the molecular level revealed that distinct lipid species, including those which are known to be DHA‐containing, were altered in dementia [30, 31]. For instance, Yuki et al. demonstrated that PC molecular species containing stearate and DHA, namely PC (18:0/22:6), or PC 40:6, were selectively depleted in the temporal gray matter of AD brains, and correlated negatively with disease duration but not with amyloid deposition [31]. By contrast, another major DHA‐containing species, PC (16:0/22:6), or PC 38:6, was not significantly different between AD and controls [31]. Another study by Lam et al. using postmortem brain tissues from subcortical ischemic vascular dementia, mixed dementia (AD and vascular/cerebral infarcts) patients, and controls, reported alterations in specific lipid molecular species such as PC 40:6, PE 40:6, and PS 40:6 in the respective dementia groups, particularly the mixed dementia group [30]. However, there is to date a dearth of research in DLB, widely considered the second commonest cause of neurodegenerative dementias. Therefore, the current study addresses an important knowledge gap using an LC–MS/MS lipidomics approach for comprehensive profiling of the phospholipid species in well‐characterized cohorts of neuropathologically confirmed DLB and controls.

Given the frequent but variable burden of concomitant amyloid pathology in DLB, as well as previous preclinical work suggesting an effect on lowering soluble Aβ with dietary DHA supplementation [45], we performed measurements of soluble Aβ₄₂, widely considered to be more neurotoxic than amyloid plaques [7] in human cortex, and demonstrated a negative correlation with all DHA‐containing PC species found to be reduced in DLB. In addition, we have shown for the first time a negative correlation between pSer129 α‐synuclein with three of the DCP species, suggesting associations between DHA and markers of LB burden. Our data, therefore, give impetus to future investigations into the mechanistic links amongst DHA signaling, α‐synuclein phosphorylation, and DLB‐specific pathology.

In this study, we also explored correlations between significantly altered DCP species and immunoreactivities of the synaptic proteins, Rab3A, and neurogranin. We showed that majority of the DCP species correlated positively with Rab3A and neurogranin, supporting previous findings that DHA plays critical roles in promoting synaptic protein levels [46, 47]. Rab3A is a presynaptic vesicle protein that reflects the recycling pool of synaptic vesicles. In our previous study, Rab3A levels in the parietal cortex of DLB patients were found to be negatively correlated with LB burden and cognitive impairment, highlighting its involvement in DLB synaptopathology [1]. Similarly, animal studies [48] have suggested associations between cognition, synaptic plasticity, and neurogranin, a postsynaptic protein involved in the regulation of synaptic transmission through its interaction with the calcium‐binding protein calmodulin [48]. Neurogranin is amongst the best‐established CSF biomarkers for synapse loss or dysfunction, with studies showing that high baseline CSF neurogranin levels predicted future cognitive decline and correlated with longitudinal reductions in hippocampal volume [49, 50, 51].

4.2. Clinical and research implications

The brain has a limited ability to synthesize DHA from precursors, and although passive diffusion of DHA has been shown [52], the majority of brain DHA is transported from the peripheral circulation via the sodium‐dependent lysophosphatidylcholine (LPC) symporter, Major Facilitator Superfamily Domain containing 2A (MFSD2A) [32]. Given the observed alterations of specific DCP species in DLB brains, we examined disrupted DHA transport as a possible mechanism underlying the observed alterations by measuring MFSD2A immunoreactivities in DLB brain and additionally examined the levels of FABP5, previously linked to brain lipid homeostasis, and found significant decreases of MFSD2A contrasting with increased FABP5. Interestingly, it was previously reported that the deletion of MFSD2A resulted in a 90% reduction in the brain uptake of ¹⁴C‐LPC‐DHA, which was significantly stronger than effects arising from FABP5 deletion (~36% reduction) [32, 33]. Furthermore, inactivating mutations in MSFD2A are known to lead to microcephaly [53], underscoring the critical roles that MFSD2A plays in brain lipogenesis [54]. We, therefore, speculate that whilst increases in FABP5 might represent a compensatory mechanism toward DHA reductions, it is the deficits in MFSD2A‐associated DHA transport which is the predominant factor underlying the observed reductions in brain DCP species in DLB patients. Future in vivo studies are needed to assess the impacts of MFSD2A dysregulation on brain DCP species. Additionally, LPC‐DHA and PC‐DHA targeted therapeutic strategies should be assessed for DLB patients.

4.3. Limitations of study

There are a few limitations to our study. First, as with most lipidomics‐based approaches, one limitation is the lack of available reference standards for each of the phospholipid species being examined. Without such reference standards, it is necessary to presume that detection efficiency was constant within each headgroup class and that the ion signals recorded for each transition are proportional to the concentrations of the species present. Nonetheless, given that the aim of our study is to compare the alteration in DHA‐containing species between DLB and controls, relative changes in quantities are of value irrespective of uncertainty in the absolute quantities. Next, we reported lipid species at the level of sum composition (e.g., PC 38:6), rather than at the level of fatty acyl composition (e.g., PC 16:0/22:6). Therefore, resolution of lipid isobars/isomers is limited, and follow‐up with more advanced platforms with individual species resolution is warranted. In addition, we were unable to confirm the presence of DHA in two of the significantly altered, putative DHA‐containing groups, due to their low abundance (see “Results” Section 3.3). However, based on the nomenclature of DHA and other fatty acid chains in the brain, it may be reasonable to postulate presence of DHA in these species. Finally, whilst our samples were derived from neuropathologically well‐characterized DLB subjects as previously described [16], it is worth noting that a relatively large proportion of them had significant Alzheimer pathology (as indicated by Braak staging of V–VI, see Table 1), a finding which has also been reported by others [14, 15, 16]. This suggests that AD processes may drive some of the observed lipid changes, and further work beyond the correlations between soluble Aβ42, pSer129 α‐synuclein with their respective DHA changes (Table 2) is needed to better delineate potential AD‐ versus DLB‐specific mechanisms underlying lipid dysregulation in neurodegenerative dementias.

5. CONCLUSION

In conclusion, we found that six DCP species were specifically decreased in the parietal cortex of DLB brains. Several of these DHA species correlated with biochemical measures of soluble Aβ₄₂ and pSer129 α‐synuclein, which may suggest their involvement in DLB neuropathology. We have also reported the positive associations between these lipid species and the synaptic proteins, elucidating the potential role of these lipid species in synaptic function. Taken together with previous studies of other dementias, we postulate that phospholipid species may be differentially altered in AD, VaD, and DLB, suggesting their potential utility as biofluid‐derived lipid panel‐based markers, especially when combined with other classes of signaling lipids such as the spshingosine‐1‐phosphate [55] and ceramide [56, 57]. Moreover, our findings help advance the addition of DLB as a potential target of DHA supplementation‐based therapeutic strategies, similar to those evaluated for AD [58]. Further examination of the enrichment of specific DCP subclass and species through these DHA supplementation is warranted. Finally, investigations on the pathophysiology and mechanisms underlying DHA alterations, as well as validation of biomarker utility in independent cohorts are required.

AUTHOR CONTRIBUTIONS

Joyce R. Chong, Dag Aarsland, Mitchell K.P. Lai conceived the study and designed the project; Deron R. Herr, Markus R. Wenk, Dag Aarsland, David L. Silver, Christopher P. Chen, Amaury Cazenave‐Gassiot provided domain expertise. Joyce R. Chong, Yuek Ling Chai, Huayang Xing, Amaury Cazenave‐Gassiot performed the experiments. Paul T. Francis, Clive Ballard, and Dag Aarsland provided postmortem and clinical data; Joyce R. Chong, Yuek Ling Chai, Amaury Cazenave‐Gassiot, Mitchell K.P. Lai analyzed the data. Joyce R. Chong and Mitchell K.P. Lai drafted the manuscript. All authors critically revised the manuscript for important intellectual content. All authors have read and approved this manuscript for submission and publication.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to declare.

ETHICS STATEMENT

Ethics approval for the collection and study of brain tissues have received Institutional Review Board approval in both the UK (08/H1010/4) and Singapore institutions (NUS 12‐062E), and informed consent was obtained from participants' next‐of‐kin prior to removal of brain.

Supporting information

Supplementary Table S1: Antibodies used in immunoblotting.

Supplementary Table S2: Average amount of each phospholipid species in control and DLB expressed as the percentage of total amount of phospholipids.

Supplementary Figure S1: Representative immunoblot for monomeric pSer129 α‐synuclein and monomeric total α‐synuclein in the insoluble fraction.

Chong JR, Chai YL, Xing H, Herr DR, Wenk MR, Francis PT, et al. Decreased DHA‐containing phospholipids in the neocortex of dementia with Lewy bodies are associated with soluble Aβ₄₂, phosphorylated α‐synuclein, and synaptopathology. Brain Pathology. 2023;33(6):e13190. 10.1111/bpa.13190

DATA AVAILABILITY STATEMENT

Data reported in this study will be provided by the Corresponding Author on reasonable request.