Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jan 15.
Published in final edited form as: Biosens Bioelectron. 2022 Nov 3;220:114862. doi: 10.1016/j.bios.2022.114862

Multi-Omics Analysis of Magnetically Levitated Plasma Biomolecules

Ali Akbar Ashkarran 1,#, Hassan Gharibi 2,#, Dalia Abou Zeki 3, Irina Radu 3, Farnaz Khalighinejad 3, Kiandokht Keyhanian 4, Christoffer K Abrahamsson 5, Carolina Ionete 3,*, Amir Ata Saei 2,6,*, Morteza Mahmoudi 1,*
PMCID: PMC9750732  NIHMSID: NIHMS1851084  PMID: 36403493

Abstract

We recently discovered that superparamagnetic iron oxide nanoparticles (SPIONs) can levitate plasma biomolecules in the magnetic levitation (MagLev) system and cause formation of ellipsoidal biomolecular bands. To better understand the composition of the levitated biomolecules in various bands, we comprehensively characterized them by multi-omics analyses. To probe whether the biomolecular composition of the levitated ellipsoidal bands correlates with the health of plasma donors, we used plasma from individuals who had various types of multiple sclerosis (MS), as a model disease with significant clinical importance. Our findings reveal that, while the composition of proteins does not show much variability, there are significant differences in the lipidome and metabolome profiles of each magnetically levitated ellipsoidal band. By comparing the lipidome and metabolome compositions of various plasma samples, we found that the levitated biomolecular ellipsoidal bands do contain information on the health status of the plasma donors. More specifically, we demonstrate that there are particular lipids and metabolites in various layers of each specific plasma pattern that significantly contribute to the discrimination of different MS subtypes, i.e., relapsing-remitting MS (RRMS), secondary-progressive MS (SPMS), and primary-progressive MS (PPMS). These findings will pave the way for utilization of MagLev of biomolecules in biomarker discovery for identification of diseases and discrimination of their subtypes.

Keywords: Plasma biomolecules, magnetic levitation, multiple sclerosis, proteomics, lipidomics, metabolomics

1. Introduction

Magnetic levitation of diamagnetic materials/objects in paramagnetic solutions is a well-documented density-based analysis and separation technique that has been used for a wide range of biological applications from cell sorting and tissue engineering to bone regeneration and self-assembly of living materials (Abrahamsson et al. 2020; Baday et al. 2019; Deshmukh et al. 2021; Durmus et al. 2015; Gao et al. 2022; Ozefe and Arslan Yildiz 2020; Parfenov et al. 2020a; Parfenov et al. 2020b; Puluca et al. 2020; Sözmen and Arslan-Yıldız 2022; Tasoglu et al. 2013; Tasoglu et al. 2015; Tocchio et al. 2018; Yang et al. 2019). There was a growing interest in the past years to use MagLev in cell biology and other novel applications as a label-free technology with a disease detection capacity (Delikoyun et al. 2021; Sarigil et al. 2019; Yaman and Tekin 2020). MagLev also has been used for scaffold-free fabrication of three-dimensional (3D) cell cultures in a weightlessness environment which is suitable for mechanobiology, drug discovery and developmental biology (Anil-Inevi et al. 2021; Anil-Inevi et al. 2018; Sarigil et al. 2021). However, this technique cannot levitate nanoscale biomolecules (e.g., plasma proteins), at least in part, due to i) the Brownian motion effect, and ii) the instability of the biomolecules arising from their interactions with high concentrations of conventional paramagnetic salts (e.g., GdCl3 and MnCl2) (Ashkarran and Mahmoudi 2021a),(Ashkarran et al. 2020b).

To overcome these limitations, we recently introduced superparamagnetic iron oxide nanoparticles (SPIONs), constituting a novel paramagnetic liquid that is able to minimize Brownian motion and reliably/reproducibly levitate plasma biomolecules. This is due mainly to SPIONs’ higher magnetic susceptibility based on their superparamagnetic properties as compared to paramagnetic salts, and their interactions with biomolecules in terms of inducing instability to their structures. Therefore, the use of SPIONs allows safer and faster levitation of nanoscale plasma biomolecules, compared to conventional paramagnetic solutions (Ashkarran and Mahmoudi 2021a; Ashkarran et al. 2020b). One of our striking observations from using SPIONs was that upon injection of the plasma into the MagLev system, plasma biomolecules began forming several ellipsoidal layers (Ashkarran et al. 2020a). Another striking finding was that the evolving magnetically levitated plasma biomolecules provided useful information regarding the health of plasma donors (Ashkarran et al. 2020a). To uncover the mechanism behind the formation of these unique patterns, we conducted proteomics analysis on the ellipsoidal layers but found no remarkable difference in their proteome profiles (Ashkarran et al. 2020a).

To mechanistically understand the underlying phenomena behind the formation of ellipsoidal bands in the magnetically levitated plasma biomolecules, here we report multi-level omics analysis on the formed ellipsoidal bands using lipidomics and metabolomics analysis. To investigate whether such differences in the biomolecular composition of various levitated ellipsoidal bands actually reflect health information, we conduct a proof-of-concept study on plasma samples of individuals who suffer from various types of multiple sclerosis (MS); we used MS subtypes as model diseases because discriminating among them is a major clinical challenge in autoimmune medicine.

MS is a common, unpredictable, immune-mediated, complex demyelinating disorder that affects the central nervous system (CNS) (Reich et al. 2018; Vaughn et al. 2019; Wootla et al. 2012). There are four main types of MS: relapsing-remitting MS (RRMS), secondary-progressive MS (SPMS), primary-progressive MS (PPMS), and progressive-relapsing MS (PRMS) (Axtell et al. 2010; DeLuca et al. 2020). Currently, clinicians must rely on combining the results of various clinical examinations including magnetic resonance imaging (MRI) and cerebrospinal fluid (CSF) analysis to determine the disease phenotype (Absinta et al. 2016; Cheng et al. 2017; Vrenken et al. 2010). Therefore, there is serious need for development of new techniques for early identification of different clinical phenotypes and patients at risk for progressive disease in order to accelerate biomarker detection that can help us understand MS pathogenesis, monitor disease progression, prescribe appropriate therapy early on, and contribute to novel therapeutics (Knowlton et al. 2015; Knowlton et al. 2017).

2. Materials and methods

2.1. Materials

SPIONs (30 mg/ml, commercially known as ferumoxytol) was purchased from Feraheme (www.feraheme.com) and diluted with phosphate-buffered saline (PBS 1X, HyClone) solution to desired concentration for all experiments. Plasma proteins from patients with three different types of MS were provided from the MS center at the university of Massachusetts Medical Center, Worcester, MA. For calibration of the MagLev system, fluorescent polyethylene microparticles with known densities and standard-density solid-glass particles were obtained from Cospheric (www.cospheric.com) and American density materials (www.americandensitymaterials.com), respectively.

2.2. MagLev platform

The standard MagLev platform used in the present work is depicted in Supplementary Figure 1. The standard MagLev set-up is one of the simplest configurations: have two blocks of N42-grade neodymium (NdFeB) cubic magnets (25.4 mm length, 25.4 mm width, and 50.8 mm height), 2.5 cm distance of separation and (like other MagLev platforms) the N poles face each other. Disposable plastic cuvettes were cut to 25mm and used as levitation containers.

All common MagLev devices have two blocks of coaxial identical permanent magnets with face-to-face similar poles and a separation distance of “d”. The details of the relationship between the density of an object and its levitation height within the MagLev system are well understood and reported by our group and others (Alseed et al. 2021; Ge et al. 2020; Turker and Arslan-Yildiz 2018).

.Briefly, the “density-levitation height” equation can be derived simply by considering the anti-Helmholtz configuration of magnetic fields, which results in a B-field only in the z direction (i.e., B-fields in XY plane cancel out each other) (Ashkarran and Mahmoudi 2021b; Ge et al. 2018):

Bz∼B0−2B0Zd (1)

The mechanical equilibrium of the levitating object is achieved due to the balance of gravitational and magnetic forces within the MagLev system, in a 3D Cartesian coordinate system by assuming the z-axis as a symmetry axis of the MagLev:

(χs−χm)μ0(Bx∂Bz∂x+By∂Bz∂y+Bz∂Bz∂z)−(ρs−ρm)g=0 (2)

One can simply derive the relation between the density of the object and its levitation height as:

h=(ρs−ρm)gμ0d2(χs−χm)4B02+d2 (3)

Where, ρm and ρs (kg/m3) are the density of the paramagnetic medium and sample respectively, g (m/s2) is the gravitational acceleration, μ0 (T.m.A−1) is the permeability of free space, d (m) is the distance between the magnets, B0 (tesla) is the magnitude of the magnetic field at the surface of the magnets, χm and χs (unitless) are the magnetic susceptibilities of the paramagnetic medium and the sample, respectively(Ashkarran et al. 2022).

2.3. Characterization

Two blocks of cubic-shaped NdFeB permanent magnets (grade N42, Model # NB044) were obtained from magnet4less (www.magnet4less.com). Magnetic field strength (~ 0.5 T on both magnets’ surfaces) between the magnets was measured by a gauss meter (vector/magnitude Gauss meter model VGM, Alphalab). Levitation profiles of the particles were recorded by a Nikon D750 digital camera containing a 105 mm Nikkor Microlens and a millimetre-scale ruler.

3. Results

We have recently shown that complex protein environments (i.e., human plasma proteins) form a unique ellipsoidal pattern in the presence of SPIONs when levitated in the MagLev system. Formation of the plasma pattern starts a few minutes after injection of the proteins into the MagLev and evolves over three hours (Ashkarran et al. 2020b). Such specific and highly reproducible plasma patterns form for plasma proteins (not single proteins or standard density microspheres; see Supplementary Figure 2 and Supplementary Figure 3 for levitation profiles of single proteins and fluorescent polyethylene microspheres) regardless of the type of the human plasma (e.g., healthy or disease related). The multiple ellipsoidal bands and corresponding particular plasma patterns that appear may be related to i) configurational or structural variation of individual proteins, ii) protein-protein interactions that generate lesser or greater density solution constructs, and/or iii) interaction of proteins and other biomolecules (i.e., lipids and metabolites) with the magnetic field of the MagLev system. Due to the different physicochemical properties of human plasma biomolecules (e.g., difference in density, charge, molecular weight, viscosity, hydrophilic/hydrophobic ratio, mechanical properties, and protein-protein interaction), they respond differently to an external magnetic field, creating a “fingerprint” for each individual human plasma.

To understand the mechanism behind the formation of plasma patterns (see Figure 1 for overall flow of the study), plasma biomolecules of various healthy individuals (n=3) and MS patients with three different types of MS [(i.e., RRMS (n=14), SPMS (n=5), and PPMS (n=3)] were magnetically levitated (Figure 1; see Supplementary Figures 4 to 6 and Supplementary movies 1 to 3 for details on levitation progress and patterns of various types of MS). The plasma samples (n=25) were first levitated in the MagLev system, where they created six ellipsoidal bands of plasma biomolecules, validating the high reproducibility of the patterns of levitated plasma proteins regardless of the source of plasma. It is noteworthy these bands are the layers that appear over the time in all human plasma samples after injection of human plasma to the MagLev system. We observed 5 clear layers from top to bottom of the MagLev column and could successfully distinguish them from each other at 0.06 mg/ml concentration of SPIONs however, in our preliminary experiments we have extracted the bottom part of the paramagnetic solution and found that there are also some proteins in this region, although the band is not very clear and visible by naked eye. Therefore, to make sure we are not missing any invisible layer or proteins, the bottom part of the solution was considered as layer 6 and extracted for possible presence of some proteins.

Figure 1: Schematic representation of the study and plasma protein patterns within the MagLev system.

Figure 1:

Open in a new tab

Formation of ellipsoidal patterns over time from levitated plasma biomolecules (n=25).

To mechanistically explore the driving force that dictates formation of ellipsoidal bands, we extracted the bands layer by layer from top to bottom of the MagLev column and carried out multi-omics analysis of their biomolecular composition to identify proteomic, lipidomic, and metabolomic profiles in all 25 plasma samples.

3.1. Proteomics

We analyzed the bulk plasma proteomes of the 25 plasma samples (3 healthy and 22 MS patients). Quantifying 210 proteins with at least two peptides for each protein, the data were normalized by the sum of all the protein intensities in each sample (Supplementary Data 1). For analysis of various layers in each individual MS sample, 3 healthy samples were used as control and compared with 22 MS patients. For each protein, the abundance was divided by the median intensity of the 3 healthy samples, and the determined value (fold change) was log2 transformed. The corresponding heatmap, presented in Supplementary Figure 7a, represents the 65 proteins quantified across all 25 plasma samples (see Supplementary Data 1 for details of all detected proteins).

Wilcoxon non-parametric sum test was used for testing significance, comparing fold change in 3 healthy samples vs. 22 MS patient samples. Four proteins passed the 0.05 threshold for p value, and the relative abundance of these proteins in all MS vs. healthy samples is shown in boxplots in Supplementary Figure 7b. Volcano plots for each MS patient vs. control (highlighting aforementioned four proteins) are shown in Supplementary Figure 8. Three of these proteins are C1QB (Grewal et al. 1999), IGKC (Torkildsen et al. 2010), and APOE,(Burwick et al. 2006) which have previously been associated with MS pathophysiology, suggesting the validity of our outcomes. However, none of these proteins has previously been identified as a MS biomarker. The last protein was PROS1, and its connection to MS has not been shown before. It is noteworthy that we have previously observed there is no significance difference among proteomics profile of various layers of the produced plasma patterns in the MagLev system (Ashkarran et al. 2020a).

3.2. Lipidomics

To further investigate the effect of the MagLev system on other biomolecules, we conducted lipidomic analysis. Twenty-four plasma samples (3 healthy individuals and 21 MS patients) were subjected to MagLev, and 6 different layers of the emerging plasma patterns of each sample (144 layer in total) were analyzed (one of the samples was removed due to technical issues), and 312 lipid species were reliably identified and quantified across the samples. The raw intensities were normalized by the sum of all lipid intensities in each sample (Supplementary Data 2). Further normalization was performed to adjust the median intensity of lipids to 1 (log2 of 0). For each lipid, abundance was divided by the median intensity of the corresponding layers in 3 healthy samples, and the determined value (fold change) was log2 transformed. The corresponding heatmap depicted in Figure 2 shows the overall lipid intensities across all 144 samples (protein cluster assignment information in Supplementary Data 3). Generally, the lipid profiles show greater variability among individuals compared to the proteomics profiles. While many lipid species are presented uniformly across all samples, lipids in clusters 1, 4, and 5 show great variability between individuals. Therefore, we can deduce that lipids should be partially responsible for the differential MagLev patterns of different MS types vs. healthy individuals.

Figure 2. Heatmap of plasma lipidomics profiles for MS patients vs. healthy subjects.

Figure 2.

Open in a new tab

The relative abundance of plasma lipids in healthy subjects vs. patients with different types of MS shows both the consistency of some lipid features but also great individual variability among the subjects (layers 1 to 6 are shown from left to right for each subject).

In principal component analysis (PCA), principal components (PCs or eigenvectors) are a measure of variation of the dataset. PC1 is the component most explaining the variation in the data, followed by PC2 (till PCn). We performed a PCA of normalized lipid intensities across all the samples and the results are shown in Figure 3a. The PC1 could separate the RRMS samples from others to some extent. The lipids that made the greatest contribution to the separation of samples along the first and second PCA components are highlighted in the PCA loading scatter plot (Figure 3b).

Figure 3. Exploratory analysis of the lipidomics data of MagLev layers for MS vs. healthy samples.

Figure 3.

Open in a new tab

(a) PCA analysis of lipid profiles separates the RRMS samples from other MS types and healthy controls to some extent. But the MS types or different layers are not separated, indicating great variability among subjects. Different colors represent different MS types, and the shapes represent various layers. (b) The lipid species that contribute most to the separation of samples along the first and second principal components are highlighted in the PCA loading scatter plot.

Wilcoxon non-parametric sum test was used to test significance, comparing the fold change of each lipid species in 3 healthy samples vs. 21 MS patient samples in each individual layer; results are shown for individual layers in Figure 4, highlighting the significant outliers. Significant outliers among lipid species can be noted in most layers, especially in layers 1 and 3-6. Most of the outliers are unique to a specific layer.

Figure 4. The relative abundance of lipids in MagLev layers of MS patients vs. healthy subjects.

Figure 4.

Open in a new tab

The lipid species that differ significantly between MS patients and healthy individuals in each MagLev layer (Wilcoxon p value < 0.05).

A number of the identified lipids including glycerophosphochcoline (GPCho) and LysoPC (Boulanger et al. 2000; Del Boccio et al. 2011) have been previously associated with MS pathophysiology, indicating the validity of our analyses. However, none of these lipids have previously been established as MS biomarkers.

Furthermore, we analyzed the data for each layer separately for each MS type and presented the results in volcano plots (Supplementary Figure 9 and Supplementary Data 4). While several lipid species differed significantly in different MagLev layers for SPMS and RRMS, there were no significant outliers in PPMS plots (therefore not shown).

Partial Least Squares analysis is a supervised multivariate data analysis tool that can separate two (or more) groups based on similarity (unlike PCA that is unsupervised). PLS-discriminant analysis (PLS-DA) can also be used to highlight the features/variables that make two groups or systems different (here features would be metabolites and groups can be MS and healthy samples). In other words, PLS-DA can reveal the features with the largest discriminatory power between two groups (variables closest to the model extremities in horizontal axis or P1). A full description of PLS-DA is given in our previous papers (Khoubnasabjafari et al. 2021; Saei et al. 2019; Saei et al. 2021). Therefore, in addition to PCA, we performed PLS analysis (models were built for each layer separately) to investigate whether the model can distinguish healthy individuals from MS patients based on the lipidomics data in each layer. We found that PLS models of all layers could separate healthy from MS samples (see the details of the obtained results for each individual layer from layer 1 to layer 6 in Supplementary Figures 10 and 11). Panel “a” of the figures shows the PLS model, while panel “b” shows the contribution of specific lipids separating the samples. The lipids on the x extremities of panel B are those contributing most to the separation between healthy and MS samples.

Moreover, we investigated whether the PLS models can also discriminate various types of MS based on the lipidomics data from each individual layer separately. Details of the results from PLS models and their loadings are depicted in Supplementary Figures 12 and 13). Among the different MS types, RRMS is better separated from the other MS types, and only layer 5 data clearly separated all the MS types. Therefore, with regards to lipidomics analysis, MagLev layer 5 might be of higher diagnostic value in MS biomarker discovery, which can be potentially further improved by a larger cohort size. In the PLS model for layer 1, SPMS and PPMS are cleanly separated from RRMS in the first component, and then PPMS is separated from the other MS types along the second component (Supplementary Figure 12a).

3.3. Metabolomics

We next performed metabolomics analysis on various layers of the plasma patterns of 24 subjects (3 healthy individuals and 21 MS patients) in the MagLev system. Similarly, 6 different layers of the plasma patterns for each sample (144 layers in total) were separated and analyzed (similar to the lipidomics analysis, one of the MS samples was removed due to technical issues). 4,343 metabolite species were quantified across the samples, with 382 metabolites identified. Raw intensities were normalized by the sum of all metabolite intensities in each sample (Supplementary Data 5). Further normalization was performed to adjust the median intensity of metabolites to 1 (log2 of 0). For each metabolite, abundance was divided by the median intensity of the corresponding layers in 3 healthy samples, the determined value (fold change) was log2 transformed, and the corresponding heatmap (Figure 5) shows the overall metabolite intensities across all 144 samples (cluster assignment information presented in Supplementary Data 6). While many metabolites are presented uniformly across samples, metabolites in clusters 4, 9, and 10 show great variability between individuals. Therefore, similar to lipids, metabolites are also partially responsible for the differential MagLev patterns of different MS types vs. healthy individuals.

Figure 5. Heatmaps of plasma metabolome profiles for MS patients and healthy subjects.

Figure 5.

Open in a new tab

The relative abundance of plasma metabolites in MagLev layers of plasms from healthy subjects vs. patients with different types of MS shows the consistency of some metabolite features but also individual variability among subjects (layers 1 to 6 are shown from left to right for each subject).

A PCA was performed on the normalized metabolite intensities across all the samples, and the results are shown in Figure 6a. The first PCA component distinguished healthy samples from PPMS in the majority of cases. Furthermore, the healthy samples could also be differentiated from RRMS to some extent. The metabolomics profiles of the healthy and SPMS, however, were quite similar. The metabolites that contribute most to the separation of samples along the first and second PCs are highlighted in the PCA loading scatter plot (Figure 6b).

Figure 6. PCA analysis of plasma metabolite profiles show discrimination between samples of healthy origin vs. certain MS types.

Figure 6.

Open in a new tab

(a) PCA analysis of the samples shows that the plasma metabolite profile can distinguish healthy individuals from MS patients (especially PPMS) to some extent, but the MS types or layers are not separated, indicating great variability among subjects, and (b) The metabolites that contribute most to the separation of samples are highlighted in the PCA loading scatter plot. Unidentified metabolites are denoted with NA. The corresponding metabolite IDs are shown after an underscore “_” and can be found in corresponding Supplementary Data files.

Wilcoxon non-parametric sum test was used for testing significance, comparing the fold change of each metabolite in 3 healthy samples vs. 21 MS patient samples in each individual layer, and the results are shown for individual layers in Figure 7, highlighting the significant outliers.

Figure 7. The relative abundance of metabolites in MagLev layers of MS patients vs. healthy individuals.

Figure 7.

Open in a new tab

The top metabolites that are significantly different between MS patients and healthy individuals are highlighted in each layer (p value < 0.05). The majority of significant outliers are unique to individual layers. The metabolites not identified are denoted with NA. The corresponding metabolite IDs are shown after an underscore “_” and can be found in corresponding Supplementary Data files.

A number of these metabolites including homocysteine and glucoheptonic acid have been previously associated with MS pathophysiology, indicating the validity of our analyses (Mititelu et al. 2006). However, none of these metabolites has previously been established as an MS biomarker.

Furthermore, we analyzed the data for each layer separately for each MS type, and the results are presented in volcano plots (Supplementary Figure 14 and Supplementary Data 7). Similar to the lipidomics analysis, while several lipid species were significantly altered in different MagLev layers for SPMS and RRMS, there were no significant outliers in PPMS plots (not shown). This indicates that the lipid and metabolite profiles of SPMS and RRMS are very distinct from those of normal patients.

In addition to PCA, we performed PLS analysis (models were built for each layer separately) to investigate whether the model can separate healthy individuals from MS patients based on metabolite profiles. PLS models of all layers were able to distinguish healthy from MS samples (see details for each individual layer from layer 1 to layer 6 in Supplementary Figures 15 and 16). Panel “a” of the figures shows the PLS model, while panel “b” shows the contribution of specific metabolites separating the samples. The metabolites on the x extremities of the panels “b” are those contributing most to the separation. Moreover, we showed that the PLS models can also discriminate various types of MS based on the metabolomics data from each individual layer separately (Supplementary Figures 17 and 18). Based on component 1 and 2 contributions, layers 1 and 2 could best discriminate the MS types compared to data in the other layers. Therefore, these MagLev layers might have a higher diagnostic value in biomarker discovery for MS in larger cohorts.

4. Discussion

We applied multi-level omics analyses to mechanistically understand the reason for formation of various ellipsoidal bands when levitating plasma biomolecules in the MagLev system. Our findings suggest that the significant variations in composition of lipids and metabolomes are the main driving force for formation of various ellipsoidal bands. In addition, we probed multi-omics information across various plasmas to determine whether differences in ellipsoidal patterns hold potential for disease diagnosis and biomarker discovery.

Applying proteomics, lipidomics, and metabolomics to each individual layer of the plasma patterns from healthy subjects and MS patients, we show that lipidome and metabolome profiles (but not proteome profiles) of human plasma are the major determinants of the produced patterns and have significant diagnostic potential (Del Boccio et al. 2016). While the majority of lipidome and metabolome profiles present in different layers have discriminatory power in biomarker discovery, we demonstrate in this proof-of-concept study that there are particular biomolecules in certain layers of the plasma patterns that have higher diagnostic potential. Although biomarker discovery was not the main focus of the current study, our findings suggest several biomolecules in the plasma patterns that merit further study for discovering novel biomarkers for MS.

Proteomics analysis of various layers of the plasma patterns originating from the MagLev system revealed no significant differences between healthy subjects and MS patients, nor did it fully distinguish different MS subtypes. However, bulk analysis of the plasma samples identified a number of proteins that were significantly different between healthy controls and MS patients. There is cumulative evidence of the involvement of these interesting proteins in neurodegeneration and neuroinflammation (Han et al. 2008). For instance, C1qb plays diverse neuroprotective roles against pathogens and inflammations in the CNS, and levels of C1qb may be disease specific (Cho 2019). IGKC and APOE are other critical proteins that are significantly altered (independent of age and sex) when compared with controls in amyotrophic lateral sclerosis (ALS) disease (Katzeff et al. 2020).

There is growing evidence that changes in the levels of plasma lipids/metabolites may result in neurodegenerative diseases (Di Paolo and Kim 2011). Lipids and their metabolic products play a critical role in the brain, and aberrations in lipid profiles increases the risk of many types of neurodegenerative disorders, including MS. However, to gain a better understanding of the possible role of various biomolecules in the discrimination of MS subtypes, we need to study larger cohorts, due to the large variability among patients with regards to age, sex, disease type, drugs, weight, BMI, etc. (see demographics information in the SI). Nevertheless, it is well documented that many neurodegenerative diseases are related to pathology and accumulation of disordered lipids and metabolites, which causes toxicity to nerve cells and neurodegeneration (Di Paolo and Kim 2011; Ferreira et al. 2020). The lipidomics profile of various layers of MagLev patterns reveals the critical role of particular categories of lipids in various layers of the produced patterns, including glyceroPhosphoCholine (GPC), posphatidylcholines (PC), glycerophosphoethanolamine (GPEtn), sphingomyelins (SM), and triacylglycerols (TG). Our results suggest that there is a specific combination of different lipid categories in each individual layer of the produced plasma patterns that may serve as a “fingerprint” for detection/discrimination of various diseases (Higgs 2010).

Metabolomics is an emerging approach in life science and precision medicine for biomarker discovery in various neurodegenerative diseases such as MS (Bhargava and Calabresi 2016; Noga et al. 2012). Previous studies showed that specific altered metabolites and lipids in plasma samples varied with AD progression (Bhawal et al. 2021). The metabolomic profile of the produced plasma patterns revealed many significantly changing metabolites in different MS types particularly RRMS and SPMS (red dots in Supplementary Figure 13) in all the layers. For example, it has been reported that glutamate may be related to inflammatory and neurodegenerative processes evident in MS (Kostic et al. 2013; Pitt et al. 2000). We have also found that glucoheptonic acid metabolite is significantly different in layers one and two of the produced plasma patterns within the MagLev system. Homocysteine, another significant metabolomic marker of various neurodegenerative disorders including MS, is found in layer one of the plasma patterns (Mititelu et al. 2006). In fact, the level of homocysteine in plasma is a good predictor of both RRMS and SPMS subtypes of MS. Although it has been reported that elevated plasma homocysteine may occur in both mild and progressive disease courses of MS, the concentration of homocysteine metabolite is significantly higher in SPMS compared with RRMS subtypes (Oliveira et al. 2018). In fact, one of the main advantages of this techniques is the capacity of MagLev for biomarker discovery since our findings show that there are particular biomolecules (i.e., lipids and metabolomes) in each layer of the plasma patterns that are significantly different with other layers which have a high diagnostic value (Figures 6 and 9). Since particular lipids and metabolomes are observed in each individual layer of the plasma patterns for each specific MS subtypes, therefore magnetic levitation of plasma biomolecules can be used for rapid MS and its subtypes diagnosis. In other word, for rapid diagnosis purposes we can isolate plasma form MS patients’ blood, levitate them in the MagLev system and after formation of the plasma patters the layers can be extracted for omics analysis. If certain types of the lipids/metabolomes are observed in lipidomics and metabolomic profiles of a sample, we may conclude the patients is diagnosed with MS and also a specific MS subtype. Presence of such metabolites (presented in Supplementary Figure 13) individually or in combination in particular layers of the produced plasma patterns in the MagLev system could predict various types of MS with high diagnostic value.

5. Conclusions

In summary, using multi-omics analysis, we have shown significant differences in the biomolecular composition of various ellipsoidal bands formed by levitating blood plasma in the MagLev system when using SPIONs as a paramagnetic medium. Based on the in-depth compositional analysis of the ellipsoidal bands, it seems that biomolecular variations of each ellipsoidal band and their interactions may contribute, at least in large part, to the formation of distinct band/patterns. In addition, we explored the possible role of ellipsoidal bands in disease detection by comparing the biomolecular composition of each band across various subtypes of MS. Our findings reveal that among other molecular species in the plasma, lipid and metabolite profiles are a major source of variability between different MS types, while plasma proteome profiles are less discriminatory between MS patients and healthy individuals. Therefore, the MagLev technique may be used for lipid and metabolite detection for disease diagnosis through comparing the biomolecular composition of various levitated ellipsoidal bands.

Supplementary Material

Supp Data 7
Supp Data 1
Supp Data 3
Supp Data 2
Supp Data 4
Supp Data 6
Supp Data 5
Movie S1
Download video file (2.8MB, avi)
Movie S2
Download video file (2.9MB, avi)
Movie S3
Download video file (2.5MB, avi)
SI

Acknowledgements

MM gratefully acknowledges financial support from the U.S. National Institute of Diabetes and Digestive and Kidney Diseases (grant DK131417-01). AAS was supported by the Swedish Research Council (grant 2020-00687) and the Swedish Society of Medicine (grant SLS-961262, 1086 Stiftelsen Albert Nilssons forskningsfond). We would like to thank Khashayar Afshari for his valuable discussions. Parts of the Figure 1 are adapted from Bio Render (BioRender.com).

Footnotes

CRediT authorship contribution statement

Ali Akbar Ashkarran: Conceptualization, Methodology, Experiments, Data acquisition, Formal analysis, Project administration, Writing, reviewing & editing original draft. Hassan Gharibi: Conceptualization, Methodology, Formal analysis, Project administration, Writing, reviewing & editing original draft. Dalia Abou Zeki: Writing, reviewing & editing. Irina Radu: Writing, reviewing & editing. Farnaz Khalighinejad: Sample preparation, Writing, reviewing & editing. Kiandokht Keyhanian: Writing, reviewing & editing. Christoffer K. Abrahamsson: Writing, reviewing & editing. Carolina Ionete: Supervision, Project administration, Review & editing. Amir Ata Saei: Supervision, Project administration, Review & editing. Morteza Mahmoudi: Conceptualization, Supervision, Project administration, Reviewing & editing original draft, Funding acquisition.

Declaration of competing interest

The authors declare that they have no conflict of interest.

Data availability

Excel files containing the analyzed raw data are provided in SI (Supplementary Data 1 to 7). Mass spectrometry data will be deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with dataset identifiers.(Vizcaíno et al. 2014)

References

  1. Abrahamsson CK, Nagarkar A, Fink MJ, Preston DJ, Ge S, Bozenko JS, Whitesides GM, 2020. Angewandte Chemie - International Edition 59(2), 874–881. [DOI] [PubMed] [Google Scholar]
  2. Absinta M, Sati P, Reich DS, 2016. Nature Reviews Neurology 12(6), 358–368. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Alseed MM, Rahmani Dabbagh S, Zhao P, Ozcan O, Tasoglu S, 2021. Advanced Optical Technologies 10(2), 109–121. [Google Scholar]
  4. Anil-Inevi M, Delikoyun K, Mese G, Tekin HC, Ozcivici E, 2021. Biotechnology and Bioengineering 118(12), 4771–4785. [DOI] [PubMed] [Google Scholar]
  5. Anil-Inevi M, Yaman S, Yildiz AA, Mese G, Yalcin-Ozuysal O, Tekin HC, Ozcivici E, 2018. Scientific Reports 8(1), 7239. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Ashkarran AA, Mahmoudi M, 2021a. Trends in Biotechnology 39(3), 311–321. [DOI] [PubMed] [Google Scholar]
  7. Ashkarran AA, Mahmoudi M, 2021b. Physics Education 56(5), 055020–055027. [Google Scholar]
  8. Ashkarran AA, Olfatbakhsh T, Ramezankhani M, Crist RC, Berrettini WH, Milani AS, Pakpour S, Mahmoudi M, 2020a. Advanced Healthcare Materials 9(5), 1901608. [DOI] [PubMed] [Google Scholar]
  9. Ashkarran AA, Sharifi S, Abrahamsson CK, Mahmoudi M, 2022. Analytica Chimica Acta 1195, 339369. [DOI] [PubMed] [Google Scholar]
  10. Ashkarran AA, Suslick KS, Mahmoudi M, 2020b. Analytical Chemistry 92(2), 1663–1668. [DOI] [PubMed] [Google Scholar]
  11. Axtell RC, De Jong BA, Boniface K, Van Der Voort LF, Bhat R, De Sarno P, Naves R, Han M, Zhong F, Castellanos JG, Mair R, Christakos A, Kolkowitz I, Katz L, Killestein J, Polman CH, De Waal Malefyt R, Steinman L, Raman C, 2010. Nature Medicine 16(4), 406–412. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Baday M, Ercal O, Sahan AZ, Sahan A, Ercal B, Inan H, Demirci U, 2019. Advanced Healthcare Materials 8, 1801517. [DOI] [PubMed] [Google Scholar]
  13. Bhargava P, Calabresi PA, 2016. Multiple Sclerosis 22(4), 451–460. [DOI] [PubMed] [Google Scholar]
  14. Bhawal RA-O, Fu QA-OX, Anderson ET, Gibson GE, Zhang SA-O, 2021. International Journal of Molecular Science 22(1422-0067 (Electronic)), 13188. [Google Scholar]
  15. Boulanger Y, Labelle M, Khiat A, 2000. Brain Research Reviews 33(2-3), 380–389. [DOI] [PubMed] [Google Scholar]
  16. Burwick RM, Ramsay PP, Haines JL, Hauser SL, Oksenberg JR, Pericak-Vance MA, Schmidt S, Compston A, Sawcer S, Cittadella R, Savettieri G, Quattrone A, Polman CH, Uitdehaag BMJ, Zwemmer JNP, Hawkins CP, Ollier WER, Weatherby S, Enzinger C, Fazekas F, Schmidt H, Schmidt R, Hillert J, Masterman T, Hogh P, Niino M, Kikuchi S, Maciel P, Santos M, Rio ME, Kwiecinski H, Zakrzewska-Pniewska B, Evangelou N, Palace J, Barcellos LF, 2006. Neurology 66(9), 1373–1383. [DOI] [PubMed] [Google Scholar]
  17. Cheng Y, Sun L, Xie Z, Fan X, Cao Q, Han J, Zhu J, Jin T, 2017. Journal of Neuroscience Research 95(10), 1973–1983. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Cho K, 2019. Aging and Disease 10(3), 652–663. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Del Boccio P, Pieragostino D, Di Ioia M, Petrucci F, Lugaresi A, De Luca G, Gambi D, Onofrj M, Di Ilio C, Sacchetta P, Urbani A, 2011. Journal of Proteomics 74(12), 2826–2836. [DOI] [PubMed] [Google Scholar]
  20. Del Boccio P, Rossi C, di Ioia M, Cicalini I, Sacchetta P, Pieragostino D, 2016. Proteomics - Clinical Applications 10(4), 470–484. [DOI] [PubMed] [Google Scholar]
  21. Delikoyun K, Yaman S, Yilmaz E, Sarigil O, Anil-Inevi M, Telli K, Yalcin-Ozuysal O, Ozcivici E, Tekin HC, 2021. ACS Sensors 6(6), 2191–2201. [DOI] [PubMed] [Google Scholar]
  22. DeLuca J, Chiaravalloti ND, Sandroff BM, 2020. Nature Reviews Neurology 16(6), 319–332. [DOI] [PubMed] [Google Scholar]
  23. Deshmukh SS, Shakya B, Chen A, Durmus NG, Greenhouse B, Egan ES, Demirci U, 2021. Communications Biology 4(1), 697. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Di Paolo G, Kim TW, 2011. Nature Reviews Neuroscience 12(5), 284–296. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Durmus NG, Tekin HC, Guven S, Sridhar K, Yildiz AA, Calibasi G, Ghiran I, Davis RW, Steinmetz LM, Demirci U, 2015. Proceedings of the National Academy of Sciences of the United States of America 112(28), E3661–E3668. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Ferreira HB, Neves B, Guerra IM, Moreira A, Melo T, Paiva A, Domingues MR, 2020. Multiple Sclerosis and Related Disorders 44, 102189. [DOI] [PubMed] [Google Scholar]
  27. Gao QH, Wen B, Kang Y, Zhang WM, 2022. Biosensors and Bioelectronics 204, 114052. [DOI] [PubMed] [Google Scholar]
  28. Ge S, Nemiroski A, Mirica KA, Mace CR, Hennek JW, Kumar AA, Whitesides GM, 2020. Angewandte Chemie - International Edition 59(41), 17810–17855. [DOI] [PubMed] [Google Scholar]
  29. Ge S, Wang Y, Deshler NJ, Preston DJ, Whitesides GM, 2018. Journal of the American Chemical Society 140(24), 7510–7518. [DOI] [PubMed] [Google Scholar]
  30. Grewal RP, Morgan TE, Finch CE, 1999. Neuroscience Letters 271(1), 65–67. [DOI] [PubMed] [Google Scholar]
  31. Han MH, Hwang SI, Roy DB, Lundgren DH, Price JV, Ousman SS, Fernald GH, Gerlitz B, Robinson WH, Baranzini SE, Grinnell BW, Raine CS, Sobel RA, Han DK, Steinman L, 2008. Nature 451(7182), 1076–1081. [DOI] [PubMed] [Google Scholar]
  32. Higgs R, 2010. Nature Reviews Neurology 6(1), 6. [Google Scholar]
  33. Katzeff JS, Bright F, Lo K, Kril JJ, Connolly A, Crossett B, Ittner LM, Kassiou M, Loy CT, Hodges JR, Piguet O, Kiernan MC, Halliday GM, Kim WS, 2020. Scientific Reports 10(2045-2322 (Electronic)), 13741. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Khoubnasabjafari M, Mogaddam MRA, Rahimpour E, Soleymani J, Saei AA, Jouyban A, 2021. Critical Reviews in Analytical Chemistry. [DOI] [PubMed] [Google Scholar]
  35. Knowlton SM, Sencan I, Aytar Y, Khoory J, Heeney MM, Ghiran IC, Tasoglu S, 2015. Scientific Reports 5, 15022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Knowlton SM, Yenilmez B, Amin R, Tasoglu S, 2017. Journal of Visualized Experiments 120(120), 55012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Kostic M, Zivkovic N, Stojanovic I, 2013. Reviews in the Neurosciences 24(1), 71–88. [DOI] [PubMed] [Google Scholar]
  38. Mititelu RR, Albu CV, Bacanoiu MV, Padureanu V, Padureanu R, Olaru G, Buga AM, Balasoiu M, 2006. Journal of Neurology, Neurosurgery and Psychiatry 9(2359-7232 (Electronic)), e135. [Google Scholar]
  39. Noga MJ, Dane A, Shi S, Attali A, van Aken H, Suidgeest E, Tuinstra T, Muilwijk B, Coulier L, Luider T, Reijmers TH, Vreeken RJ, Hankemeier T, 2012. Metabolomics 8(2), 253–263. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Oliveira SR, Flauzino T, Sabino BS, Kallaur AP, Alfieri DF, Kaimen-Maciel DR, Morimoto HK, de Almeida ERD, Lozovoy MAB, Reiche EMV, Dichi I, Simão ANC, 2018. Metabolic Brain Disease 33(1573-7365 (Electronic)), 1393–1399. [DOI] [PubMed] [Google Scholar]
  41. Ozefe F, Arslan Yildiz A, 2020. Analyst 145(17), 5816–5825. [DOI] [PubMed] [Google Scholar]
  42. Parfenov VA, Khesuani YD, Petrov SV, Karalkin PA, Koudan EV, Nezhurina EK, Pereira FDAS, Krokhmal AA, Gryadunova AA, Bulanova EA, Vakhrushev IV, Babichenko II, Kasyanov V, Petrov OF, Vasiliev MM, Brakke K, Belousov SI, Grigoriev TE, Osidak EO, Rossiyskaya EI, Buravkova LB, Kononenko OD, Demirci U, Mironov VA, 2020a. Science Advances 6(29). [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Parfenov VA, Mironov VA, Koudan EV, Nezhurina EK, Karalkin PA, Pereira FD, Petrov SV, Krokhmal AA, Aydemir T, Vakhrushev IV, Zobkov YV, Smirnov IV, Fedotov AY, Demirci U, Khesuani YD, Komlev VS, 2020b. Scientific Reports 10(1), 4013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Pitt D, Werner P, Raine CS, 2000. Nature Medicine 6(1), 67–70. [DOI] [PubMed] [Google Scholar]
  45. Puluca N, Durmus NG, Lee S, Belbachir N, Galdos FX, Ogut MG, Gupta R, Hirano KI, Krane M, Lange R, Wu JC, Wu SM, Demirci U, 2020. Advanced Biosystems 4, 1900300. [DOI] [PubMed] [Google Scholar]
  46. Reich DS, Lucchinetti CF, Calabresi PA, 2018. New England Journal of Medicine 378(2), 169–180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Saei AA, Beusch CM, Chernobrovkin A, Sabatier P, Zhang B, Tokat ÜG, Stergiou E, Gaetani M, Végvári Á, Zubarev RA, 2019. Nature Communications 10(1), 5715. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Saei AA, Beusch CM, Sabatier P, Wells JA, Gharibi H, Meng Z, Chernobrovkin A, Rodin S, Näreoja K, Thorsell AG, Karlberg T, Cheng Q, Lundström SL, Gaetani M, Végvári Á, Arnér ESJ, Schüler H, Zubarev RA, 2021. Nature Communications 12(1), 1296. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Sarigil O, Anil-Inevi M, Firatligil-Yildirir B, Unal YC, Yalcin-Ozuysal O, Mese G, Tekin HC, Ozcivici E, 2021. Biotechnology and Bioengineering 118(3), 1127–1140. [DOI] [PubMed] [Google Scholar]
  50. Sarigil O, Anil-Inevi M, Yilmaz E, Mese G, Tekin HC, Ozcivici E, 2019. The Analyst 144(9), 2942–2953. [DOI] [PubMed] [Google Scholar]
  51. Sözmen AB, Arslan-Yıldız A, 2022. Biosensors and Bioelectronics 10, 100137. [Google Scholar]
  52. Tasoglu S, Gurkan UA, Wang S, Demirci U, 2013. Chemical Society Reviews 42(13), 5788–5808. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Tasoglu S, Yu CH, Liaudanskaya V, Guven S, Migliaresi C, Demirci U, 2015. Advanced Healthcare Materials 4(10), 1469–1476. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Tocchio A, Durmus NG, Sridhar K, Mani V, Coskun B, El Assal R, Demirci U, 2017. Advanced Materials 30(4), 1705034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Torkildsen O, Stansberg C, Angelskår SM, Kooi EJ, Geurts JJG, Van Der Valk P, Myhr KM, Steen VM, Bø L, 2010. Brain Pathology 20(4), 720–729. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Turker E, Arslan-Yildiz A, 2018. ACS Biomaterials Science and Engineering 4(3), 787–799. [DOI] [PubMed] [Google Scholar]
  57. Vaughn CB, Jakimovski D, Kavak KS, Ramanathan M, Benedict RHB, Zivadinov R, Weinstock-Guttman B, 2019. Nature Reviews Neurology 15(6), 329–342. [DOI] [PubMed] [Google Scholar]
  58. Vizcaíno JA, Deutsch EW, Wang R, Csordas A, Reisinger F, Ríos D, Dianes JA, Sun Z, Farrah T, Bandeira N, Binz PA, Xenarios I, Eisenacher M, Mayer G, Gatto L, Campos A, Chalkley RJ, Kraus HJ, Albar JP, Martinez-Bartolomé S, Apweiler R, Omenn GS, Martens L, Jones AR, Hermjakob H, 2014. Nature Biotechnology 32(3), 223–226. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Vrenken H, Seewann A, Knol DL, Polman CH, Barkhof F, Geurts JJG, 2010. American Journal of Neuroradiology 31(3), 541–548. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Wootla B, Eriguchi M, Rodriguez M, 2012. Autoimmune diseases 2012, 969657. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Yaman S, Tekin HC, 2020. Analytical Chemistry 92(18), 12556–12563. [DOI] [PubMed] [Google Scholar]
  62. Yang K, Wu J, Santos S, Liu Y, Zhu L, Lin F, 2019. Biosensors and Bioelectronics 124-125, 150–160. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supp Data 7
NIHMS1851084-supplement-Supp_Data_7.csv (3.7MB, csv)
Supp Data 1
NIHMS1851084-supplement-Supp_Data_1.csv (42.7KB, csv)
Supp Data 3
NIHMS1851084-supplement-Supp_Data_3.csv (6.9KB, csv)
Supp Data 2
NIHMS1851084-supplement-Supp_Data_2.csv (723KB, csv)
Supp Data 4
NIHMS1851084-supplement-Supp_Data_4.csv (332.3KB, csv)
Supp Data 6
NIHMS1851084-supplement-Supp_Data_6.csv (80.7KB, csv)
Supp Data 5
NIHMS1851084-supplement-Supp_Data_5.csv (10.2MB, csv)
Movie S1
Download video file (2.8MB, avi)
Movie S2
Download video file (2.9MB, avi)
Movie S3
Download video file (2.5MB, avi)
SI
NIHMS1851084-supplement-SI.pdf (9.8MB, pdf)

Data Availability Statement

Excel files containing the analyzed raw data are provided in SI (Supplementary Data 1 to 7). Mass spectrometry data will be deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with dataset identifiers.(Vizcaíno et al. 2014)

RESOURCES