Multiparametric platform for profiling lipid trafficking in human leukocytes

Summary Systematic insight into cellular dysfunction can improve understanding of disease etiology, risk assessment, and patient stratification. We present a multiparametric high-content imaging platform enabling quantification of low-density lipoprotein (LDL) uptake and lipid storage in cytoplasmic droplets of primary leukocyte subpopulations. We validate this platform with samples from 65 individuals with variable blood LDL-cholesterol (LDL-c) levels, including familial hypercholesterolemia (FH) and non-FH subjects. We integrate lipid storage data into another readout parameter, lipid mobilization, measuring the efficiency with which cells deplete lipid reservoirs. Lipid mobilization correlates positively with LDL uptake and negatively with hypercholesterolemia and age, improving differentiation of individuals with normal and elevated LDL-c. Moreover, combination of cell-based readouts with a polygenic risk score for LDL-c explains hypercholesterolemia better than the genetic risk score alone. This platform provides functional insights into cellular lipid trafficking and has broad possible applications in dissecting the cellular basis of metabolic disorders.


In brief
Insights into cellular dysfunction underlying hypercholesterolemia are lacking. Pfisterer et al. establish an automated analysis platform enabling quantification of multiple cellular readouts, including lipid uptake, storage, and mobilization, from different white blood cell populations. This approach provides personalized insights into the cellular basis of hypercholesterolemia and obesity.

INTRODUCTION
Hypercholesterolemia is one of the most common metabolic disorders and a major risk factor for cardiovascular disease (CVD). It is characterized by an accumulation of low-density lipoprotein cholesterol (LDL-c) in the blood (Boré n et al., 2020). In familial hypercholesterolemia (FH), mutations, most commonly in the LDL receptor (LDLR) gene, lead to increased LDL-c. However, FH represents only 2.5% of all hypercholesterolemia patients. For the remainder, polygenic and lifestyle effects appear as the main contributing factors (Abul-Husn et al., 2016;Khera et al., 2016;Ripatti et al., 2020;. So far, we have little information on how cellular lipid trafficking and storage are altered in individual patients. However, systematic assessment of LDL uptake and other mechanisms related to hypercholesterolemia could provide insights into disease MOTIVATION We have limited information on how cellular lipid uptake and processing differ between individuals and influence the development of metabolic diseases, such as hypercholesterolemia. Available assays are labor intensive, require skilled personnel, and are difficult to scale to higher throughput, making it challenging to obtain systematic, functional cell-based data from individuals. To overcome this problem, we established a scalable automated analysis pipeline enabling reliable quantification of multiple cellular readouts, including lipid uptake, storage, and mobilization, from different white blood cell populations. This approach provides personalized insights into the cellular basis of hypercholesterolemia and obesity. mechanisms and treatment outcomes in a personalized manner. The majority of high-risk hypercholesterolemia patients do not achieve their LDL-c target levels (Ray et al., 2020). This could be due to sub-optimal treatment, non-adherence to therapy, and/or cellular programs limiting drug efficacy. Increased evidence from cancer therapy demonstrates that cell-based assays can provide better targeted and more effective personalized treatment strategies (Snijder et al., 2017). Regarding hypercholesterolemia, we need to establish scalable and reliable assays that allow systematic profiling of functional defects in individual persons and evaluate how to utilize such assays to better explain factors contributing to hypercholesterolemia in individual patients. The currently used cell-based assays for studying the etiology of hypercholesterolemia are quantification of cellular LDL uptake or LDLR cell surface expression using flow cytometry. These readouts have been mostly utilized to characterize the severity of LDLR mutations in FH patients (Benito-Vicente et al., 2018;Romano et al., 2010). However, LDLR surface expression and LDL uptake are highly variable among FH patients (Tada et al., 2009;Thedrez et al., 2018;Urdal et al., 1997). This not only speaks for the importance of functional cell-based assays but also calls for additional cellular readouts to better characterize the heterogeneity of lipid metabolism in individual subjects.
LDLR expression and cellular LDL internalization are tightly regulated. Low cholesterol levels in the endoplasmic reticulum (ER) signal cholesterol starvation and trigger increased LDLR expression, while high cholesterol in the ER downregulates LDLR expression. Excess ER cholesterol is stored as cholesterol ester in lipid droplets (LDs), from where it can be mobilized upon need (Ikonen, 2008;Luo et al., 2020). We therefore considered that quantification of cellular LDs and their dynamic changes upon altering lipoprotein availability may provide additional information for assessing the cellular basis of hypercholesterolemia.
Here, we established sensitive and scalable analyses for automated quantification of fluorescent lipid uptake, storage, and removal in primary lymphocyte and monocyte populations and defined lipid mobilization as an additional parameter measuring how efficiently cells deplete their lipid stores. We found marked differences in the parameters established in both FH and non-FH study groups and highlight their potential to provide deeper insights into the cellular mechanisms of hypercholesterolemia.

RESULTS
Automated pipeline for quantification of hypercholesterolemia-related functional defects in primary human leukocytes Several cell types, such as lymphocytes, monocytes, and Epstein-Barr virus (EBV) immortalized lymphoblasts, have been used for measuring LDL uptake (Chan et al., 1997;Schmitz et al., 1993). While EBV lymphoblasts show the highest LDL uptake, cell immortalization is time consuming and alters cellular functions (Chan et al., 1997;Piccaluga et al., 2018). We therefore set up an automated imaging and analysis pipeline for sensitive quantification of LDL uptake and LDLR surface expression from less than two million peripheral blood mononuclear cells (PBMCs) ( Figure 1A). Cryopreserved PBMCs were recovered in 96-well plates at defined densities and incubated with lipid-rich control medium (CM) (10% fetal bovine serum [FBS]) or lipid poor medium (LP) (5% lipoprotein-deficient serum) for 24 h. Cells were labeled with fluorescent LDL particles (DiI-LDL) for 1 h, washed, and automatically transferred to 384-well plates for staining and automated high-content imaging ( Figure 1A). After adhesion to coated imaging plates, lymphocytes remain small while monocytes spread out, enabling a crude classification of leukocyte populations based on size: PBMCs with a cytoplasmic area <115 mm 2 were classified as a lymphocyte-enriched fraction (from here on lymphocytes) and those with a cytoplasmic area >115 mm 2 as monocyte-enriched fraction (from here on monocytes; Figures S1A-S1C).
In CM, DiI-LDL uptake into lymphocytes and monocytes was more than 2-fold above the background of non-labeled cells . Lipid starvation further increased DiI-LDL uptake in both cell populations, as expected ( Figures 1C and 1D). We quantified about 700 monocytes and 2,300 lymphocytes per well (Figure S1D), aggregated the single-cell data from individual wells, and averaged the results from 2-4 wells for each treatment (Figure S1D). For both cell populations, we defined 2 readouts: cellular DiI-LDL intensity (DiI-Int), reflecting DiI-LDL surface binding and internalization, and DiI-LDL organelle number (DiI-No), reflecting internalized DiI-LDL ( Figures 1E and 1F). This resulted in 4 parameters: monocyte (Mo) DiI-Int, lymphocyte (Ly) DiI-Int, Mo DiI-No, and Ly DiI-No. In both cell populations, DiI-Int was inhibited by adding surplus unlabeled LDL, arguing for a saturable, receptormediated uptake mechanism ( Figure S1E).
In lipid-rich conditions, Mo DiI-Int was slightly higher than Ly DiI-Int ( Figure 1E), and upon lipid starvation, Mo DiI-Int increased more substantially, providing a larger fold increase than Ly DiI-Int ( Figure 1E). Furthermore, Mo DiI-No was roughly 10-fold higher than Ly DiI-No, with both parameters showing a 5-fold increase upon lipid starvation ( Figure 1F). Thus, DiI-LDL uptake into monocytes was better than into lymphocytes, but both cell populations responded to lipid starvation. As EBV lymphoblasts are often a preferred choice for LDL uptake studies (Chan et al., 1997), we compared LDL uptake between EBV lymphoblasts and monocytes ( Figures S1F and S1G). This showed that DiI-Int signal after lipid starvation was roughly similar in EBV lymphoblasts and monocytes, implying that the primary cells provide high enough DiI-LDL signal intensities without cell immortalization ( Figure S1G (B) Quantification of monocyte (Mo) and lymphocyte (Ly) cellular DiI-LDL intensities (Int.), organelle numbers (No.), and pan-uptake normalized to 2 controls (100%); 2 to 3 independent experiments, each with duplicate or quadruplicate wells per patient (8-16 wells per patient for pan-uptake). Cys325Tyr and Ser580Phe were described in Figures 1G and 1H. Significant changes to control two were calculated with Welch's t test. To enable data comparison between experiments, we included 2 controls. Each control consisted of a mixture of large-scale PBMC isolations from 4 healthy blood donors, with the cells cryopreserved at a defined density for one-time-use aliquots. In each experiment, Mo DiI-Int, Ly DiI-Int, Mo DiI-No, and Ly DiI-No were normalized to these controls. We also introduced a combinatorial score, pan-LDL uptake (or pan-uptake), representing the average of Mo DiI-Int, Ly DiI-Int, Mo DiI-No, and Ly DiI-No. We then assessed the intraindividual variability of these 5 readouts in 3 individuals on 2 consecutive days ( Figure S1H). The intraindividual variability was low for a cell-based assay, especially in monocytes, with 7.6% for Mo DiI-No, 12% for Mo DiI-Int, and 13% for pan-uptake. The values were only moderately higher in lymphocytes, with DiI-Int 15% and DiI-No 21% variability ( Figure S1I).
We next validated our LDL uptake measurements in PBMCs of 2 He-FH patients with highly elevated LDL-c and reduced LDL uptake in EBV lymphoblasts (Cys325Tyr and Ser580Phe mutations in LDLR; Figure S1J). For both patients, Mo and Ly DiI-No as well as Mo DiI-Int were reduced by more than 45%, Ly DiI-Int was less profoundly decreased, and pan-uptake was reduced by over 50% (Figures 1G, 1H, and S1J). Together, these data indicate that our analysis pipeline enables quantification of multiple LDL uptake parameters in major leukocyte cell populations and distinguishes defective LDLR function therein.
Heterogeneous LDL uptake and LDLR surface expression in He-FH patients We next used this pipeline to characterize 21 He-FH patients from the metabolic syndrome in men (METSIM) cohort study ; Table S1). The patients' mutations reside in the LDLR coding region and range from pathogenic to likely benign variants ( Figure 2A). Quantification of DiI-Int and DiI-No for monocytes and lymphocytes provided relatively similar results for each individual ( Figure 2B). However, there were substantial differences in these parameters between individuals, including patients harboring identical LDLR mutations (Figure 2B). This was most pronounced for FH-North Karelia (Pro309Lysfs*59), a pathogenic loss-of-function variant but also evident for FH-Pogosta (Arg595Gln) and FH-Glu626Lys (Figures 2A and 2B). These observations imply that, in He-FH, regulatory mechanisms may enhance the expression of the unaffected LDLR allele and/or stabilize the encoded protein. In support of this notion, we obtained a strong correlation between monocyte LDLR surface expression and DiI-Int, DiI-No, and pan-uptake scores for the same individuals (pan-uptake; R = 0.58; p = 0.006; Figures 2C and S2A).
Interestingly, the pan-uptake score showed a tendency for lower values in FH-North Karelia carriers as compared with those carrying the likely pathogenic FH-Pogosta and likely benign Glu626Lys variants ( Figure S2B). This is in agreement with higher LDL-c concentrations in FH-North Karelia patients (Lahtinen et al., 2015). While LDL uptake did not correlate with circulating LDL-c for the entire study group ( Figure S2C), this correlation was highly significant for monocyte DiI-Int, DiI-No, and the pan-uptake scores for the 11 He-FH patients on statin monotherapy (Mo DiI-Int: R = À0.75; p = 0.0081; Figure 2D). Notably, three of the individuals with the lowest monocyte DiI-Int had a 2-fold higher LDL-c concentration than the 3 individuals with the highest monocyte DiI-Int; Figure 2E), suggesting that the LDL-clowering effect of statin is reflected by monocyte LDL uptake. This is likely due to the higher LDL uptake capacity of monocytes as compared with lymphocytes ( Figures 1E and 1F).
LDL uptake in non-FH individuals with normal or elevated circulating LDL-c As most hypercholesterolemia patients do not carry LDLR mutations, we also investigated cellular LDL uptake in PBMCs from 20 biobank donors with elevated LDL-c levels (LDL-c > 5 mM) (hLDL-c) and from 19 donors with normal LDL levels (LDL-c 2-2.5 mM) (nLDL-c) from the FINRISK population cohort (Borodulin et al., 2018; Table S2). DNA sequencing confirmed that common Finnish LDLR variants were not present among these subjects.
We quantified DiI-Int and DiI-No for monocyte and lymphocyte populations as well as the pan-uptake score for nLDL-c and hLDL-c individuals. This revealed a large interindividual variation in LDL uptake ( Figure 3A). Both groups included persons with severely reduced LDL internalization, although the lowest pan-LDL uptake scores were among the hLDL-c individuals (Figure 3A). Overall, pan-uptake and Ly DiI-No were reduced in hLDL-c compared with nLDL-c subjects, but the differences were not significant ( Figures S3A and S3B). Of note, reduced pan-uptake, Mo DiI-Int, and Ly DiI-No correlated with increased serum LDL-c levels in the hLDL-c subgroup, but the correlations relied on a single individual with a very high serum LDL-c concentration (pan-uptake: R = À0.49; p = 0.028; Figure S3C).

Assessment of cellular lipid storage and mobilization in leukocytes
Cells store excess lipids in LDs, and this is related to lipid uptake: when peripheral cells have sufficient lipids available, they typically exhibit LDs and, in parallel, lipid uptake is downregulated. We therefore also included the staining of LDs in the automated analysis pipeline ( Figure 1A). Staining of PBMCs in lipid-rich conditions (CM) with the well-established LD dye LD540 ) revealed that lymphocytes and monocytes displayed LDs in a heterogenous fashion ( Figure 4A), with lymphocytes showing fewer LD-positive cells and fewer LDs per cell than monocytes (Figures 4B and 4C). We then visualized the changes in LD abundance upon overnight lipid starvation in lipoprotein-deficient medium (LP; Figures 4B-4F). This resulted in a pronounced decrease in lipid deposition: in CM, 9% of lymphocytes and 25% of monocytes contained LDs, but upon lipid starvation, these were reduced to 6% (Ly) and 12% (Mo; Figure 4D).
Due to the lower LD abundance in lymphocytes, we focused on monocytes and defined 3 readouts for them: (1) percentage of LD-positive cells (LD-Pos), (2) cellular LD number in LD-Pos (LD-No), and (3) total cellular LD area in LD-Pos (LD-Area). On average, LD-Pos cells showed 2.9 LDs in lipid-rich conditions and 1.8 LDs upon lipid starvation ( Figure 4E), while the total LD area decreased from 1.35 mm 2 in lipid-rich conditions to 0.8 mm 2 upon lipid starvation ( Figure 4F).
When quantifying LD parameters from several subjects, we observed substantial differences between individuals in how LDs changed upon starvation. To systematically quantify these differences, we established a parameter, lipid mobilization score that reflects how efficiently cellular lipid stores are depleted under lipid starvation ( Figure 4G). Lipid mobilization scores were calculated for each of the LD readouts, LD-Pos, LD-No, and LD-Area, by dividing the results obtained in lipid-rich conditions with those obtained after lipid starvation ( Figure 4G). Furthermore, we established a pan-mobilization score by averaging LD-Pos, LD-No, and LD-Area scores ( Figures 4G and 4H), with LD-Pos providing the highest mobilization score but also the highest variability ( Figure 4H).
To further assess the reliability of the LD mobilization parameters, we determined their intraindividual variation using the same samples as for analyzing intraindividual variation of DiI-LDL uptake ( Figures S1I and S1J). This showed a modest intraindividual variation for the lipid mobilization scores ( Figure S4A), with an average of 8% for pan-mobilization, 10% for LD-Pos, 11% for LD-No, and 13% for LD-Area ( Figure S4B).

Cellular lipid mobilization in He-FH patients
When lipid mobilization was analyzed from the He-FH samples of the METSIM cohort, we found that the pan-mobilization score was significantly reduced in He-FH individuals carrying the FH-North Karelia and Glu626Lys variants ( Figure 4I). This suggests that defective LDLR function may be accompanied by reduced lipid mobilization. We also studied whether the combination of a lipid mobilization score with LDL uptake improves identification  Figure 2D). When monocyte DiI-Int was combined with the pan-mobilization score, larger differences between patients were observed, providing a better separation of individuals with high and intermediate LDL-c ( Figure 4J). Moreover, the difference in LDL-c concentration between the 3 individuals with the highest versus lowest score was more significant than when using monocyte DiI-Int alone ( Figure 4K versus Figure 2E). This suggests that the combined LDL uptake and lipid mobilization assays may help to better pinpoint those He-FH cases that remain refractory to statin monotherapy.

Cellular lipid mobilization is reduced in non-FH patients and correlates with LDL uptake
We then investigated whether monocytes from nLDL-c and hLDL-c biobank donors displayed differences in lipid mobilization. Analogously to LDL uptake, we observed a large variability for the pan-and individual mobilization scores in this cohort (Figure 5A). Interestingly, pan-mobilization, LD-No, and LD-Area were significantly reduced in the hLDL-c compared with nLDLc subjects ( Figures 5A, 5B, S5A, and S5B). This prompted us to scrutinize whether lipid mobilization correlates with LDL-uptake-related parameters in this cohort. All mobilization scores correlated positively with the pan-uptake score (R = 0.42; p = 0.0095 for pan-mobilization; Figure 5C). Furthermore, pan-, LD-No, and LD-Area mobilization scores correlated negatively with total cholesterol, apo-B concentrations (Figures S5C and S5D), and with age (R = À0.38, p = 0.019 for pan-mobilization; Figure 5D).

Hybrid scores of genetic and functional cell-based data show improved association with hypercholesterolemia
The hLDL-c biobank donors of the FINRISK population cohort displayed an increased LDL-c polygenic risk score (LDL-PRS) ( Figure 6A). LDL-PRS did not correlate with LDL uptake or lipid mobilization ( Figures S6A and S6B), suggesting that LDL-PRS and cellular LDL uptake monitor, in part, distinct processes. Interestingly, combination of LDL-PRS with pan-uptake reduced the variation and made it easier to discriminate the nLDL-c and hLDL-c groups, providing an 8-times-better p value as compared with LDL-PRS only ( Figure 6B). Furthermore, combination of the pan-mobilization score with LDL-PRS drastically improved the discrimination between groups ( Figure 6C), and combining all 3 parameters, i.e., LDL-PRS, pan-uptake, and pan-mobilization, provided the best discrimination power and lowest p value (Figure 6D). To further highlight the benefits of combining genetic and functional cell data, we calculated the odds ratio (OR) for elevated LDL-c by comparing individuals with the highest 30% of the scores to the remaining subjects. Interestingly, combining LDL-PRS with either pan-uptake or pan-mobilization doubled the OR, and using a hybrid score combining all 3 readouts resulted in a 5-fold higher OR ( Figure 6E). The odds for having elevated LDLc was 21 times higher for a person within the highest 30% of the triple hybrid score, as compared with the remaining subjects, highlighting the strength of functional hybrid scores. This is further supported by calculating the OR for 25%, 30%, 35%, and 40% of the individuals with the highest LDL-PRS, double or triple hybrid scores, and the remaining subjects, which in almost all instances provided higher OR for hybrid scores than for LDL-PRS ( Figure S6C).

DISCUSSION
In this study, we established a multiplexed analysis pipeline to quantify lipid uptake and mobilization in primary leukocytes and used it to analyze over 300 conditions (combinations of assays and treatments) from 65 individuals. The automated cell handling, staining, and imaging procedures enable highthroughput applications. Key advantages of the method are (1) large-scale internal standards allow comparison of experimental results over time; (2) automated cell quantification avoids researcher bias, increasing reliability of results; (3) semi-automated workflow can be scaled to increase throughput; (4) cell immobilization on coated surfaces allows flexibility in sample handling and facilitates automation, (5) lymphocyte-and monocyte-enriched cell populations can be detected based on cell spreading on coated surfaces; and (6) subcellular resolution enables quantification of internalized LDL and LDs, yielding additional scores derived from them. In conventional flow cytometry assays, cells are quantified when passing through a capillary, providing mean cellular intensities without subcellular resolution.
The cells need to be in suspension, and cell aggregation can obstruct the capillary. This complicates cell handling and requires centrifugation steps for cell washing, making it more challenging to automate the assays. Consequently, the first two aspects can be readily included in flow cytometry assays while the latter four rely on a high-content, high-resolution imaging platform. Several of the observations made using this analysis pipeline are supported by previous findings obtained using manual assays, thereby validating our results. We showed that monocytes display higher LDL uptake activities than lymphocytes, in accordance with previous findings (Schmitz et al., 1993). The highly variable LDL uptake observed by us between individuals, including He-FH patients with identical LDLR mutations, also agrees with earlier reports (Tada et al., 2009;Thedrez et al., 2018;Urdal et al., 1997). Furthermore, we observed an association of low cellular LDL uptake with increased circulating LDL-c in He-FH patients on statin monotherapy, in line with studies utilizing radiolabeled LDL (Gaddi et al., 1991;Hagemenas and Illingworth, 1989;Hagemenas et al., 1990;Sun et al., 1998). However, this finding was not readily reproduced by using (I) Pan-mobilization for controls (combined control one and two from 5 experiments), FH-North-Karelia (n = 7), FH-Pogosta (n = 3), and FH-Glu626 (n = 5). (J) Correlation of combined monocyte mean DiI-LDL intensities (Mo Int) and pan-mobilization with circulating LDL-c. (K) LDL-c concentration for 3 patients with the highest (high) and lowest (low) combined score as in (J). *p < 0.05 and **p < 0.01. Error bars represent SEM.
8 Cell Reports Methods 2, 100166, February 28, 2022 Article ll OPEN ACCESS fluorescently labeled LDL particles in lymphocytes (Homma et al., 2015;Raungaard et al., 2000). Indeed, our results indicate that monocytes provide an improved detection window and a better correlation between cellular LDL uptake and circulating LDL-c.
We also found that reduced LDL uptake correlated with increased BMI and waist circumference, two obesity indicators. Metabolic syndrome is typically linked to dyslipidemia characterized by decreased high-density lipoprotein cholesterol (HDL-c), elevated LDL-c with increased small, dense LDL particles, and increased plasma triglycerides (Klop et al., 2013). Our results suggest that, besides VLDL overproduction and defective lipolysis of triglyceride (TG)-rich lipoproteins (Boré n et al., 2020), reduced LDL clearance may contribute to dyslipidemia in overweight individuals. This fits with the observed reduction of LDLR expression in obese subjects (Mamo et al., 2001).
Moreover, we employed the platform to quantify cellular LDs, established a parameter termed lipid mobilization score, and demonstrated its ability to provide additional data on individual differences on lipid handling. Lipid mobilization correlated with LDL uptake, implying that efficient removal of stored lipids was typically paralleled by efficient lipid uptake. Moreover, combining monocyte LDL uptake and lipid mobilization data facilitated the detection of He-FH cases that remained hypercholesterolemic on statin. In the FINRISK population cohort, lipid mobilization outperformed LDL uptake in distinguishing individuals with high (>5 mmol/L) and normal LDL-c (2-2.5 mmol/L), with impaired lipid mobilization associating with elevated LDL-c. Hence, lipid mobilization shows potential to highlight additional aspects of cellular lipid metabolism underlying hypercholesterolemia in individual patients.
Polygenic risk scores (PRSs) provide tools for cardiovascular risk profiling and are increasingly included in clinical care guidelines of hypercholesterolemia (Boré n et al., 2020;Mach et al., 2019). We found that the hypercholesterolemia subjects of the FINRISK cohort had an increased LDL-PRS, but this did not correlate with LDL uptake or lipid mobilization, arguing that the cell-based parameters cover in part different territories than PRS. In agreement, the combination of LDL uptake, lipid mobilization, and LDL-PRS improved the segregation of hyper-and normocholesterolemic subjects. An increased LDL-PRS is associated with a higher incidence of coronary artery disease (Ripatti et al., 2020). We therefore anticipate that the cell-based assays may provide additional information for future integrated CVD risk calculations. These, in turn, might facilitate the detection of hypercholesterolemia risk at younger age when clinical manifestations are not yet overt, enabling faster initiation of treatment and improved disease prevention (Wiegman et al., 2015). In summary, the automated analysis platform established here enables systematic assessment of cellular lipid trafficking in accessible primary cell samples of human origin. Besides hypercholesterolemia, this approach can be useful in other metabolic disorders, as well as diseases not previously linked to cellular lipid imbalance. As an example of the latter, we recently uncovered aberrant LD size distribution in MYH9-related disease patient neutrophils using quantitative imaging .

Limitations of the study
We analyzed 65 individuals as a proof of concept for the analysis platform. While this outperforms most previous studies measuring lipid uptake in primary cells, further validation in larger study groups will be required to assess its potential clinical utility. Such studies will be feasible due to the high automation level of the platform, enabling processing of samples from several thousand subjects per year. In particular, the finding that combined LDL uptake and lipid mobilization assays may improve the detection of He-FH cases that remain refractory to statin monotherapy relies on the small number of such individuals in the current study and awaits validation with additional He-FH patients on cholesterol-lowering medication.
Regarding the cellular origin of hypercholesterolemia, we infer parameters related to whole-body metabolism and in particular liver function from PBMCs. Evidently, primary hepatocytes would provide more direct information but are not accessible on a routine basis. PBMCs are easily obtained from standard blood collections. Moreover, our data demonstrate that PBMC-derived parameters can correlate with readouts deriving from the whole body level.
Currently, the analysis platform is set up to quantify two cellular parameters: LDL uptake and lipid storage in droplets. In the present conditions with minimally modified cells, only a fraction of cells (9% of lymphocytes and 25% of monocytes) (E) Odds ratio (OR) for 30% of the individuals with the highest LDL-PRS, double or triple hybrid scores, and the remaining subjects, calculated with the Fisher's exact probability test; n = 36. The OR for genetic and the hybrid scores are above one, indicating that a person with a high score is more likely to have elevated LDL-c. The significance tests evaluate the likelihood that an OR different from 1 has been obtained by chance. For the combination of LDL-PRS with the functional cell data, this likelihood is very low and our results are significant, while for LDL-PRS alone, this is not the case. nLDL-c n = 18 and hLDL-c n = 18; *p < 0.05, **p < 0.01, and ***p < 0.001; Welch's t test.
contained LDs. Further extensions of the assay can be envisaged, for example, by employing exogenous lipid loading to induce LDs with a specific content prior to lipid mobilization. In the future, the utility of the platform can also be further extended by the inclusion of additional fluorescence-based readouts amenable to high-content imaging and quantification.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

ACKNOWLEDGMENTS
We thank Anna Uro for technical assistance, HiLIFE-and Biocenter-Finlandsupported Helsinki BioImaging infrastructures for help with microscopy, Katariina Ö ö rni for help with LDL preparation, and Abel Szkalisity for help with image analysis. We thank THL Biobank for providing samples and data for this study (study nos. 2016_15, 2016_117, and 2018_15)  A.S.d.F., G.F., J.K., and M.L. provided patient samples and clinical data. L.P. and P.H. established image analysis and processing tools. S.G.P. and E.I. wrote the manuscript. All authors reviewed and revised the manuscript.

DECLARATION OF INTERESTS
A patent application covering the use of the here-suggested patient stratification methods has been filed (application: FI 20206284), in which University of Helsinki is the applicant and E.I. and S.G.P. are the inventors.

INCLUSION AND DIVERSITY
We worked to ensure gender balance in the recruitment of human subjects. The author list of this paper includes contributors from the location where the research was conducted who participated in the data collection, design, analysis, and/or interpretation of the work. Tada, H., Kawashiri, M., Noguchi, T., Mori, M., Tsuchida, M., Takata, M., Nohara, A., Inazu, A., Kobayashi, J., Yachie, A., et al. (2009). A novel method for determining functional LDL receptor activity in familial hypercholesterolemia: application of the CD3/CD28 assay in lymphocytes. Clinica Chim. Acta 400, 42-47.