Lipidomic comparison of 2D and 3D colon cancer cell culture models

Abstract Altered lipid metabolism is one of the hallmarks of cancer. Cellular proliferation and de novo synthesis of lipids are related to cancer progression. In this study, we evaluated the lipidomic profile of two‐dimensional (2D) monolayer and multicellular tumor spheroids from the HCT 116 colon carcinoma cell line. We utilized serial trypsinization on the spheroid samples to generate three cellular populations representing the proliferative, quiescent, and necrotic regions of the spheroid. This analysis enabled a comprehensive identification and quantification of lipids produced in each of the spheroid layer and 2D cultures. We show that lipid subclasses associated with lipid droplets form in oxygen‐restricted and acidic regions of spheroids and are produced at higher levels than in 2D cultures. Additionally, sphingolipid production, which is implicated in cell death and survival pathways, is higher in spheroids relative to 2D cells. Finally, we show that increased numbers of lipids composed of polyunsaturated fatty acids (PUFAs) are produced in the quiescent and necrotic regions of the spheroid. The lipidomic signature for each region and cell culture type highlights the importance of understanding the spatial aspects of cancer biology. These results provide additional lipid biomarkers in colon cancer cells that can be further studied to target pivotal lipid production pathways.


| INTRODUCTION
Cancer is a major global public health problem and is the second leading cause of death in the United States. 1 Colorectal cancer is projected to be the third most diagnosed cancer, and the third leading cause of cancer-related death in the United States for 2022. 2  Three-dimensional (3D) cell cultures enable substantial improvements in mimicking the tumor microenvironment and recapitulating other critical aspects of tumor biology better than traditional monolayer (2D) culturing platforms. Although 2D cell cultures have minimal cell-cell contacts and a homogenous phenotype, the multicellular tumor spheroid (MCTS) is an example of a 3D cell culture that provides a platform for examining cellular function and metabolism in cancer. 3,4 Nutrient and oxygen concentration gradients are formed in spheroids, much like tumors, and consist of proliferating cells in the outer layers and hypoxic, nutrient-deprived necrotic cells in the inner regions. 5,6 In between these vastly different cell populations, viable quiescent cells are also present. 6 Studies on the altered lipid metabolism of colorectal cancers show it to be a promising and targetable vulnerability. 7  in which they found altered levels of sphingolipids, such as sphingomyelins (SM) and ceramides (Cer). 10 Several SM species were significantly up-regulated in 5-FU-resistant samples, whereas Cer species were significantly down-regulated. Sphingolipids are bioactive molecules that have critical roles in regulating cancer cells. SMs have prosurvival function whereas Cers can mediate cell death. 11 The use of copper oxide nanoparticles has also been investigated to induce toxicity in HCT 116 cells, where a dose-dependent increase of certain ceramide species was observed, as well as the upregulation of triacylglycerols, and phosphatidylcholines. 12 Co-culturing cells have increasingly been recognized as a technique to understand cell-to-cell communication and to better model the tumor microenvironment. 13 Gong 17 Using three independent patient cohorts, they have identified lipidomic changes in tumor samples compared with normal tissue. Despite a high degree of variation among the cohorts, which the authors suggested was due to patients' diet, prescribed drugs and other confounding metabolic disorders, significant differences in glycerolipids, glycerophospholipids, and sphingolipids were identified.
They found certain sphingomyelin (SM) and triacylglycerol (TG) species to be elevated and proposed a lipid signature based on TG levels which differentiated them from nontumor tissue. This study provided a comprehensive analysis of CRC patient samples and suggested putative markers to diagnose patients better and to monitor potential drug therapies.
One of the critical aspects of using spheroids as a tumor model is that their cellular composition is similar to an in vivo tumor. As spheroids grow, radially symmetric chemical gradients develop. Many of the current approaches in generating tumor spheroids can be produced at a large scale, generate uniform structures, and be used for cancerrelated studies with live-cell imaging, 18 immunohistochemistry, Western blotting, mass spectrometry-based proteomics, 19,20 and metabolomics. [21][22][23] Recent drug toxicity studies of cancer therapeutics were conducted using mass spectrometry imaging (MSI) of spheroids, which provides the spatial distribution of not only the drug therapeutic but also drug metabolites. 24,25 This label-free imaging technology can also be used to profile other biochemical changes, such as small molecule metabolites involved in the Krebs cycle and lipids. [26][27][28] Within colorectal cancer research, MSI has been previously used to characterize lipid profiles of tumor tissues and the adjacent environment. 29 To mitigate the loss of spatial information, serial trypsinization can be subsequently utilized after harvesting spheroids. Using a dilute trypsin solution, cells can be sequentially peeled off from the spheroid through multiple cycles and washes. McMahon et al. previously obtained discrete cell populations from HT-29 spheroids representing different regions of the 3D culture by subjecting them to serial trypsin treatments, quantifying proteins in HT-29 spheroids involved in several cellular metabolism pathways. 30 Keithley et al. also utilized serial trypsinization to probe the sphingolipid metabolic profile at different regions of HCT 116 spheroids. They were able to detect fluorescent probes that were specifically labeled for the outer, middle, and core regions after serial trypsinization in a reproducible manner and showed cells from different regions exhibited differences in metabolism. 31 In this work, we conducted comprehensive lipidomics profiling of the HCT 116 colon carcinoma cell line as 3D MCTS and 2D monolayers. For each biological replicate, we conducted serial trypsinization to further elucidate the lipidomic changes in different regions of MCTS before lipid extraction. Liquid chromatography coupled to highresolution quadrupole time-of-flight mass spectrometry was utilized to identify and quantify lipids. We found several classes of lipids that are highly altered in spheroids compared with 2D monolayer cultures.
The lipidomic profile of the other regions shows similar features as 2D monolayers, whereas the middle (quiescent) and core (necrotic) regions diverge from this similarity and show unique lipidomic features that suggest altered lipid metabolism and different energy storage. Triacylglycerols and certain classes of sphingolipids were elevated in the hypoxic and necrotic regions of HCT 116 spheroids, suggesting the 3D microenvironment of spheroids affects the lipid spatial distribution. Further, lipids consisting of fatty acyls with longer carbon-chain and a higher degree of fatty acyl unsaturation are observed in spheroids at higher levels relative to 2D monolayers.
Finally, we discuss our lipidomic results in the context of recent studies on the effects hypoxic and low-pH regions of the tumor microenvironment.

| Spheroid culturing
The colon carcinoma cell line HCT 116 was purchased from the Amer- Spheroids were cultured in agarose-coated 96-well plates as previously described. 32 to make a gelatin array. As discussed by Johnson et al., the molds provide better alignment of the spheroids along the z-axis of the gelatin mold so multiple spheroids can be cryosectioned together. 34 Briefly, gelatin powder was dissolved in water and warmed in a pressure cooker that is set to "warm" mode. To form the gelatin base using the mold, 550 μl of warm gelatin was carefully pipetted into the mold and allow it to harden. Using a wide-bore pipette, spheroids were transferred with minimal 1Â PBS. Residual PBS was removed around the spheroids to avoid deterioration during the cryosectioning process.
Small drops of warm gelatin were placed onto each individual spheroid to embed them in place. The mold was then placed in a À80 C freezer for 10 min to allow it to harden. The spheroids were capped by carefully pipetting 190 μl of warm gelatin, and the sample was stored in a À80 C freezer until the cryosectioning step.
Cryosectioning was performed on a Leica CM 1950 cryostat (Leica Biosystems, Buffalo Grove, Illinois, USA). Spheroids were sectioned into 12 μm slices and thaw-mounted onto glass slides. Glass slides were then stored in a À80 C freezer until the staining procedure was conducted.

| Data analysis and statistics
Agilent (.d) data files were converted to ABF format using ABF Converter. MS-DIAL version (4.48) was used for peak peaking, alignment, and identification. 37,38 The analysis parameter settings for MS-DIAL are available in Table S2. Further, EQUISPLASH lipidomics internal standards were used to normalize and quantify the endogenous lipids To further assess the reliability of the in silico database matches of the identified lipids, we compared the retention time behavior of each lipid class relative with their double-bond number and total number of carbons in their fatty acyl chains. All retention time behavior plots are included in the supporting information. Furthermore, lipid nomenclature used in this study follows the rules from Liebisch et al. 39 Lipids with a known fatty acyl composition based on its tandem-MS data in negative ion mode were annotated at the "molecular species level." Lipids that have uncertain fatty acyl composition were annotated at the "species level" that contains the total carbon and total double bond (or double bond equivalents) of the fatty acyl residues. Additionally, manual filtering between polarities were conducted to minimize multiple annotations of the same lipid. In silico annotation of the tandem-MS of cholesterol was conducted using MS-FINDER version (3.52). 40 R-based packages lipidR and SCOPE were utilized to obtain bioinformatics. 41,42 Lipids that were identified as having the same lipid carbon : double bond assignment were not filtered and annotated as either an "_A" or "(1)" for SCOPE and lipidR, respectively.  Figure 1B).  Based on the normalized total lipid content, the percent composition of each lipid subclass was determined between 2D monolayer and spheroid regions. As listed in Table 1  . Statistical significance was determined using ordinary one-way ANOVA with Bonferroni's multiple comparisons test against 2D culture samples. Adjusted P-value indicators are *P < 0.05, **P < 0.005, ***P < 0.0005, ****P < 0.0001 (Figure 4).

F I G U R E 4
Legend on next page.

| Glycerolipids
To provide a comprehensive spatial view of the lipid alteration within spheroids, we utilized SCOPE by Odenkirk et al. to generate a circular dendrogram that clusters lipids by their SMILES data and provides a heatmap of log 2 fold change ratios relative to a control. 41 This enabled a spatial comparison of each spheroid layer to further visualize the lipidomic changes at both a subclass-wide and even on a lipid speciesspecific level. Using the 2D monolayer data for our baseline comparison, Figure 2 illustrates all the lipid abundance alterations in the three spheroid regions. Based on this circular dendrogram, there are entire lipid subclasses that appear to be altered in specific spatial regions of the spheroids. For example, the triacylglycerols (TG) is an example of a subclass that are elevated in spheroids. TG are a subclass of neutral lipids that are produced by cells as either a precursor for lipid membrane synthesis, produced to sequester free fatty acids, or as a storage for energy. TGs have also been observed to accumulate in cells exposed to hypoxia due to the induction of lipin 1, as a catalyst in the triglycerol biosynthesis, and by hypoxia-inducible transcription factors (HIFs). 43 Quantitatively, comparing the normalized molar abundance of the TG subclass further reveals the difference between 2D monolayer cultures and the spheroid regions. As illustrated in Figure 3C, there is significantly more TG content in all spheroid regions than in 2D monolayers. At the same time, diacyglycerol (DG) molar abundance remains largely remains unchanged, except for in the outer region where it is slightly elevated. To investigate the lipid species-specific alteration for each spheroid layer, Figure 3A illustrates a heatmap as a Phosphatidylethanolamine (PE) is the second most abundant glycerophospholipid, which constitutes 20%-50% of total phospholipids. 45 We quantified PE lipids to be third most abundant in HCT 116 2D monolayers and 3D spheroids as shown in Table 1 branes. 44 A total of 18 PI lipids were detected and quantified ( Figure S4). As seen in Figure 3D, there were no significant changes in F I G U R E 4 Triacylglycerol (TG) abundance profile in HCT 116 spheroids and retention time behavior of TGs in reversed-phase chromatography. (A) Relative to monolayer cultures, higher abundance of TG lipids are present in 3D cultures that consists of higher degree of double bond unsaturation. This distinction is more amplified in the middle and core layers than in the outer layers. The numerical values inside each pixel correspond to the fold change ratio (spheroid layer/2D monolayer) of that lipid species. (B) Retention time behavior of all TG species detected and identified by MS/MS in the study; 0 dB through 7 dB represent the cumulative number of double bonds in the fatty acyl chains, whereas the total carbon represents the cumulative number of carbons in the fatty acyl chains.
the amount of PI lipids between the spheroid regions and 2D monolayer.  Figure 3D). The major PS species detected with highest molar abundance is PS(18:0_18:1), which interestingly, had an increasing amount from the outer to the core regions (supporting information).

| Sphingolipids
Sphingolipids are a diverse group of lipids comprising of ceramides Within the spheroids, higher levels of the identified Cer, HexCer, SHexCer, and SM lipids were found in the middle and core regions, which represents the quiescent and necrotic regions of the spheroid.
Quantitatively, the total normalized abundance of each subclass is illustrated in Figure 3B.

| Cholesterol, acylcarnitines, and free fatty acids
Free cholesterol was determined to not be significantly altered across spheroid regions and 2D monolayers ( Figure S12B). Acylcarnitines are esters of l-carnitine and fatty acids and are involved in the β-oxidation of fatty acids in the mitochondria. Long-chain acyl-CoA synthetase (LACS) facilitates the transformation of fatty acids into acyl-CoAs, which are then further transformed into acylcarnitines by carnitine palmitoyl-transferase 1 and 2 (CPT1 and CPT2). Finally, acylcarnitines are transported across the mitochondrial membrane by carnitine/acylcarnitine translocase (CACT). 46 In the current study, eight acylcarnitine species were detected and quantified across the spheroid layers and 2D cultures. As shown in Figure S7, all acylcarnitine species have relatively elevated levels across the spheroid layers as compared with 2D monolayers, although the core region has a more significant amount of CAR compared to 2D cultures ( Figure 3A). CAR (22:0) is observed to have the greatest fold change among the acylcarnitine species, with a gradual increase from the outer layer to the core. The retention time behavior for the acylcarnitines is shown in Figure S9A.
Free fatty acids were also detected and quantified in this study, with seven species being detected and quantified across samples.
Fatty acids comprised the majority of the lipids detected and quantified in all sample types, as shown in Table 1 S9B). Overall, the FA subspecies remains unchanged across 2D monolayers and spheroid regions ( Figure 3A).

| Chain length comparison across spheroid layers relative to 2D cultures
We next evaluated the total lipid chain length of the fatty acyl moieties identified across samples. As shown in Figure 5, shorter chain lengths (less than 34 carbons) are elevated at the same levels across the spheroid regions relative to 2D cultures. At the same time, higher production of longer chain lengths appears to be more prominent in the spheroid culture, where the most elevated production of longer chain lengths appears to be in middle and core samples. This suggests different fatty acid metabolism pathways are being utilized between 2D and 3D cultures.
F I G U R E 5 Global lipid chain length analysis as a function of log 2 fold change ratio between spheroid layer to monolayer culture. Longer chain lengths (total carbon number of 34) are higher in abundance at middle and core layers relative to (2D) monolayer cultures.

| DISCUSSION
The present lipidomics study aims to characterize the lipidomic profile of 3D and 2D cell cultures of the HCT 116 colorectal cancer cell line.
3D cell culture platforms such as spheroids provide a valuable tool for studying the tumor microenvironment in cancer research as they mimic many of the characteristics that tumors exhibit in the body, such as pH and oxygen gradients. It is widely accepted that cancer cells undergo metabolic changes to support their needs to proliferate.
Further, it has been increasingly recognized that lipid metabolism is altered in cancers, which includes increased fatty acid production and uptake from the tumor microenvironment. 47 As discussed in reviews, cancer cells can acquire fatty acids by either (1)  sources of FAs that could be shuffled to fulfill cellular needs. A less discussed property of LDs is that they can prevent lipotoxicity by providing cells a buffering capacity to store lipids. They achieve this capacity by sequestering free FAs as TGs to prevent them from incorporating with cytotoxic lipids at high levels such as ceramides or acylcarnitines. 52 Interestingly, this phenomenon has been observed in several metabolic diseases such as obesity, non-alcoholic fatty liver disease, and cardiovascular disease. 52 Indeed, the total TG amount present in the spheroid regions are much more elevated than in 2D cultures ( Figure 3A). Furthermore, using Oil Red O staining, we observed an accumulation of LDs in the inner regions of HCT 116 spheroids, which correspond to the increased TG accumulation in spheroids relative to the 2D cultures ( Figure S11B). Our findings correlated with previous studies using 3D cell culture to study the lipidomic rewiring in several cancers. 23,53,54 However, using serial trypsinization, we were able to further decipher specific lipidomic alterations within the spheroid. As our LC-MS-based lipidomics profiling shows, higher TG levels are present in the middle and core regions of spheroids, where a large portion of the TG species contain PUFAs.
Interestingly, it has been suggested that PUFAs are preferentially sequestered in LDs as they can be vulnerable to lipid peroxidation, leading to ferroptosis. At the same time, it was recently reported that extracellular low pH conditions (acidosis) can lead to the activation of TGF-β2 signaling, thus inducing epithelial-mesenchymal transition (EMT) and PKC-zeta-mediated translocation of CD36, which facilitates FA uptake that are stored in LDs as TGs. 53 Fatty acid availability can come either from exogenous sources or from de novo synthesis by fatty acid synthase (FASN). 47,55 Aberrant FASN expression is implicated in several cancers, which is known to induce de novo lipogenesis and is involved in cell proliferation, survival, and invasion. 56 As a result, the ability to synthesize FAs can make cancer cells independent of the limited environmental availability of precursor substrates. Indeed, there has been considerable interest in the development of FASN inhibitors, which include the nextgeneration therapeutics such as TVB-2640 and yields high anti-tumor potential and limited systematic toxicity in clinical trials. These promising results from early-phase trials were an improvement from the first-generation therapeutics such as C75, orlistat, and cerulenin; although showing great promise as anti-tumor agents, side effects were present as characterized by severe weight loss. 49 Cancer cells can activate multiple pathways to obtain its metabolic needs. In addition to exogenous FA uptake, a FADS2-dependent route of producing PUFAs is also activated. Alternatively, SCD-1, an enzyme involved in converting saturated fatty acids to monounsaturated fatty acids (MUFA) with double bonds at the Δ9 position, has been shown to be overexpressed in cancer. 47 However, SCD-1 requires O 2 to be active, and thus, O 2 deprivation can inhibit the activity of the enzyme. 57 In addition, as tumors grow, their need for unsaturated lipids shifts from endogenous production to the uptake from the microenvironment. FADS2 is a key enzyme in the production of polyunsaturated fatty acids and is more highly expressed in several cancers, including colorectal cancers. 58 In addition to fatty acid dysregulation, the accumulation of lipid droplets in tumors and 3D cell culture was observed, which provides storage for TGs and cholesteryl esters. 23 In the current study, we observed a class-wide increase of TGs in all layers of the spheroid.
However, we were unable to detect intact cholesteryl esters in this study as the cholesterol moeity appeared to have been cleaved off, perhaps during the electrospray process. This potential cleavage was gleaned from the data when we generated the extracted ion chro- Ceramide is the central lipid of all sphingolipids as it can be further metabolized or transformed by various enzymes. 61 The accumulation/synthesis of ceramide in response to cellular stress is welldocumented and is known to induce apoptosis, necroptosis, and endoplasmic reticulum (ER) stress. 11,62,63 Interestingly, ceramide accumulation is observed to occur as a secondary effect after radiation treatment and chemotherapy treatment. In other diseases, ceramide and other sphingolipids have been associated with neuronal cell death, which was previously observed in a mouse model of Niemann-Pick disease, type C (NPC). 64 It is worth noting that individual ceramide species generated exert different functions in the cell. For example, long chain ceramides containing fatty acyls 16:0, 18:0, and 20:0 are known to be antiproliferative, whereas ceramides containing very long fatty acyls 24:0 or 24:1 promote cellular proliferation. 65 These ceramides are associated to specific ceramide synthases (CERS), which have propensities toward particular acyl-CoA to bond with sphingosine. 63,66 From our current data, it seems to indicate higher levels of sphingolipids in all regions of the spheroid, with the middle region having the highest abundance of all four subclasses ( Figure 3B). This could indicate higher levels of ceramide-mediated cell death is occurring, but further studies are warranted to further investigate these alterations.
Higher production of acylcarnitines were observed in all three spheroid layers. Acylcarnitines are intermediate forms of free FAs that are destined for the mitochondria, where it is further metabolized to acyl-CoA and carnitine by CPT2. Furthermore, the acyl-CoA is additionally involved in a multistep process known as β-oxidation, which ultimately feeds into the Krebs cycle. 47

| CONCLUSION
In conclusion, this lipidomic profiling revealed several lipid class-wide alterations and lipid-species specific changes in HCT 116 spheroids.