Longitudinal Proteomic Changes in HCT 116 Colon Cancer Spheroids During Growth

Abstract

The FDA Modernization Act 2.0 permits data from advanced microphysiological systems, such as spheroids, to be used as a testbed for drug candidates entering phase 1 clinical trials. Despite their increasing adoption, spheroids of varying growth durations are often used interchangeably as disease models. While transcriptomic studies have been employed to monitor spheroids over time, proteomics has primarily been used to validate their utility as model systems and assess drug responses rather than for longitudinal studies. Here, we apply data independent acquisition with gas phase fractionation (DIA-GPF) proteomics to investigate temporal changes in HCT 116 spheroids every 2 days throughout 18 days of growth, identifying 6,835 proteins across all samples. Differential expression analysis reveals that day 2 spheroids more closely resemble monolayer cells than spheroids cultured for extended periods. Gene ontology (GO) term analysis of differentially expressed proteins indicates that relative to monolayer cells DNA replication is downregulated, while glycolysis is upregulated during spheroid maturation. Parallel reaction monitoring (PRM) experiments targeting thymidylate synthase and fructose-bisphosphate aldolase C validate the initial proteomic findings and corroborate the trends observed in the GO term analysis. These results highlight the importance of growth duration when spheroids are used as a model for avascular tumors.

Graphical Abstract

INTRODUCTION

A major obstacle in discovering new therapeutics is the high failure rate of drugs in clinical trials. One cause of this low success rate is inadequate methods to assess the efficacy of candidate therapeutics.¹ Historically, drugs have been evaluated using two-dimensional (2D) monolayer cell cultures, but the resulting data is often unable to be reproduced in more complex biological systems.^2-5 One reason for this discrepancy is the failure of the 2D model to recapitulate in vivo cellular microenvironments.^1,6 Before 2023, all drugs that entered human clinical trials in the United States needed to be evaluated preclinically in animal models. However, the passing of the FDA Modernization Act 2.0 updated this requirement, allowing advanced microphysiological systems to be used for nonclinical drug screening prior to clinical trials.⁷ One such alternative system is three-dimensional (3D) cell cultures, which offer several benefits over 2D monolayer cultures, including multidimensional cellular contacts, as well as chemical and nutritional gradients within the model. When compared with animal models, 3D cell cultures pose a significantly lower financial and ethical burden.

Spheroids are one type of 3D cell culture that can be generated from immortalized cell lines to generate a mass of cells more closely recapitulating a tumor’s cellular microenvironment.^8,9 The exact conditions for spheroid growth and the time of harvest can vary for different cell types.^10,11 Spheroids can be formed with or without molecular scaffolds that bring cells close together to ensure the formation of cell-to-cell contacts.¹² Compared to monolayer cell culture, spheroids also have both protein and gene expression profiles that are more similar to in vivo tumors than monolayer cells alone, which can be enhanced by the addition of multiple cell types.^4,5,13,

Proteomics has often been used to characterize cancer spheroids for method validation, drug response, and other applications; however, it is less frequently employed to evaluate the longitudinal changes in this model system.^14-17 Pomeshchik et al. compared the proteomes of multiple model systems used for studying Alzheimer’s disease, including induced pluripotent stem cell-derived spheroids after 100 days of in vitro culture, post-mortem patients’ brains, and animal grafts derived from dissociated spheroids after 50 days of in vitro growth. Their proteomics analysis concluded that the grafted animal model most closely resembled the early stages of Alzheimer’s disease. In this case, proteomics was used to compare disease models rather than to monitor them over time.¹⁸

For longitudinal studies of spheroids—tracking how the model system changes over time—transcriptomics is more commonly utilized, particularly in studies of neurological disorders. Nickel et al. integrated transcriptomic, epigenetic, and drug response data from spheroids generated from multiple glioblastoma lines to make predictions about drug efficacy.¹⁹ Another longitudinal transcriptomic study of human cortical spheroids identified genes associated with axonal processes as early risk factors for schizophrenia.²⁰ Additionally, transcriptomic analyses of multiple cell types in cortical spheroids modeling Down syndrome revealed cell-specific changes and the development of excitatory neurons over time.²¹ These and many other studies have demonstrated the value of studying spheroids as they age, yet proteomics remains underutilized for this purpose.

The goal associated with the utilization of spheroids in cancer research is to mimic the tumor microenvironment as closely as possible while maximizing the throughput. Initially, spheroids were grown using the liquid overlay technique by the Sutherland lab, who grew spheroids approximately 2 weeks, until growth plateaus were evident.^10,22 At this time point, three general cellular populations were observed in the spheroids, including a necrotic core, a quiescent middle layer, and an outer proliferative layer.¹⁰ The presence of these three cellular populations modeled intratumor heterogeneity. These cellular populations are also a reflection of radially symmetric chemical gradients observed in avascularized tumors, including oxygen, glucose, ATP, and lactate (Figure 1A).²³ The Sutherland lab used a microelectrode to determine the amount of oxygen present throughout the spheroid, finding that a majority of the spheroid is hypoxic.²⁴ Hematoxylin and Eosin staining confirms the presence of cellular heterogeneity, with secondary necrosis at the center of the spheroid, with a layer of senescent cells present just inside the outermost layer of cells.²⁴ We hypothesized that these changes could be visualized using two stains for viability and one stain for apoptosis. NucBlue Live ReadyProbe Reagent (Hoechst 33342) is cell-permeable and binds DNA to all cells. NucRed Dead 647 ReadyProbes Reagent is cell-impermeable and only binds to DNA in cells with a compromised membrane. CellEvent Caspase-3/7 Green ReadyProbes Reagent is a cell-permeable nucleic acid binding dye conjugated to a peptide sequence. If Caspase 3/7 is active in cells, it will cleave the short peptide sequence and free the dye to yield a green fluorescent signal. In these images, blue staining is assumed to correspond to live cells, red staining is for cells with a permeable membrane (most likely dead), and green staining corresponds to apoptotic cells (Figure 1B). Notably, between days 7 and 14, the amount of apoptotic staining increases, consistent with previous flow data.²⁵ This fluorescent data highlight the increase in cellular heterogeneity as the spheroids grow. Spheroids are a better tumor mimic compared to two-dimensional monolayer cell culture because they develop nutritional and chemical gradients over time and can be used to test cellular response to drug treatment.^{2,13,14,24,26,27}

Figure 1.

Characteristics of spheroids reprinted from Hirschhaeuser et al. (A).²³ Whole spheroid fluorescent imaging for spheroids that have been grown for 7 and 14 (B) days. All cells are stained blue, cells with a permeable cell membrane are stained red, and cells in the early stages of apoptosis are stained green.

In our research group, we have followed the Sutherland lab protocol and grew our spheroids for 14 days before experiments are conducted. Specifically, we have used the liquid overlay technique to generate spheroids to mimic avascularized colon tumors with HCT 116 cells.^14,25-32 Despite their rising popularity, there is no uniform time for spheroids to be harvested. In the literature, the time of HCT 116 spheroid harvest (generated by the liquid overlay technique) varies greatly, from 2 days to 2 weeks, with most studies being done on spheroids grown for a few days.^33-37 Because HCT 116 spheroids grow radially outward with newly divided cells on the outer edges of the cellular mass, it is feasible that the age of a spheroid can impact the results of a study. Previous experiments using flow cytometry indicate that the amount of apoptotic cells increase as spheroids grow.²⁵ Another study found that HCT 116 spheroids at day 6 exhibited a higher survival rate than both day 3 spheroids and monolayer cells when exposed to several commonly prescribed chemotherapeutics, including fluorouracil (5-FU), irinotecan, oxaliplatin, and melphalan.³³ The impact of growth time on HCT 116 spheroids as a model system has not been systematically studied despite a wide range of growth periods displayed in the literature.

In the current study, we examined the longitudinal proteomic changes of HCT 116 spheroids formed by liquid overlaying over 18 days of growth. We determined that there are changes in the proteome of HCT 116 spheroids as they grow using data-independent acquisition with gas phase fractionation (DIA-GPF). Principal component analysis (PCA) and hierarchical clustering indicate that spheroids grown for only a few days are more similar to monolayer cells than older spheroids. Whole spheroid fluorescent imaging complimented the observed proteomic differences by displaying an increase in apoptosis between 1 and 2 weeks of growth. Gene ontology (GO) analysis of differentially expressed proteins compared to monolayer indicates that middle aged spheroids and older spheroids are more dependent on glycolysis for energy. Older spheroids have DNA replication and cell cycle genes downregulated. To validate our initial exploratory proteomics findings, we conducted parallel reaction monitoring (PRM) on proteins that our initial experiment suggested had changed in abundance throughout growth. Thymidylate synthase, fructose-bisphosphate aldolase C, and ferritin light chain followed the trends observed in the original DIA experiment and supported the biological changes indicated by the GO term analysis. Taken together, the growth time of HCT 116 spheroids has implications for their use as a model system in cancer research.

EXPERIMENTAL SECTION

Chemicals

McCoy’s 5A media was obtained from Life Technologies. Fetal bovine serum (FBS), Bicinchoninic acid (BCA) protein assay, DAPI to Hoechst 33242, Caspase-3/7 Cell Event ReadyProbes, and NucRed Dead ReadyProbes Reagent were obtained from Thermo Scientific. Penicillin-streptomycin was obtained from Invitrogen. LC-MS grade solvents, including water (H₂O), acetonitrile (ACN), H₂O:formic acid (CH₂O₂) (100:0.01 v/v), and ACN:CH₂O₂ (100:0.1 v/v) were obtained from Burdick & Jackson. Sodium dodecyl sulfate (SDS), sodium fluoride, sodium orthovanadate, dithiothreitol (DTT), and iodoacetamide were obtained from Sigma-Aldrich. Ethylenediaminetetraacetic acid (EDTA) free protease cocktail inhibitor was obtained from Roche Diagnostics. Suspension traps (S-traps) were obtained from ProtiFi. Trypsin was obtained from Promega.

Cell Culture and Spheroid Formation

HCT 116 cells, a human colon carcinoma cell line, were purchased from American Type Cell Culture (ATCC) and cultured in McCoy’s 5A medium, supplemented with 10% fetal bovine serum, 1% l-glutamine, and 1% penicillin-streptomycin. Upon resuscitation, cells were cultured in a 2D monolayer until 80% confluency with the supplemented McCoy’s 5A medium and underwent a minimum of 2 passages before spheroid formation. The cells tested negative for the mycoplasma. The liquid overlay technique was used for spheroid formation. The wells in the 96-well plates were coated with 1.5% agarose in unsupplemented McCoy’s 5A media, and 7,000 cells per 200 μL aliquots were seeded in each well. After seeding, the 96 well plates were centrifuged at 1,000 × g for 10 min to ensure spheroid formation after 2 days of growth. Spheroids were incubated at 37 °C and 5% CO₂ for the duration of growth. Starting on day 4, 50% volume media changes occurred every 48 h. Three biological replicates of approximately 30 spheroids were harvested every 2 days, starting on day 2 and ending on day 18. Upon harvest, cells were collected in a Petri dish of warmed phosphate-buffered saline (PBS). Spheroids were washed in PBS to remove excess medium and centrifuged at 4,000 × g. The supernatant was removed, and the spheroids were resuspended in HPLC-grade H₂O. After two washes, the supernatant was completely removed and samples were stored at −80 °C until all samples were collected and ready for lysis.

Fluorescent Imaging of Whole Spheroids

A working solution of dye was prepared by adding 2 drops of DAPI to Hoechst 33242, Caspase-3/7 Cell Event ReadyProbes, and NucRed Dead ReadyProbes Reagent per milliliter of supplemented media. Spheroids on days 7 and 14 were subjected to a half-media change approximately 24–28 h before whole spheroid imaging or embedding. 100 μL of the dye working solution was added to the spheroids, which were then returned to the incubator for continued growth overnight. Spheroids were harvested, washed three times with warmed PBS, and placed in a white-walled, clear-bottom 96-well plate. Whole spheroids were imaged at 10× magnification on a Molecular Devices ImageExpress Pico, using the DAPI, FITC, and Cy5 filters with exposure times of 5, 40, and 600 ms, respectively. The image intensity scale was 4–507 for each dye. The presented images are a series of 9 z-stacked images, centered around the plane of best fluorescent focus in 10 μm steps (Figure 1B).

Proteomic Sample Preparation

Protein from harvested spheroids was extracted via lysis in buffer (6% SDS in 50 mM Tris-HCl, pH 8) containing 1 mM sodium fluoride, 1 mM sodium orthovanadate, 1 mM glycerophosphate, and one EDTA-free protease inhibitor tablet. Each sample was treated with 250 μL of lysis buffer and sonicated twice for a total of 1 min at 15% power (3 s on, 2 s off). Clarification was achieved by centrifugation at 15,000 rpm for 10 min at room temperature. Protein concentration was determined by using a BCA assay. A total of 100 μg of protein per sample was lyophilized and resuspended in 50 μL of lysis buffer. Disulfide bonds were reduced with 20 mM DTT at 65 °C for 10 min, followed by centrifugation at 1,250 rpm. Alkylation was performed using an excess of iodoacetamide (~40 mM) at room temperature in the dark for 30 min, followed by centrifugation at 13,000 g for 8 min. Samples were acidified to a final concentration of 1.2% (v/v) phosphoric acid, ensuring a final pH of ≤ 1. S-trap mini columns were used for detergent removal and protein digestion. Binding buffer (100 mM triethylammonium bicarbonate (TEAB), 90% methanol) was added at a 1:6 sample-to-buffer ratio. Samples were loaded onto S-trap mini columns, centrifuged at 4,000 × g for 30 s, and reloaded to ensure complete protein retention. The columns were washed three times with 400 μL of binding buffer, followed by a final spin to dry. Digestion was performed by adding 125 μL of 50 mM TEAB containing trypsin at a enzyme:protein ratio of 1:50 enzyme-to-protein ratio (w/w). Samples were incubated at 37 °C overnight. Peptides were sequentially eluted using 80 μL of 50 mM TEAB, H₂O:CH₂O₂ (100:0.02 v/v), and ACN:H₂O:CH₂O₂ (50:50:0.2 v/v/v). Peptides were lyophilized and resuspended in 1 mL of H₂O:CH₂O₂ (100:0.01 v/v).

Proteomic LC-MS

Samples were desalted using Waters HLB extraction cartridges (10 mg of sorbent, 30 μm particle size, 80 Å pore size). Cartridges were activated with 2 mL of 80% ACN:-H₂O:CH₂O₂ (80:20:0.01), then equilibrated with 2 mL of H₂O:CH₂O₂ (100:0.01 v/v). After sample loading, cartridges were washed with 2 mL of H₂O:CH₂O₂ (100:0.01 v/v) and eluted with 1 mL of ACN:H₂O:CH₂O₂ (80:20:0.01). Desalted peptides were lyophilized and resuspended in 200 μL of H₂O:CH₂O₂ (100:0.01 v/v). Sample concentrations were determined by using a Nanodrop. A pooled sample was generated by combining equal masses from each sample for GPF library generation.³⁸ All samples, including the pooled sample, were lyophilized and resuspended to a final concentration of 200 ng/μL in H₂O:CH₂O₂ (100:0.01 v/v) prior to liquid chromatography-mass spectrometry (LC-MS) analysis. A total of 2 μL from each sample was injected into a Waters NanoAcquity LC system equipped with a nanoEase M/Z Peptide BEH C18 column (75 μm inner diameter × 200 mm length) maintained at 50 °C with a flow rate of 400 nL/min. Peptides were eluted using the following gradient increasing mobile phase B (ACN:CH₂O₂ [100:0.01 v/v]) from 2% to 28% over 90 min, from 28% to 85% over 2 min, held at 85% for 10 min, decreased from 85% to 2% over 1 min and held at 2% for 40 min. Mobile phase A was H₂O:CH₂O₂ (100:0.01 v/ v). Proteomic data were acquired on a Thermo QE-HF mass spectrometer in DIA mode, using staggered 16 m/z-wide windows from 400–1000 m/z. A GPF library was generated using the pooled sample with 4 m/z-wide staggered windows across 100 m/z from 400–1000 m/z to construct the chromatogram library.

Proteomic Data Analysis

Each injection was evaluated by comparing the MS total ion current (TIC) across the run to ensure a consistent spray. Sample and GPF library files were converted to.mzML format using MSconvert. GPF library files were searched against the Pan-Human database in EncyclopeDIA to generate an empirically corrected library. For all sample analyses, Human FASTA served as the background proteome. In Skyline.mzML files were searched against the generated spectral library, and peptides were quantified using area under the curve. The data was annotated in Skyline before being exported as a.csv file compatible with MStats.^38,39 Samples were normalized using equalized medians. Initial file processing in MStats filtered out peptides that were nonunique, had a q-value greater than 0.01, were present in fewer than three samples, or had a single charge state. Proteins with only one unique peptide were removed.⁴⁰ Postnormalization, the quality of each run was evaluated using box-and-whisker plots (Figure S1).

GO Term Figure

ShinyGO 0.82 was used to generate GO term figures for each time point.⁴¹ Statistically significant up- and down-regulated proteins in spheroids were identified by comparison to monolayer cells at each time point. These up- and down regulated genes were separately analyzed in ShinyGO 0.82. The background for this enrichment was 6.835 proteins identified across all samples. Pathways consistently observed across multiple days were selected for heatmap visualization. The log intensities from the ProteinLevelData from MSstats was collected for genes associated with the selected pathways. Heatmaps displaying protein abundance over time were generated using Heatmapper.ca.⁴² Gene sets for each pathway were obtained from either the KEGG database or the Curated Elsevier Pathway database.⁴³ The scale type was row, the clustering method was the average link, and the distance measurement method was Euclidian.

PRM LC-MS

PRM samples were grown independently of the global DIA proteomics samples. Spheroids were grown and prepared for mass spectrometry analysis as previously described, with the exception that desalting was performed using Pierce Peptide Desalting Spin Columns according to the manufacturer’s protocol. Initial method optimization was completed using a pooled sample with equal mass amounts from the three time points selected for PRM analysis: monolayer cells on days 6, and day 18. Target proteins were chosen based on the relevant biological pathways indicated by GO analysis, the number of unique peptides detected in the DIA experiment, the number of fragments observed, and their presence in SRM Atlas.⁴⁴ Two unique target peptides were detected for each protein. One μL of each sample (500 ng) was injected on to a Thermo Ultimate 3000 equipped with a PepMapTM RSLC C18 column (75 μm inner diameter × 25 cm length) maintained at 40 °C with a flow rate of 300 nL/min. Peptides were eluted using the following gradient, increasing mobile phase B (ACN:CH₂O₂ [100:0.01 v/v]) from 2% to 5% over 5 min, from 5% to 16% over 40 min, from 16% to 25% over 10 min, from 25% to 32% over 10 min, from 32% to 90% over 1 min, held at 90% for 2 min, and decreased from 90% to 2% over 1 min. Mobile phase A was H₂O:CH₂O₂ (100:0.01 v/v). Data were collected on a Thermo Orbitrap Fusion. Our method included an MS1 scan taken throughout the gradient in the Orbitrap (MS OT) from 300 to 2,000 m/z with a resolution of 120,000. Precursor masses of each target peptide were isolated with the quadruple and fragmented with HCD at a collision energy of 32 in the ion trap. Target peptides were detected over a 2- or 4 min retention time window, with an isolation window of 2.5 m/z also in the ion trap.

PRM Data Analysis

Each injection was evaluated by comparing the MS1 TIC across the run to ensure a consistent spray. All data analysis was performed in FreeStyle. The presence of each target peptide was confirmed by visualizing the elution profile of at least three fragment ions for that peptide. More than 8 points across the peak were observed for each target. A quantitative ion was selected for each peptide based on fragment intensity and consistency across samples. Peaks were manually integrated in FreeStyle. Target abundance was normalized by dividing the abundance of a target peptide by the sum of the abundances for both housekeeping gene target peptides.

RESULTS AND DISCUSSION

We hypothesized that significant longitudinal changes would occur in the proteome of spheroids during an 18-day growth period. To evaluate this hypothesis, HCT 116 spheroids were grown by using the liquid overlay technique. Starting at 2 days of growth, three biological replicates of approximately 30 spheroids were harvested every 2 days and stored at −80 °C. A total of 9 plates of spheroids were used in this experiment. After 18 days of growth, the proteomic sample preparation for all samples was completed. All samples were lysed, protein extracted, and digested simultaneously. DIA-GPF bottom-up proteomics was performed on an Thermo QE-HF coupled to a Waters nano-LC using reverse-phase chromatography (Figure 2). The primary goal of this initial experiment was to determine how protein expression changes during spheroid growth. HCT 116 cells grown in a monolayer were also harvested as a control. Upon examination of the TIC for each injection, we noticed a stark loss of signal for sample 2D replicate 2 (Figure S2). Due to this loss of signal, we excluded sample 2D replicate 2 from our analysis. After searching the data in EncyclopeDIA, data formatting was completed in Skyline, and data processing was carried out in R using the MSstats package. After initial filtering and normalization, the box-and-whisker plots showed consistent signal intensities for all injections (Figure S2). Approximately 6,835 unique proteins and 64,650 peptides were identified across all samples.

Figure 2.

Sample workflow overview generated in BioRender. HCT 116 spheroids were harvested every 2 days of growth. Samples were lysed in 6% SDS, reduced, alkylated, and digested with trypsin. Samples were desalted and quantified, and a pooled sample was created. Finally, samples were analyzed on an Orbitrap HF instrument with a Waters nano-LC using a C18 reversed phase column by DIA.

A PCA of all samples showed that biological replicates for each day clustered together (Figure 3A). Strikingly, the samples sorted themselves from oldest to youngest along the first principal component, with day 2 spheroids closest to the monolayer samples and day 18 being the farthest away. Spheroids age days 4, 6, and 8 and the pooled sample separated themselves spatially from the neighboring days. Unlike the other early time points, day 2 spheroids clustered near the monolayer cells. However, starting at day 10, there appeared to be more overlap between each of the time points and the subsequent days. These data provided the initial evidence that the proteome of spheroids early in their growth is more similar to that of monolayer cells than to older spheroids. PCA analysis relies on the reduction of many variables to determine which samples are like one another and thus cluster spatially together. To obtain a fuller picture of sample relation, a heatmap of protein expression using Euclidean clustering was generated (Figure 3B). The initial branch point in this heatmap included the monolayer, day 2, day 4, day 6, and two of the day 8 biological replicates. Interestingly, in this first branch, the next split separated the monolayer and day 2 from the rest of the early time points, which correlates well with the overlap between these two samples in the PCA plot. The second branch contained most of the samples, including two replicates of day 8, the pooled samples, and all other later time points, which, in the PCA plot, occupied space near one another. As shown in the PCA plot, the latest time points considered in this study were the farthest away from the monolayer and day 2 spheroids. The trends observed in both the PCA plot and the hierarchical clustering data suggest that there are differences in the proteome of HCT 116 spheroids as they grow.

Figure 3.

PCA plot (A) and a heatmap of protein expression using Pearson Correlation (B) for each sample and its replicates after initial filtering in MSstats.

To investigate these changes further, the proteome of each time point was compared against the monolayer cells, revealing significantly upregulated and downregulated proteins (an adjusted p < 0.01) for each day of spheroid growth (Figure 4). The p values shown in Figure 4 have been adjusted to correct for multiple hypothesis testing. Interestingly, the number of statistically significant upregulated and down-regulated proteins increases throughout spheroid growth (Figure S3), with most proteins showing no significant changes. To gain a broader understanding of how biological processes are changing throughout growth, GO term figures were generated using the proteins that were statistically significant in the volcano plots using ShinyGO 0.82. From these enrichment plots, several pathways were common across multiple days with an enrichment false discovery rate less than 0.01. For example, starting at day 6, glycolysis, ECM-receptor interaction, fructose and mannose metabolism, and galactose metabolism are upregulated while the cell cycle is down-regulated. Starting from day 10, carbon metabolism is upregulated, while DNA replication and RNA polymerase is downregulated. Some pathways were unique for certain days. For instance, day 2 had cell adhesin molecules upregulated, and day 18 had the ribosome downregulated. These data depict a model system that shifts its metabolic preferences over 18 days of growth.

Figure 4.

Significantly up and down regulated proteins for representative early time points, including day 2 (A), day 4 (B), day 6 (C), day 8 (D), day 10 (E), day 12 (F), day 14 (G), day 16 (H), and day 18 (I) compared to the monolayer cells. A p-value of <0.01 was used and statistically significant upregulated proteins are shown in blue, while down regulated proteins are shown in red.

Of the analyzed biological pathways, we were particularly interested in how the genes associated with DNA replication and glycolysis would change due to HCT 116 spheroids being a colon cancer cell model. HCT 116 cells exhibit microsatellite instability due to defects in mismatch repair pathways, leading to an accumulation of DNA mutations over time. This is an established mechanism contributing to genomic instability in colon cancer.⁴⁵ Glycolysis is closely linked to the onset and progression of colorectal cancer and has long been observed as the preferred metabolic pathway in cancer cells, even under aerobic conditions.^46,47 DNA replication genes show high expression in the monolayer and early spheroid time points but lower expression starting around day 14 (Figure 5A). While this may seem counterintuitive given that cancer cells are known for their rapid proliferation, spheroid growth occurs radially outward, limiting nutrient access to central cells. A previous study found that in HCT 116 spheroids the proportion of viable cells decrease and the amount of apoptotic cells increase throughout growth.²⁵ We reasoned that DNA synthesis would be occurring less frequently in older spheroids, which consist primarily of nonviable cells. Conversely, glycolysis associated genes show low expression in the monolayer cells and day 2 spheroids and higher expression starting at day 10 (Figure 5B). In our model system, where oxygen availability varies among cells, glycolysis appears to become the dominant metabolic pathway as spheroids mature. Notably, the expression of glycolysis-related genes remains similar between monolayer cells and early spheroids (days 2, 4, and 6). Overall, GO analysis indicates biological changes consistent with the characteristics of an avascular tumor as HCT 116 spheroids develop over time.

Figure 5.

A heat map displaying the expression values for genes associated with DNA replication (A) and glycolysis (B). Feature level protein data from MSstats were log2 transformed and then normalized using z-scores in Heatmapper.ca.

The DIA-GFP proteomics data presented here support that the time of spheroid harvest is a critical decision for researchers using this model system. Here we have shown that there are significant longitudinal proteomic differences in HCT 116 spheroids on a global scale. In consideration of the FDA modernization 2.0, it is critical that researchers choose a complex microphysical system and a time scale that best reflects the disease they are trying to model.

To validate the initial exploratory proteomic experiment and the observed biological changes in spheroids, we performed a PRM experiment. Our global data set was used for the selection of proteins for PRM detection. Briefly, proteins were considered for the PRM experiment, if their abundance changed as the spheroids grew. To determine this change, box-and-whisker plots were generated for each protein detected using MSstats. These plots estimate protein abundance based on the normalized intensity of unique peptides for each protein. An example box and whisker plot for fructose-bisphosphate aldolase C (ALDOC) is displayed in Figure S4A. Three target proteins were selected: thymidylate synthase (TYMS), fructose-bisphosphate aldolase C, and ferritin light chain (FTL). When selecting target peptides, we prioritized those with consistent retention times and 3–5 fragments detected in our DIA-GPF data set. Ultimately, two unique peptides observed in our initial data set and present in SRM Atlas were selected as surrogates for target protein abundance.

Fructose-bisphosphate aldolase C and ferritin light chain were observed to increase in abundance during growth, while thymidylate synthase decreased in abundance. The observed trend for each target protein is logical given each protein’s function. Fructose-bisphosphate aldolase C is one of three paralogues in humans that catalyze the fourth step in glycolysis. Ferritin light change is one of the many subunits of the ferritin protein that facilitates intracellular iron storage. Recent studies have found that iron storage is important in colorectal cancer cell growth and proliferation.^48,49 Thymidylate synthase converts deoxyuridine monophosphate (dUMP) to deoxythymidine monophosphate (dTMP) which is essential for de novo synthesis of deoxythymidine monophosphate.

We chose to use relative quantification of each target by also monitoring the abundance of a housekeeping gene in our PRM experiment. For this role, we selected glyceraldehyde-3-phosphate dehydrogenase (GAPDH) because it is commonly used as a loading control in Western blotting and showed a consistent abundance across all injections in the box-and-whisker plots (Figure S4B). PRM samples were grown, processed, and analyzed independently of the global DIA proteomics samples. Peak integration was done manually in Freestyle after confirming the presence of at least 3 fragment ions for each target peptide. The peak area for each target peptide was normalized by taking the detected peak area of the quantitate ion for a target peptide and dividing by the sum of the peak areas for both peptides detected for the housekeeping gene. This normalization achieved precent coefficient of variances (CV) between biological replicates equal to or less than 30%. Percent CVs for technical injections were below 10%. Results of the relative quantification by PRM for thymidylate synthase, fructose-bisphosphate aldolase C, and ferritin light chain are displayed in Figure 6A,B,C, respectively. Each point in the bar graph correlates to the normalized peak area of one biological replicate for each condition. In correlation with our DIA-GPF data set, there was a statistically significant difference between the monolayer cells and day 6 spheroids as well as day 18 spheroids for every peptide detected. For both thymidylate synthase and fructose-bisphosphate aldolase C there was a statistically significant difference between day 18 and day 6 spheroids, as well.

Figure 6.

PRM results for thymidylate synthase (A), fructose-biphosphate aldolase C (B), and ferritin light chain (C). Each bar graph displays the average normalized peak area of three biological replicates for each day, with error bars indicating the standard deviation. The area for each biological replicate is indicated by the induvial data points for each bar graph. An unpaired type 2 students t test was used to determine statistical differences. An * indicates a p value ≤ 0.05, ** indicates a p value ≤ 0.01, and *** indicates a p value ≤ 0.001.

Together, these findings highlight that spheroid growth is accompanied not only by global proteomic longitudinal changes the proteome but also by alterations in specific protein targets as spheroids mature. Thus, studies done with spheroids harvested after 2 days of growth should not be compared to studies that harvest spheroids after 14 days of growth. Even though these studies are hypothetically done using the same model system, the data support that the length of growth time significantly impacts the final cellular system.

In this work, we focus on HCT 116 spheroids formed using the liquid overlay technique. However, these findings are likely applicable to other advanced microphysiological systems. The FDA Modernization Act 2.0 not only highlights the progress of these models but also underscores the need for researchers to carefully consider the fitness of their testbed. To obtain meaningful insights, it is essential to select the appropriate model system under the right conditions. Years of research have demonstrated that monolayer cells alone are insufficient for modeling in vivo systems. Therefore, if researchers are aiming to use more advanced models that better replicate the tumor microenvironment, HCT 116 spheroids should be grown for longer than just a few days. The timing of the spheroid harvest is a critical factor that significantly impacts the overall fitness of the model.

CONCLUSION

In this paper, we observe longitudinal proteomic changes in HCT 116 spheroids across 18 days of growth. We revealed that the proteome of spheroids grown for shorter periods of time is more similar to monolayer HCT 116 cells. These proteomic changes reflect differences in metabolism and a de-emphasis on DNA replication throughout spheroid growth. PRM experiments confirm these biological changes by determining the relative abundances of proteins involved in each of these pathways. Findings made regarding the time of spheroid harvest have implications for the use of spheroids as a model system. Ultimately, the time at which spheroids are harvested is a critical factor in experimental design and must be made with the utmost consideration.

Supplementary Material

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jproteome.5c00207.

Supplemental File 1: Additional experimental information, including quality control checks, SHINYGO figures, and PRM method information (DOC) Supplemental Figure 1. Quality Control Check. Box and whisker plots of normalized injections using equalized medians done in R using the MSstats package. Supplemental Figure 2. TIC for 2D HCT116 biological replicates 1–3 in (A), (B), and (C). TIC for Day 2 spheroids biological replicates 1–3 in (D), (E), and (F). We note what appears to be a loss of ion spray in panel B, for the second biological replicate of the monolayer cells. Supplemental Figure 3. After comparison to monolayer cells, the number of up and downregulated proteins for each time point. Supplemental Figure 4. Box and whisker plots of fructose-biphosphate aldolase C (A) and Glyceraldehyde-3-phosphate dehydrogenase (B) for each sample done in R using the MSstats package. Supplemental Figure 5. SHINYGO gene enrichment figures for Day 2 (A), Day 4 (B), Day 6 (C), Day 8 (D), Day 10 (E), Day 12 (F), Day 14 (G), Day 16 (H), and Day 18 (I) compared to monolayer cells. This figure depicts upregulated pathways throughout spheroid growth using the KEGG database Supplemental Figure 6. SHINYGO gene enrichment figures for Day 2 (A), Day 4 (B), Day 6 (C), Day 8 (D), Day 10 (E), Day 12 (F), Day 14 (G), Day 16 (H), and Day 18 (I) compared to monolayer cells. This figure depicts downregulated pathways throughout spheroid growth using the KEGG database. Supplemental Table 1. PRM Target Summarization (PDF)

Supplemental File 2: SHINYGO results in table format (XLSX)

ABREVIATIONS

DIA-GPF

data independent acquisition with gas phase fractionation

PRM

parallel reaction monitoring

2D

two-dimensional

3D

three-dimensional

PCA

principal component analysis

GO

gene ontology

LC-MS

liquid chromatography–mass spectrometry

TIC

total ion current

Data Availability Statement

Raw LC-MS proteomics files and searched.elib files are available in ProteomeXChange consortium in the PRIDE repository.