Quantifying Biomass Changes of Single CD8+ T Cells during Antigen Specific Cytotoxicity

Existing approaches that quantify cytotoxic T cell responses rely on bulk or surrogate measurements which impede the direct identification of single activated T cells of interest. Single cell microscopy or flow cytometry methodologies typically rely on fluorescent labeling, which limits applicability to primary cells such as human derived T lymphocytes. Here, we introduce a quantitative method to track single T lymphocyte mediated cytotoxic events within a mixed population of cells using live cell interferometry (LCI), a label-free microscopy technique that maintains cell viability. LCI quantifies the mass distribution within individual cells by measuring the phase shift caused by the interaction of light with intracellular biomass. Using LCI, we imaged cytotoxic T cells killing cognate target cells. In addition to a characteristic target cell mass decrease of 20–60% over 1–4 h following attack by a T cell, there was a significant 4-fold increase in T cell mass accumulation rate at the start of the cytotoxic event and a 2–3 fold increase in T cell mass relative to the mass of unresponsive T cells. Direct, label-free measurement of CD8+ T and target cell mass changes provides a kinetic, quantitative assessment of T cell activation and a relatively rapid approach to identify specific, activated patient-derived T cells for applications in cancer immunotherapy.

Introduction CD8+ T lymphocyte mediated cytotoxicity is a critical component of the adaptive immune response against viruses and cancers, and is also implicated in autoimmunity [1,2]. T cell mediated cytotoxicity is typically measured by target cell death or surrogate markers of effector cell cytotoxic capacity. The canonical assays are the 51 Cr release assay and ELISPOT, both of which provide bulk measurements of whole lymphocyte population or sub-population responses [3,4]. The introduction of peptide-MHC tetramers and microfluidic platforms has allowed for surrogate measures of cytotoxicity through analysis of T cell antigen specificity and cytokine secretion [3,5,6]. Directly tracking T lymphocyte mediated cytotoxicity at the single cell level is advantageous for analyzing cytotoxic T cells (CTLs) within a mixed population, which is of particular relevance in assessing T cell recognition against cancer cells. Viable CTLs can potentially be cultured and expanded further, or the corresponding T cell receptors (TCRs) bearing optimal specificity toward immunogenic peptides can be molecularly cloned for utilization in a clinical setting [7].
Optical microscopy allows for direct identification and tracking of CTLs in the full context of target cell recognition and killing. Optical imaging methods such as epifluorescence, confocal microscopy, total internal reflection fluorescence and two photon laser scanning microscopy have been explored for the study of lymphocyte activation, but typically require antibody or conjugated protein labeling to track and quantify cells [8,9]. This limits applicability to studies of T lymphocytes due to transduction inefficiencies associated with diverse phenotypes as well as progressive differentiation towards exhaustion or senescence during in vitro culture, as is required for typical fluorescence labeling techniques [10,11]. Live cell interferometry (LCI) is a label-free optical microscopy technique which measures whole cell responses. LCI uses a Michelson-type interferometer to compare the optical thickness of living cells in a sample chamber to the optical thickness of fluid in a reference chamber in order to quantify the optical thickness difference between a cell and its surrounding media [12,13]. The optical thickness difference due to the interaction of light with cellular biomass is linearly proportional to the material density of a cell [14]. Based on this interaction, cell mass can be related to the measured phase retardation of light passing through each cell with 2% precision in total cell mass [12][13][14]. Practically, LCI yields measurements of mass and mass accumulation or loss rates of 100-400 cells simultaneously per imaging location within 1-5 h of imaging [12]. With automated measurements every 2-5 minutes to allow for accurate tracking and mass determination during cytotoxic events at 20-50 imaging locations, this technique can quantify the mass of 2,000 to 20,000 cells.
Our approach directly tracks T lymphocyte mediated cytotoxicity at the single cell level without labeling by quantifying the mass of individual CTLs and their cognate target cells. Single cytotoxic events are identified and evaluated over time within a mixed population, using the mass data to confirm individual T cell mediated cytotoxicity events. As a proof of concept, we demonstrate tracking of up to 2,000 individual CTLs with specificity toward Melanocytic Antigen Recognized by T lymphocytes (MART1) responding against human leukocyte antigen (HLA) matched MART1+ target cells [15]. Target cells are imaged by the LCI to establish a base-line mass accumulation rate. CTLs are then plated onto the target cells and individual cytotoxic events are identified as a characteristic decrease in target cell mass following contact with a corresponding T cell.
It is well established that T cells increase in size during activation [16]. This previously observed increase in size may result from a change in solute concentration or osmolality within the cell as opposed to an increase in biomass [17]. Using our approach we determine that the size increase in CTLs responding to cognate target cells is due to an increase in biomass and that biomass measurements provide robust identification of activated T cells. The capacity to measure the mass of a single CTL opens several potential downstream applications including T cell biological studies pertaining to metabolic or differentiation states in addition to the analysis of CTLs for potential use in adoptive immunotherapy protocols.

Cell Lines & PBMCs
M202, M207, [18] PC-3, PG13, and 293T cells (ATCC) were routinely maintained at 37uC in 8% CO 2 , using either DMEM or RPMI1640 Media supplemented with 5% FBS, 100 U/mL penicillin, 100 mg/mL streptomycin and 2 mmol/l-glutamine. HLA A2.1+ PBMCs derived from anonymized healthy donors were obtained from the Center for AIDS Research Virology Core Lab at UCLA and frozen following collection. Thawed PBMCs were revived in complete medium (CM) plus anti-CD3/2/28 beads for 4 d prior to retroviral infection. CM consisted of AIM-V media (Invitrogen, USA) supplemented with 25 mmol/L HEPES, 5.5610 25 mol/L b-mercaptoethanol and 300 IU/mL IL-2. PBMCs were in culture for a total of 7-10 d prior to all imaging experiments. Cells were maintained in complete media on the LCI imaging platform.
Generation of MART1 specific CD8+ T cells F5 retrovirus was collected from PG13 cells modified to produce retroviral vector consisting of the F5 TCR with specificity toward the MART1 ELAGIGLTV peptide fragment, which is expressed by the M202 and M207 cell lines used in cytotoxicity experiments. Briefly, 293T cells were transfected with the packaging vector pCL-Eco and the MSCV-based retroviral vector RV-MSCV-F5MART1 TCR. Resulting supernatants were used to transduce the murine PG13 retrovirus packaging cell line for Gibbon ape leukemia virus (GaLV) envelope-pseudotype generation. PBMCs were infected with the retrovirus containing PG13 supernatant in the presence of Retronectin (Takara, Japan) according to the manufacturer's protocol. 48-72 h after infection the cells were stained with MART1 specific tetramer (Beckman Coulter, USA) and analyzed by flow cytometry (FACSCanto, BD Biosciences, USA). CD8+ T cells were isolated by negative enrichment (Stem Cell Technologies, USA) and the enrichment efficiency was verified by flow cytometry.

IFNg measurement by flow cytometry
To verify the functional specificity of DMF5 transduced CD8+ T cells, a total of 1610 5 T cells were co-cultured with 1610 5 target cells (M202 or M207) in a 96-well flat plate with 200 ml of complete medium in a humidified incubator at 37uC and 8% CO 2 for 18 h. The concentration of IFN-gamma in the supernatant was determined by flow cytometry using the Human IFNg FlowCytomix Simplex kit following the manufacturer's protocol (eBioscience, USA cat# BMS8228FF).

LCI mass measurements
Target cells were plated onto 20 mm620 mm silicon slides treated with a 0.01% solution of poly-l-lysine (Sigma) at a density of approximately 2.5610 4 cells/cm 2 and allowed to grow in a cell culture incubator for 48 h prior to the start of imaging experiments. A silicon slide with attached target cells was placed into a custom-built, temperature and CO 2 controlled perfusionbased live cell imaging chamber and imaged for approximately 1.5 h before the addition of T cells. The T cell-target cell coculture was imaged continuously for 18 h. 30 imaging locations were chosen based on suitable density of target cells on the silicon substrate and images collected approximately once every 3 to 4 min. Imaging was performed using a modified GT-X8 optical profiler (Bruker) at 206 magnification (numerical aperture 0.28) with a 0.55x demagnifying lens to increase field of view while preserving resolution. Interference fringes were generated using a Michelson-type interferometer consisting of a beam splitter, reference mirror and a reference fluid chamber which compensates for the optical path length through the sample chamber. Images were acquired using the phase-shifting interferometry (PSI) method with illumination from a 530 nm fiber-coupled LED (Thorlabs). Intensity images represent the average intensity of the image without the interference fringes necessary for Michelson phase imaging.

Phase Unwrapping
To remove the integer-wavelength phase ambiguities inherent in quantitative phase imaging [19], we performed phase unwrapping using a custom script implemented in Matlab (Mathworks). First, we performed unwrapping based on Flynn's minimum discontinuity method [19]. Next, a training dataset was constructed by manually applying single wavelength corrections to approximately 200 sub-images of the phase data, selected for the appearance of target and T cells of interest. This training dataset was used in a linear discriminant analysis (LDA) to identify pixels which lie on the boundary of phase-wrapped regions, based on 16 sets of image statistics, including the raw image itself, the computed intensity image, and the results of various edge-finding filters applied to the wrapped phase image. LDA was followed by genetic optimization to refine the LDA results and watershed algorithm thresholds used in determining the boundaries of phasewrapped regions. Regions within the boundaries determined by the watershed algorithm applied to the final LDA result were shifted (corrected) by a phase shift of one wavelength and median filtered with a kernel size of 3.

Mass Tracking
Single cell mass measurements were performed using a custom script implemented in Matlab (Mathworks). Briefly, phasecorrected images were Gaussian low pass-filtered before image segmentation based on Otsu thresholding. Finally, objects identified by image segmentation were tracked using the particle tracking code adapted for Matlab by Daniel Blair and Eric Dufresne, based on the particle tracking algorithm by Grier et al. [20]. Cell area was determined using a local adaptive threshold based on a 200 pixel neighborhood in the quantitative phase image.

Statistics
Statistical analysis was performed using a two-tailed Welch's Student T test with unequal variances and sample sizes.

LCI for quantitative imaging of T cell mediated cytotoxicity
We developed a model system for analyzing cytotoxicity events by establishing the antigen specificity of healthy human donor CD8+ enriched lymphocytes against HLA matched target cell lines. Peripheral blood mononuclear cells (PBMCs) were transduced with an F5 anti-MART1 TCR, which is a high affinity TCR with specificity toward MART1 [15]. Target cells expressing MART1 and antigen-defined CD8+ enriched T cells were cocultured in a live-cell observation chamber on the LCI stage and imaged for a period of 18 h. ( Figure 1A). The observation chamber was temperature controlled with pH maintained by continuous perfusion of media equilibrated at 8% CO 2 . Following image collection, the light phase shift data was corrected for phase wrapping errors which are caused by the integer wavelength ambiguity inherent in quantitative phase imaging [19]. The result is a map of phase shifts across each cell that can be converted into a map of local dry mass density ( Figure 1B). The total dry mass of a cell is quantified as the sum of the local densities [12,14,21]: where m is cell dry mass, Ql is the measured phase shift, k is the mass conversion factor, and A is projected area. The mass conversion factor [21,22], which is a measure of the change in density per unit change in refractive index (Dr/Dn), is taken as k = 5.56 pg/mm 3 [14]. This parameter, k, is measured as a change in refractive index relative to the refractive index of water, Therefore, the cell mass measured in this manner is the cell dry mass, or the mass of everything within the cell other than water.
With this equation, the measured dry mass of the activated T cell in Figure 1B is 240 pg, the target cell mass is 840 pg and the unactivated T cells have an average dry mass of 65 pg.

Antigen-specific T cells and maintenance of viability on the imaging platform
To generate antigen-defined CTLs, we infected HLA A2.1+ healthy donor PBMCs with the F5 TCR by retroviral transduction and enriched for CD8+ cells by magnetic separation to remove magnetically labeled non-CD8+ cells (Figure 2A-B). Although CD8+ T cells have endogenous TCRs, ectopic expression of the F5 anti-MART1 TCR results in overexpression of the exogenous alpha and beta chains to allow for preferential pairing and surface expression. The majority of isolated cells were CD8+ with 75% expressing the F5 TCR on the surface, as determined by MART1 peptide tetramer stains prior to imaging. We measured interferon gamma (IFNg) accumulation in the supernatant following an 18 h co-culture period to verify that F5 redirected CD8+ T cells were specific for the cognate target cells. Results of a bead-based immunoassay analyzed by flow cytometry indicated a significant, 3.5-fold higher, IFNg release from F5 transduced CTLs upon coculture with HLA-matched MART1+ M202 target cells as compared to co-culture with an HLA-mismatched control cell line ( Figure 2C).
Target cells were imaged in standard culture media for 1.5 h prior to the start of each experiment to confirm the live cell culture imaging platform maintains viability of target cells in the absence of CTLs. M202 target cells showed a positive mass accumulation rate, indicating a healthy population and the maintenance of cell viability. (Figure 2D-E and Figure S1B). Control experiments demonstrated maintenance of both T and target cell viability during extended imaging periods ( Figure S1 and Figure S2).  (Figure 2A-B) were added to the live cell imaging chamber and imaged continuously for 18 h. This experiment duration is similar to the time period typically required for measurement of T cell activity by ELISPOT [3]. Single CTLs killing individual target cells are identified through qualitative analysis of the intensity image data as a change in appearance of the target cell following prolonged contact with a CTL ( Figure 3A-  Figure 3E-H). These mass distributions from successive image frames can be integrated to yield measurements of target cell mass over time (Equation 1 and Figure 3I). Individual cytotoxic events due to recognition of CTLs are confirmed by a characteristic decrease in target cell mass following prolonged contact (30 min to 2 h) with a corresponding CTL ( Figure 3I and Movie S1).

Mass decrease of killed target cells
Target cell mass decreased by 20 to 60% over a period of 1-4 h when successfully attacked by a CTL, as compared to an increase in total target cell mass of 15% over 4 h when not killed by CTLs ( Figure 3I This indicates that target cell death was due to the presence of antigen-specific CTLs and that the rate and extent of target cell mass decrease due to T cell mediated cytotoxicity is directly quantifiable using LCI. T cell mediated cytotoxicity is evident within the first 30 min and confirmed within the first 2-4 h following the addition of CTLs, indicating the speed of the LCI approach in measuring T cell mediated cytotoxicity (Movie S1). An estimated 95% of target cells were dead by 18 h after the addition of CTLs, while greater than 95% of control target cells appeared healthy at 18 h (Figure 3 K-L and Figure S3).

Mass increase of activated CTLs
In parallel with the decrease in target cell mass, individual activated CTLs increased in overall size by the end of a cytotoxic event (Figure 4). Individual CTL and target cell masses can be tracked through the duration of their interactions ( Figure 4A and Figure S4). CTL mass versus time data for 10 such events is summarized in Figure 4B, with CTL mass normalized relative to the mass when the target cell dramatically changed morphology (''balled-up'') at the start of a death event, which is defined as t = 0 h. In a typical trace, the target cell initially shows an increase in mass consistent with the growth rate of a healthy cell ( Figure 3M). During this period (t,0 h), CTLs show a relatively slow growth rate ( Figure 4C). Then, the target cell ''balls-up'' and detaches from the substrate, immediately prior to a very rapid loss of mass over the first 1-2 hours. During this initial period (approximately 100 min), the T cell mass accumulation rate increases significantly ( Figure 4C). As the target cell loses mass and the central cell body condenses over the next 2-5 hours, the T cell continues to increase in mass, at a slower rate than during the initial period ( Figure 4C). This change in mass accumulation rate resulted in a significant 2 to 4-fold higher cellular mass than surrounding unresponsive T cells ( Figure 4D). The total cellular mass of 116 CTLs at the endpoint of each cytotoxic event was compared to the mass of 3,900 control T cells that did not kill targets during the course of the experiment. On average, the CTLs had a 2.8-fold higher mass as compared to their non-specific or unresponsive counterparts ( Figure 4E and Figure S5A). This mass increase persisted for up to 4 h, a duration that is limited by the average period of observation prior to the activated T cell being washed away due to continuous media perfusion through the observation chamber.
The two-dimensional (2D) area of responsive versus unresponsive T cells was calculated to determine whether there was a significant difference relating to overall size. The observed 1.4-fold increase in 2D area was smaller than the 2.8-fold difference in total cell mass and did not achieve statistical significance at the p,0.05 level compared to controls ( Figure 4F and Figure S5B). These results show that the mass change of CD8+ T cells is a more robust indicator for activity than the change in cell area. Additionally, for spherical T cells, the observed 1.4-fold increase in mass corresponds to a 1.7-fold increase in volume, which is substantially lower than the observed 2.8-fold increase in mass. These results, therefore, suggest that there is also an increase in T cell density during activation, although density quantification is not possible with the present configuration of LCI measurements.

Discussion
LCI provides a quantitative label-free cytotoxicity assay through sensitive biomass measurements of single effector T cells and their affected target cells during cytotoxic events (Figure 1). The mass of killed target cells can be tracked over time to confirm a 20 to 60% decrease in mass over 1 to 4 h, consistent with a cytotoxic insult ( Figure 3). We found a significant 4-fold increase in T cell mass accumulation rate at the start of the cytotoxic event and a 2.8-fold average increase in total mass of effector T cells after recognition and killing of cognate target cells (Figure 4). The change of mass of T cells was found to be a more significant indicator of T cell activation state than measurements of 2D changes in area alone.
The mass increase we observed in activated CTLs is likely accompanied by an increase in biosynthesis driven by metabolic changes. It has been demonstrated that T cells use glucose and glutamine as their primary energy sources. Activated lymphocytes generate energy to meet protein synthesis demands by significantly increasing glucose, amino acid and fatty acid uptake from the extracellular environment [23]. Glucose deprivation studies have shown that activated T cells require glucose for proliferation and survival even in the presence of adequate levels of glutamine [24]. TCR signaling plays a critical role in regulating the transcription of the glucose transporter Glut1, enabling enhanced glucose uptake with activation [25]. Studies have shown that TCR agonists such as anti-CD3 antibodies or compounds that cause cross-linking of CD3 proteins result in a rapid and maximal induction of Glut1 expression [24,25].
A potential application of the LCI technique presented here is for the identification and isolation of single and potentially rare CTLs. A growing body of work has focused on the identification of tumor infiltrating T lymphocytes (TILs) bearing TCR recognition of autologous tumor cells [7,26]. Recent studies have indicated that these CTLs occur at relatively low frequencies, making it difficult to employ bulk or surrogate cytotoxicity assays to confirm their existence and isolation from a mixed population [27,28]. The LCI approach uses the cytotoxic interaction between CTLs and target cells as a natural amplifier of the underlying peptide-MHC-TCR recognition event which avoids false positives due to nonspecific binding. The LCI imaging platform is fundamentally compatible with a segmented culture system that will allow for isolation of rare cells that may be lost in the current open perfusion cell culture system. LCI may therefore provide a viable alternative for the identification and isolation of rare effector T cells killing autologous tumor cells or HLA-matched cancer cell lines.
T cells against cancer-associated antigens are generally anticipated to bear lower affinity TCRs if they are raised against a selfantigen and presumably escaped thymic selection and tolerance induction [29]. The affinity between the TCR and peptide-MHC is considered to play a crucial role in the outcome of T cell stimulation [30]. The classic method to assess TCR-peptide-MHC affinity entails the measurement of on and off-rates using surface plasmon resonance. The surface bound peptide-MHC-TCR interaction does not accurately mimic the multiple receptormediated interactions that occur during recognition of a target cell by a CTL. Evidence suggests that these measurements provide limited information regarding lymphocyte effector function [30,31]. In a transfection system, TCRs engineered with higher affinity for cognate peptide-MHC ligands compared to their wild type counterpart exhibited increased CTL activity [31]. An affinity model suggests that activation of T cells is related to the number of receptors engaged. Higher affinity interactions require less TCRpeptide-MHC engagements to activate a T cell into a cytotoxic state [32]. It is conceivable that higher affinity TCR-peptide-MHC interactions drive a more rapid response than their lower affinity counterparts, and the LCI approach may also potentially discriminate between these interactions.