2.1.

Cell Lines and Reagents

PANC-1 pancreatic cancer cells used in the study were obtained from American Type Culture Collection (ATCC, Rockville, Maryland) and grown according as per ATCC descriptions in media containing $50\mathrm{IU}/\mathrm{mL}$ penicillin and $50\text{-}\mathrm{mg}/\mathrm{mL}$ streptomycin (Hyclone). Dulbecco’s modified Eagle’s medium culture media and fetal bovine serum were obtained from Hyclone Laboratories (Logan, Utah). The 3-D cultures of PANC-1 were grown in overlay on a bed of Growth Factor Reduced (GFR) Matrigel™ (BD Biosciences, Bedford, Massachusetts) with media supplemented with 2% GFR Matrigel, adapted from a methodology previously described.^6,48 Cultures were plated in MatTek Glass Bottom Culture Dishes (MatTek Corporation, Ashland, Massachusetts), which enable high quality optical imaging through the dish bottom. Media were changed once every third day during the growth process.

2.2.

Chemotherapy Treatment

For treatment response, solutions of Oxaliplatin (Selleck Chemicals, Houston, Texas) in 20 and $100\text{-}\mu \mathrm{M}$ concentrations were prepared in the growth medium and added to the cultures at day 3. Cultures were incubated in media containing the chemotherapy agent for 72 h. After 3-days, media was removed and fresh media without chemotherapy was added for further monitoring of the growth process.

2.3.

DHM Image Acquisition and Processing

The growth rate and treatment response of PANC-1 tumor nodules were imaged using an implementation of a typical off-axis DHM setup, shown in Fig. 1(a). The output of an Ar-Kr laser ($\lambda =488\mathrm{nm}$) is spatially filtered and well collimated. It is then divided into two beams using a beam splitter—one to transmit through the sample (hence termed as the object beam) and the other to serve as a reference beam. The object beam is coupled into a Zeiss (Jena, Germany) Axiovert 100 Observer A1 inverted microscope. Light transmitted through the sample is collected using a Zeiss EC Epiplan-Neofluar $20\times$ microscope objective and is collected from one of the video ports of the microscope. It is then relayed onto a SPOT Insight™ 2-MP Firewire CCD camera using a long focal length lens. The reference beam is made to interfere with the object beam at the camera with a certain angle between them (off-axis geometry) to create a uniform fringe structure oriented at 45 deg with respect to the $x$ and $y$ axes of the CCD. The white light source of the microscope is used to obtain bright field images. Imaging was performed on the same cultures at discrete time points in between which cell cultures were returned to the ${\mathrm{CO}}_2$ incubator. Individual culture dishes were on the microscope stage at lab atmosphere for only brief intervals, $\sim 15\min$, at each imaging time point.

Fig. 1

(a) Schematic of the off-axis digital holographic microscopy (DHM) setup using for imaging tumor nodules. M, mirrors; BS, beam splitter; IM, inverted microscope; L, field lens; O, object beam; and R, reference beam. (b) Reconstructed unwrapped image of three-dimensional (3-D) tumor nodules. Longitudinal 3-D imaging of two nodules (I and II) are shown in panels (c) and (d). (e) Confocal images of the same nodules confirming the thickness.

There were two dishes for each group—no treatment, $20\mu \mathrm{M}$, and $100\text{-}\mu \mathrm{M}$ oxaliplatin treatment. We acquired approximately 10 images per dish at randomly selected sites. As the field-of-view was large, there were about three multicellular nodules per image. Therefore, statistically each group represents the imaging of about 60 nodules per day. All images were processed and organized by timepoint and treatment group to generate analysis of the treatment-dependent growth behavior. Additionally, the entire experiment was rerun three times with a minimal variation of treatment timepoints and other parameters with almost identical results.

Reconstruction of the digitally recorded holograms is numerically performed by a standard computer. A detailed phase reconstruction of the object wave procedure is described elsewhere.^18,19,49,50 In short, the digital hologram is Fourier transformed, high pass filtered to eliminate the dc, and the twin image is suppressed. It is then inverse Fourier transformed and unwrapped to obtain the true phase information of the object. In order to obtain 3-D morphological parameters, the phase map $\varphi (x,y)$ is converted to a height map $h(x,y)$ as

2.4.

Measurement of Refractive Index of 3-D Tumor Nodules

A side by side comparison of DHM phase unwrapping and confocal fluorescence $z$-stacks of the same microscope fields was made to obtain the refractive index of the PANC1 3-D cultures. First, DHM images were obtained and the orientation of the nodules was marked. Cultures were then stained with calcein acetoxymethylesther (calcein AM, BD Biosciences), a fluorescent label for viable cells, and placed on the confocal microscope stage in precisely the same orientation so that exactly the same nodules could be imaged. In this manner, the optical path length obtained by DHM, combined with the coregistered physical object height from confocal imaging, is used to solve using Eq. (1) for the refractive index for individual nodules (relative to the known index of the surrounding medium) and calculated a mean value.

2.5.

Fluorescence Imaging and Analysis

DHM quantification of the treatment response was also validated by an established method based on the quantification of the signal from the fluorescent reporters of cell viability,⁹ as shown on day 11, where one set of oxaliplatin treatment cultures was stained with ethidium bromide (Promega Corporation, Madison, Wisconsin) and calcein (Life technologies, Oregon). For this purpose, $4\text{-}\mu \mathrm{M}$ ethidium bromide and $2\text{-}\mu \mathrm{M}$ calcein AM were prepared in DPBS (dulbecco’s phosphate-buffered saline—balanced salt) solution obtained from HyClone Laboratories (Logan, Utah). Multichannel fluorescence images were obtained for multiple fields of each replicate of each treatment group on a Zeiss Axio Observer inverted fluorescence microscope using standard filter cubes to provide excitation of EtBr and calcein at 528 and 494 nm and the corresponding emissions were at 617 and 517 nm, respectively. Collected images were processed to estimate the total residual volume that was normalized. The total residual volume represents the sum of the volumes of individual viable nodules, while its normalized value was obtained as the ratio of total volume to that of the group and that of the no treatment control group.

3.1.

DHM Volume Analysis

We monitored the growth of PANC-1 tumor nodules, grown for 11 days, in 3-D. Ensembles of holograms were acquired and digitally processed to obtain unwrapped phase images as described above. On day 3, cultures were assigned into groups that were treated with 20 and $100\text{-}\mu \mathrm{M}$ oxaliplatin, in addition to no treatment. On day 6, the oxaliplatin was removed and the tumors were allowed to grow. Figure 2(a) displays the variation of the tumor nodule diameter (solid lines) and thickness (dashed lines) averaged over reconstructions from large spatial fields containing many nodules in untreated (red) and treated ($20\mu \mathrm{M}$ in blue and $100\mu \mathrm{M}$ in green) cultures. In the case of untreated samples, both the thickness and the diameter were steadily increased during the entire period of study. However, when the treatment was initiated, the tumor nodule’s diameter and thickness were inhibited. The nodule size continued to decrease under the influence of oxaliplatin, but growth resumed upon removal of the cytotoxic agent as expected for regrowth following incomplete cell killing.

Fig. 2

DHM imaging experiments in 3-D tumor models in untreated (red) and treated ($20\mu \mathrm{M}$ in blue and $100\mu \mathrm{M}$ in green) cultures using off-axis holography. (a) Plot showing the discrete time intervals at which the averaged tumor nodule diameter and thickness are obtained. The solid and dashed lines are B-spline connection between the points showing typical behavior. (b) Plot shows the variation of the tumor nodule volume where bars showing standard error are displayed on the calculated volume data points. They are omitted for clarity in (a) due to the high density of data points. (c) Representative quantitative phase reconstructions obtained from longitudinally imaging. The days at which these images are recorded are represented by vertical lines in the plots (a) and (b). The vertical color bar represents the object height in microns (obtained from unwrapped phase).

A provocative insight revealed in Fig. 2(b) is the delay in response to the treatment. The $100\text{-}\mu \mathrm{M}$ treatment shows a rapid response to intervention with nodule diameter and thickness decreasing within hours of drug administration, whereas in the case of the $20\text{-}\mu \mathrm{M}$ treatment, the tumor nodules continued growing for an additional day before decreasing in volume. Figure 2(b) represents the plot representing the average integrated 3-D volumes of PANC-1 nodules, averaged over reconstructions from large spatial fields containing many nodules in untreated (red) and treated ($20\mu \mathrm{M}$ in blue and $100\mu \mathrm{M}$ in green) cultures. Bars showing the standard error, calculated from an ensemble of 60 images that were obtained per group per day, are displayed on the calculated volume data points but are omitted for clarity in Fig. 2(a) due to the high density of data points. The treatment ON and OFF time lines are indicated by arrows. It clearly displays the effect of the treatment response. As the untreated tumor nodules are steadily growing in size, the treated cultures show a decrease in the nodule volume as anticipated. Once again the noticeable difference between 20 and $100\text{-}\mu \mathrm{M}$ cultures growth inhibition is clearly observable. In Fig. 2(c), the diameter, thickness, and corresponding volume behaviors of the tumor nodules is illustrated by phase reconstructed images. The time periods for which the representative images are selected are shown as dashed vertical lines in Figs. 2(a) and 2(b). The vertical color bar adjacent to the phase reconstructed images is the object height in microns (obtained from unwrapped phase). On the first day, Day 0, as all three groups would have similar phase-reconstructed images, only one image is shown (first column). However, for 90, 144, and 230 h (on day 4, day 5, and day 10), we show the phase reconstruction images for all three groups.

Although measurement of the cellular dry mass has also been shown to be a powerful quantitative tool by which to quantify growth behavior,^11,12,55 the tumor volume is a more natural metric to use in this study, which is focused on monitoring therapeutic response.^56–60 Although tumor volume is a deeply established metric for treatment monitoring in clinical oncology and is widely implicated as a key prognostic indicator, it is also a logical choice for evaluation of oncology agents in preclinical systems. Although changes in the measured volume can, in fact, be driven by the osmolality of the surrounding medium in controlled experiments such as in present study, the main influence driving tumor volume change is the exposure to a cytotoxic agent which is known to reduce tumor volume by killing some fraction of the cancer cells. In addition, the current study was performed by changing the culture media at regular intervals to maintain a constant nutrient/salt balance. Therefore, while osmolality may also cause some degree of fluctuation in the measured tumor volume, the dominant effect in this treatment-focused study is due to the death of a dose-dependent fraction of the cells in each multicellular tumor nodule.

3.2.

DHM Volume and Estimated Volume Comparison

When studying the growth rate characteristics using two-dimensional (2-D) projections obtained by qualitative phase contrast or darkfield snapshots, it is a common practice to assume the tumor nodules have a perfectly spherical geometry and only the radius is used to calculate their volumes. We termed this calculated volume as the “estimated volume.” However, as the DHM images reveal 3-D information, the height of the cell is different from the radius. Therefore, the estimated volume will be different from the actual value. A bar plot comparing the tumor nodule volume to the estimated volume is shown in Fig. 3. In the untreated cultures, for the first 5 days of growth monitoring, the estimated volume is higher than that of the DHM volume. On day 6, the estimated volume is almost same as the actual volume and is lower as the days progressed. Thus, when the tumor nodule is small, the volume calculated with the assumption of spherical geometry tends to overestimate. We observed a similar trend during the treatment response period. These observations suggest that in the early stages or during the growth inhibition period, the estimated volume may lead to inaccurate growth inhibition conclusions.

Fig. 3

Bar graph showing difference between the average integrated 3-D volumes of untreated tumor nodules obtained using DHM procedure (red) and volume estimating considering the diameter alone (green).

3.3.

Volume Growth Rate Analysis

From the tumor nodule volume plot, we calculate the volume growth rate by linear fitting various intervals. Figure 4 demonstrates the ability to quantitatively report normal and estimated growth rates and the inhibitory effect with this method. By comparison, it is evident that the 2-D imaging overestimates the growth rate, tumor inhibition, and tumor regrowth. Although the growth inhibition is delayed by a day between various dosage concentrations, their slope is the same. However, the regrowth rate for $100\mu \mathrm{M}$ is slight more than that for the $20\mu \mathrm{M}$.

Fig. 4

Tumor nodule volume growth rate plot. Plot shows the discrete time intervals at which the averaged tumor nodule diameter was obtained. The solid lines are B-spline connection between the points showing typical behavior.

3.4.

Validation with Fluorescence Imaging

Fractional changes in tumor volume obtained by DHM were also independently validated by an established protocol based on quantification of the signal from fluorescent reporters of cell viability,⁹ as shown in Fig. 5. For this experiment, on day 8, cultures were assigned to groups that were either treated with $10\text{-}\mu \mathrm{M}$ oxaliplatin, or received no treatment. Figure 5 shows that the longitudinal volumetric growth monitoring via DHM before and during treatment administration beside terminal quantitative fluorescence measurements from separately stained cultures are in excellent agreement.

Fig. 5

Plot depicting validation of tumor volume obtained by DHM to that of the values obtained using quantification of signal from fluorescent reports of cell viability.⁹