Label-free drug response evaluation of human derived tumor spheroids using three-dimensional dynamic optical coherence tomography

This study aims at demonstrating label-free drug-response-patterns assessment of different tumor spheroids and drug types by dynamic optical coherence tomography (D-OCT). The study involved human breast cancer (MCF-7) and colon cancer (HT-29) spheroids. The MCF-7 and HT-29 spheroids were treated with paclitaxel (Taxol; PTX) and the active metabolite of irinotecan SN-38, respectively. The drugs were applied with 0 (control), 0.1, 1, and 10 μM concentrations and the treatment durations were 1, 3, and 6 days. A swept-source OCT microscope equipped with a repeated raster scanning protocol was used to scan the spheroids. Logarithmic intensity variance (LIV) and late OCT correlation decay speed (OCDS\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$_l$$\end{document}l) algorithms were used to visualize the tumor spheroid dynamics. LIV and OCDS\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$_l$$\end{document}l images visualized different response patterns of the two types of spheroids. In addition, spheroid morphology, LIV, and OCDS\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$_l$$\end{document}l quantification showed different time-courses among the spheroid and drug types. These results may indicate different action mechanisms of the drugs. The results showed the feasibility of D-OCT for the evaluation of drug response patterns of different cell spheroids and drug types and suggest that D-OCT can perform anti-cancer drug testing.


Introduction
Cancers are ones of the deadliest human diseases [1,2].The burden of cancer and the related mortality rate could be reduced through early detection and provision of appropriate treatment in a patient-specific manner.Hopefully, cancer patients may have a high chance of recovery if they are diagnosed early and treated appropriately using suitable anti-cancer drugs.However, there are several types of cancer based on which cell type has been mutated.The type of cancer varies from patient to patient and thus the suitable drug will also vary patient by patient.
One of the most effective approaches to understand the cancer's biology and select anti-cancer drugs appropriately is to cultivate patient-derived tumor cells as a three dimensional (3D) cluster, so-called multi-cellular tumor spheroid (MCTS).An MCTS closely mimic the main features of in vivo solid tumors, including their structural organization and the gradients of their oxygen, pH, and nutrients [3][4][5][6].In addition, the MCTS is similar to the in vivo solid tumor in several ways, including growth kinetics, cell interactions, metabolic rates, and its response or resistance to the applied chemotherapy and radiotherapy [7,8].MCTSs have therefore been used widely in the field of anti-cancer drug investigations.An anti-cancer drug's efficacy can be evaluated based on its impact on the shape and volume alteration of the MCTS, the dissociation of the MCTS cell aggregations, and the MCTS cells viability degradation [9][10][11].
MCTS evaluations can be performed using several conventional methods, e.g., staining histology [12,13], fluorescence microscopy [14][15][16], and bright field microscopy [17,18].Although these methods are the gold standards, they have several limitations.First, most of these methods use chemical labeling and/or tissue slicing and are thus invasive.Second, most of them are 2D imaging modalities.Third, the image penetration depth of these methods is limited to a few hundreds of microns, which prevents entire-depth imaging of thick tissues such as the MCTS.Finally, the quantitative capabilities of these methods are low.Some of the limitations mentioned above, e.g., invasiveness and limited penetration depth, can be overcome using optical coherence tomography (OCT) [19] and its microscopic version socalled optical coherence microscopy (OCM) [20].OCT is a low-coherence interferometry-based biomedical imaging modality that provides high-resolution, label-free, non-invasive, 3D imaging over a few millimeters penetration depth.
Several hardware and software extensions have been integrated with conventional OCT to provide additional contrasts to the OCT.For example, attenuation coefficient (AC) imaging [21,22] is sensitive to tissue density, and was found to be sensitive to both the apoptotic and necrotic processes of human fibroblasts [23].The tissue polarization properties can be measured using polarization-sensitive OCT [24], which provides collagen-sensitive [25,26] and melanin-sensitive [27][28][29] contrasts.Tissue stiffness, which is an important parameter in tumor investigations, can also be evaluated via optical coherence elastography [30][31][32][33].
One interesting OCT extension is dynamic OCT.As an emerging technique, dynamic OCT provides tissue activity/sub-cellular motion contrasts.Dynamic OCT represents a combination of high-speed OCT systems with the signal processing of sequentially acquired OCT frames.Timefrequency analysis-based en face dynamic full field OCT (DFF-OCT) has been demonstrated previously and was found to be sensitive to the adenosine triphosphate (ATP) consumption-related tissue metabolism [34][35][36].High-speed scanning OCT and time spectroscopic analysis have been used to perform for cross-sectional dynamic imaging [37,38].Both of these methods are able to provide cross-sectional tissue activity/sub-cellular motion contrast.However, the methods may not be suitable for volumetric imaging applications.
Recently, we demonstrated 3D dynamic OCT, which combines a standard-speed swept-source OCT device with a custom-designed 3D scanning protocol and was shown to acquire volumetric tissue dynamics in 52.4 s [43,44].The 3D tissue dynamics were visualized and quantified using our previously proposed dynamic OCT signal analysis algorithms [38].
This 3D dynamic OCT technique visualized the viable and necrotic cell layers of MCTSs successfully.In addition, similarity between the fluorescence microscopy and en face dynamic OCT contrasts was demonstrated [44].In the same work, a preliminary study of the tumor spheroid's response to anti-cancer drug was presented.However, that study involved only one type of tumor (MCF-7 cell line, human breast cancer), and only one anti-cancer drug, paclitaxel (PTX, or Taxol), was used.
In general, treatment of different tumor cell types with different anti-cancer drugs leads to different spatial patterns of the anti-cancer drug responses.The spatial pattern differences among the different tumor spheroids and various drug types are worth to be investigated to aid in the selection of appropriate anti-cancer drugs, and these differences can be highlighted using our proposed dynamic OCT method.
In this paper, more generalized drug response studies based on the 3D dynamic OCT are Fig. 1.Time protocol for spheroid cultivation and measurement.The same time protocols were used for both MCF-7 and HT-29.60 spheroids were cultured for both cell types.Each spheroid was seeded with 1,000 cells on day-1.The drugs, i.e., PTX for MCF-7 and SN-38 for HT-29, were applied on day-5.On days-6, 8, and 11, fluorescence (FLUO) and then OCT imaging were performed.The fluorescence contrast agents were applied 3 hours before FLUO imaging.
presented.Breast and colon cancers and the combination of these cancers are among the most prevalent causes of cancer-related death worldwide [45][46][47][48].Therefore, we selected two tumor cell types for this study, including the human breast cancer (MCF-7) and human colon cancer (HT-29) cell-lines.The breast and colon cancer spheroids are treated using an anti-cancer drug (PTX) and an active metabolite of Irinotecan (SN-38), respectively.Image-based 3D observation of the time-course drug response shows different spatial patterns of the drug effects among the different tumor and drug types.In addition, quantification of the dynamic OCT and the spheroid morphology also reveals different time courses among the cell and drug types.

Protocols and samples of drug response studies
The tumor spheroid drug response evaluation study involves use of two different human-derived tumor cell-lines, including human breast adenocarcinoma (MCF-7) and human colon cancer (HT-29) cell-lines.Both cell-lines were cultured using the same protocol in line with the study protocol depicted in Fig. 1.On day-1, 1,000 tumor cells were seeded in each well of a ultra-low attachment 96-well-plate and 60 wells were prepared.The cells were cultured at a temperature of 37 • C and the cell-culture chamber was supported with 5 % CO 2 .The culture medium contained a 1:1 mixture of Eagle's minimal essential medium (EMEM; for MCF-7) or Dulbecco's modified eagle medium (DMEM; for HT-29) and F12 (Invitrogen, Waltham, MA) with a 2 % B-27 supplement (Invitrogen), 2 ng/mL of basic fibroblast growth factor (bFGF; Wako, Osaka, Japan), 2 ng/mL of epidermal growth factor (EGF; Sigma-Aldrich, St. Louis, MO), 100 U/mL of penicillin G, and 0.1 mg/mL of streptomycin sulfate (Wako, Osaka, Japan).The cells were aggregated with each other and formed one spheroid in each well on day-5.On the same day, each spheroid was then treated using 0.1 μM, 1 μM, or 10 μM of the relevant anti-cancer drugs.The MCF-7 spheroids were treated using paclitaxel (PTX, Sigma-Aldrich; also known as Taxol), while the HT-29 spheroids were treated using an active metabolite of Irinotecan (SN-38, Chemscene LLC, NJ).In addition, untreated (0 μM) spheroids were retained as control samples.After drug application the spheroids were kept within the same culture environment.
The spheroids were then extracted from the cultivation and measured using the fluorescence (FLUO) and dynamic OCT microscopes at three time-points of days-6, 8, and 11, i.e., 1-, 3-, and 6-days after drug application, respectively.Here, we also denote the days 6, 8, and 11 as the 1st, 3rd, and 6th days of treatment (TD-1, -3, and -6).At each time-point, five spheroids were measured for each of the four drug concentrations, including control (0 μM), meaning that 20 spheroids were measured in total.The measurements were performed in a room temperature of 25 • C.

Fluorescence microscopic imaging
Live/dead fluorescence assays were performed using the THUNDER imaging system (Leica Microsystems, Wetzlar, Germany) with a microscopic objective with a numerical aperture (NA) of 0.12.Two contrast agents were applied for three hours before the FLUO imaging.The first agent is calcein acetoxymethyl (Calcein-AM; Dojindo, Kumamoto, Japan), which stains the living (viable) cells and emits a green fluorescence signal.The second agent is propidium iodide (PI; Dojindo), which contrasts the dead cells by emitting red fluorescence signal.

OCT device
A polarization-sensitive Jones matrix swept-source OCT (JM-OCT) was used to perform the three-dimensional (3D) tissue dynamics imaging.A light source with a central wavelength of 1.3 μm and a scanning speed of 50,000 A-lines/s was used.The axial (in tissue) and lateral resolutions were 14 μm and 18.1 μm, respectively.The complete specification for the JM-OCT has been published elsewhere [26].Although this device is polarization-sensitive, only polarization-insensitive OCT images, which are the average of four OCT intensity images acquired through four polarization channels, were used for the tissue dynamics imaging.

OCT data acquisition protocol
The spheroid samples were transferred to our 3D JM-OCT system after FLUO imaging.For the 3D dynamics imaging, the lateral imaging field was divided into eight regions.Each region consists of 16 B-scan locations and was scanned repeatedly using a quick raster scanning protocol for 32 times in 6.55 s.Therefore, 32 sequential frames were obtained at each location in the sample with a frame repetition time of 204.8 ms.The OCT volume consists of a total of 128 locations and was captured in 52.4 s.The scanning area was 1 × 1 mm 2 .Further details about the 3D dynamics scanning protocol are available elsewhere in the literature [44].

Dynamic OCT algorithm: Logarithmic intensity variance (LIV)
The time-sequential OCT signals were processed using logarithmic-intensity-variance (LIV) algorithm [38].The LIV is a measure of the magnitude of the temporal fluctuations of the dB-scaled OCT signals at each location in the tissue as illustrated in Fig. 2(a), and it is defined as: where   (, ,   ) is the dB-scaled (logarithmic) OCT intensity,   is the sampling time for the -th frame, where  = 0, 1, 2,...., -1, and  is the total number of frames, which is 32 in our case, and  represents averaging over time.
Because the LIV is a measure of the magnitude of the signal fluctuations, it is believed to be sensitive to the magnitude of the intracellular motility.
An LIV image is presented as a pseudo-color image in which the hue and the value of the HSV (hue, saturation, value) color representation are set as the LIV and the averaged OCT intensity image, respectively.LIV is defined as the temporal fluctuation (variance) of the dB-scaled OCT intensity.As a result, LIV is expected to be sensitive to the magnitude of the tissue dynamics.
On the other hand, OCDS  represents the speed of the dynamics and is defined as the decorrelation rate, i.e., as the slope of the auto-correlation curve [red curve in (b)] of the dB-scaled OCT intensity.The slope is computed here by performing a linear regression of the auto-correlation curve over a specified delay-time range [204.The time-sequential OCT signals are also processed using yet another dynamic OCT algorithm: "late OCT correlation decay seed (OCDS  )." OCDS  is defined as the speed (rate) of autocorrelation decay of the sequentially captured dB-scaled OCT intensity and it is expected to be sensitive to the slow tissue dynamics [38].
To obtain the OCDS  , the temporal auto-correlation   (  ; , ) was first computed as: where the numerator represents the covariance between   (, ,   ) and   (, ,   +   ), and Var[ ] represents the variance.  represents the delay time, which is defined as Δ where  is an integer variable.Δ is the B-scan repetition time, which is 204.8 ms in this case.The OCDS  is then defined as the correlation decay rate, i.e., as the slope of   (  ; , ) over a specific range of  = [204.8,1228.8 ms] as illustrated in Fig. 2(b).
An OCDS  image is presented as a pseudo-color image, which is generated in the same manner as the LIV image, but in this case the hue is the OCDS  signal.Further details about the LIV and OCDS  algorithms are published elsewhere [44].

OCT-based volumetric quantification
To quantify the volumetric alterations of the morphology and viability of the spheroid cells, the spheroid volume was computed via B-scan by B-scan segmentation.The segmentation was performed using the find-connected-region plugin of the Fĳi ImageJ software with a manually defined OCT intensity threshold.The initial segmentation contained the bottom surface of the well-plate because it shows high OCT intensity.This well-plate surface was removed manually.
In addition to the spheroid volume, the mean values of the LIV and OCDS  over the entire spheroid volume were computed using the segmentation results.Furthermore, by applying empirically defined cut-offs of 3 dB 2 and 2 ×10 −4 ms −1 for the LIV and OCDS  , respectively, the necrotic cells volume (i.e., the volume of the LIV and OCDS  signals that are lower than the cut-offs in each case) is computed [44].The necrotic cell ratio (= necrotic volume / entire spheroid volume) is then computed.These computed quantities are plotted as a function of the treatment time for each drug concentration, as shown in Section 3.

MCF-7 spheroid response to PTX
Figure 3 summarizes the en face LIV, OCDS  , and FLUO images of the MCF-7 spheroids treated with PTX.In the control case (the first column), the low LIV (red) and low OCDS  (red) regions increase over time.These low-dynamics regions at the center of the spheroid may correspond to the dead cells (red) shown in the FLUO images.It is also found that the boundaries between the high and low signal regions are clearer in the OCDS  images than the LIV images.
On the first day of treatment (TD-1), the PTX-treated spheroids show similar appearances to the control sample.However, as the treatment time increases, the spheroids show shape corruption, volume reduction, and degradation of the LIV and OCDS  signals.The FLUO images show increase of dead cells over time in addition to the shape corruption.
Notably, the LIV and OCDS  images of the 0.1-μM and 1-μM cases on day-8 (TD-3) show completely different patterns (as magnified in Fig. 4).In the 0.1-μM case, tessellated low dynamics (red) patterns appear in the OCDS  image (black dotted shapes), while the LIV images shows low dynamics (red) overall.The FLUO image also shows some domains composed of dead cells (red, yellow dashed lines).In the 1-μM case, the LIV image shows two domains including a central low-dynamics (red) region and a peripheral region that is a mixture of low (red) and high (green) dynamics.In contrast, the OCDS  image shows three concentric domains of low, high, and low dynamics.The FLUO image shows dead cells (red) at the periphery (indicated by yellow arrowheads) and the center regions,with live cells (green) located in between.In both cases, the appearance of the FLUO images are more consistent with OCDS  than LIV images.
For the reference of the readers, OCT intensity images corresponding to the images shown in Fig. 3 are presented in the supplementary material (Fig. S1).The fast-scan cross-sections of LIV and OCDS  extracted at the locations indicated by the white arrow heads in Fig. 3 are also presented in a supplementary figure (Fig. S2).
Images of additional spheroids measured at under each treatment condition and at each time point are presented in a supplementary figure (Fig. S3) and are consistent with the images in Fig. 3.
In Fig. 5, the spheroid volume (a), the mean LIV (b), the mean OCDS  (c), the LIV-based necrotic cell ratio (d), and the OCDS  -based necrotic cell ratio (e) are plotted as functions of the treatment time.The spheroid volumes of the control case increases over time may be due to the growth of the spheroids.In contrast, the drug-treated spheroids show decreasing volume trends.
The mean LIV and mean OCDS  over the entire spheroid [Fig.5(b) and (c)] show decreasing trends over time for all PTX concentrations, including the control case.A significant reduction in the mean LIV was found for all PTX concentrations, except for 10μM from TD-1 to TD-3 (P = 0.047, 0.018, and 0.006 for the control, 0.1, and 1.0 μM) and from TD-3 to TD-6 (P = 0.006, 0.006, and 0.018 for the control, 0.1, and 1.0 μM).For the 10-μM samples, the TD-1-to-TD-3 and TD-3-to-TD-6 reductions were insignificant, but TD-1-to-TD-6 reduction was significant (P = 0.006).Significant reductions in the mean OCDS  were also found for the three PTX concentrations from TD-1 to TD-3 (P = 0.030, 0.006, 0.018 for the control, 0.1, and 1 μM) and from TD-3 to TD-6 (P = 0.006, 0.006, 0.030 for control, 0.1, and 1 μM).For the 10-μM case, the reduction from TD-1 to TD-3 (p = 0.006) was significant, although it was not significant from TD-3 to TD-6 (p = 0.148).This may occur because the mean OCDS  has already become very low on TD-3.All the p-values of these tests are summarized in Table S1 (supplementary).
The necrotic cell ratios obtained with both LIV and OCDS  [Fig.5(d) and (e), respectively] clearly increase over the treatment times for all PTX concentrations.Significant increases in the LIV-based necrotic cell ratios were found from TD-1 to TD-3 (P = 0.030, 0.006, and 0.047 for 0.1 μM, 1 μM and 10 μM, respectively), and from TD-3 to TD-6 (P = 0.006, 0.047, and 0.030 for 0.1 μM, 1 μM, and 10 μM, respectively) for all PTX concentrations except for the control sample.For the control sample, the observed increase was significant from TD-3 to TD-6 (P = 0.006).A significant increase in the necrotic cell ratio was also found in the OCDS  -based necrotic cell ratio from TD-1 to TD-3 for all PTX concentrations, including the control (P = 0.030, 0.018, 0.010, and 0.006 for the control, 0.1 μM, 1 μM and 10 μM).Similar increases were also found from TD-3 to TD-6 for all concentrations, except for the 10-μM case (P = 0.006, 0.006, and 0.030 for the control, 0.1 μM, and 1 μM).All statistics are summarized in Table S1 (supplementary).
The degradation in the mean LIV and mean OCDS  and the increasing necrotic cell ratios for the PTX-treated spheroids may be related to inhibition of the intracellular transport through the microtubules.This point will be discussed in detail later in Section 4.1.

HT-29 spheroid response to SN-38
Figure 6 summarizes the en face LIV, OCDS  , and FLUO images of the HT-29 spheroids after treatment with SN-38.For the TD-1 spheroids at all SN-39 concentrations, including the 0 μM (control) case, the LIV appearance is contrary to that of the MCF-7 spheroids with PTX.Specifically, the center region shows high LIV (green), and this area is surrounded by low LIV (red) regions.For the 10-μM case in particular, the low LIV region is further surrounded by a moderately high LIV periphery.In contrast, the appearance of the OCDS  images is similar to that in the MCF-7 case.Namely, the center shows low OCDS  (red), and this area is surrounded by a high OCDS  periphery (green).
For the longer treatment times (TD-3 and TD-6), the LIV images show granular patterns of high and low LIV, and clear domain structures are not observed.This appearance is contrary to that of the MCF-7, and is cell-type dependent.In contrast, most of the OCDS  images show double or triple concentric domain structures.These structures may be the result of two factors, which include drug effect from the peripheral side and hypoxia or lack of nutrients starting from the center region.The appearances of the OCDS  images are quite consistent with the FLUO images.The dynamic OCT and FLUO images show clear volume reductions and/or structural corruption under the SN-38 treatment.
Figure 7 shows magnified images of representative spheroids with both volume reduction and structural corruption.These magnified images demonstrate the high pattern correlation between the OCDS  and FLUO images, as indicated by the arrowheads and dashed lines.In contrast, the LIV patterns are not similar to the OCDS  and FLUO images.
Figure 8 shows the morphological and mean dynamic OCT signal alterations for the HT-29 spheroid treated with SN-38.It was observed that the data in the TD-6 10-μM case had much standard deviations (i.e., error bars in the figure) than the other cases for the mean LIV [Fig.8(b)], LIV-based, and OCDS  -based necrotic cell ratios [Fig.8 (d) and (e), respectively].By observing the LIV images for all five spheroids in this case [Fig.9], it was found that one spheroid case (spheroid 1) exhibited very low LIV values over the entire spheroid region.We suspect that this spheroid may be fragmented because of some cultivation issues or because of spheroid handling during the measurement.We therefore regard this case as an outlier and excluded it from the subsequent statistical analysis.The plots that include this outlier are presented with red spots and red lines, while those that exclude it are shown with orange color in Fig. 8.
The control spheroid volume [Fig.8(a)] increased from TD-1 to TD-3, and then decreased slightly from TD-3 to TD-6.This volume alteration of the control spheroid may indicate the growth and spontaneous corruption of the HT-29 spheroid over the cultivation time.In contrast, the spheroids that were treated with SN-38 showed monotonic reductions in their spheroid volumes over time for all drug concentrations.
The mean LIVs [Fig.8(b)] increased from TD-1 to TD-3 for all drug concentrations, and then decreased from TD-3 to TD-6 for the control sample (P = 0.006, Mann-Whitney test; supplementary Table S2).
The mean OCDS  in the 1-μM and 10-μM SN-38 concentration cases demonstrated significant  3).Specifically, the LIV images of the TD-1 spheroids show high LIV (green) at the center surrounded by low LIV (red).After longer treatment times, the LIV images show a granular appearance.The OCDS  images show concentric domain structures.In addition, the OCDS  and FLUO images are found to be highly correlated.The scale bars represent 200 μm in each case.reductions over time, as illustrated in [Fig.8(c)].The significant reduction in the 1-μM case was found from TD-3 to TD-6 (P = 0.006, Mann-Whitney test), while that in the 10-μM was found from TD-1 to TD-3 (P = 0.018).For all other SN-38 concentrations, the treatment-time dependency was not observed.All statistical test results are presented in supplementary Table S2.
The LIV cut-off-based necrotic cell ratio demonstrated a reduction in the necrotic cell ratio from TD-1 to TD-3 for all SN-38 concentrations, and the ratio then became stable from TD-3 to TD-6.The reduction in LIV-based necrotic cell ratio from TD-1 to TD-3 was significant in the 1-μM and 10-μM cases of SN-38 (P= 0.047 and 0.010, respectively; Mann-Whitney test).
On the other hand, the necrotic cell ratios based on the OCDS  cut-off remain stable over time for low concentrations (i.e., the control and 0.1 μM cases), while that of the 10-μM case increases from TD-1 to TD-3 (P = 0.006).The ratio of the 1-μm shows a significant increase from TD-3 to TD-6 (P = 0.006).The results of this test are presented in Table S2 (supplementary).
The morphological and dynamic-OCT signal alterations of the HT-29 spheroid treated with SN-38 may be related to the drug mechanism of SN-38, as will be discussed later in Section 4.2.The OCT intensity, cross-sectional LIV and OCDS  results for the same data presented in Fig. 6 are presented in Figs.S4 and S5 in the supplementary material.In addition, the responses of one additional HT-29 spheroid under each treatment condition are presented as supplementary Fig. S6.

Drug effects of PTX appearing on OCT and dynamic OCT
Dynamic OCT imaging of the control MCF-7 spheroid [Fig.3 (first column)] showed circular low dynamic signals (LIV and OCDS  ) surrounded by a high-dynamics shell.Similar image appearances were observed in our previous studies [38,44].The low and high dynamics observed at the spheroid core and the periphery are believed to highlight the well-known necrotic core (caused by the lack of both oxygen and nutrients supply) and the viable rim of the tumor spheroids, respectively [3,49].In addition, the increase of low dynamics area over the cultivation time [Fig.

(first column)] may indicate an increase in the necrotic area over the long cultivation time.
Similarity between the dynamics patterns of the MCF-7 spheroids treated with PTX for one day and that of the control spheroid was observed in Fig. 3 (first to third rows).This may indicate either the low efficacy of PTX or the resistance of the tumor cells when the treatment time is only one day.
In addition to the findings mentioned above, the general tendency of the structural and dynamic OCT observed in this study can be explained as follows using the drug mechanism of PTX.Microtubules (MTs) are polymers that are formed by polymerization of  and  tubulin dimers [50].This MTs have three important functions in the eukaryotic cells.First, MTs are the main constituents of the cytoskeleton, which maintains the cell structure and supports the cell with the mechanical resistance to deformation [51].Second, the MTs are involved in mitosis as the main constituents of the mitotic spindles [50,52].Third, the MTs act as platforms for intracellular transport of the vesicles and organelles [50,53].
For the functions described above, the MTs exhibit highly dynamic behavior.They exert rapid and stochastic phase transitions from extension (by polymerization) to shrinkage (by de-polymerization) [50,54].This dynamic instability allows the MTs tips to search for and bind to intracellular components, e.g., capturing chromosomes during mitosis.[50,54].
PTX is an MT-stabilizing anti-cancer drug.This drug prevents MT depolymerization and suppresses the dynamic behavior of the MTs [55,56].Therefore, it leads to cell cycle and mitotic arrest, which ultimately leads to cell death [55,56].
The structural OCT findings for the MCF-7 spheroid after treatment with PTX showed spheroid shape corruption and volume reduction over the treatment time.This spheroid shape corruption may indicate malfunctioning of the first function of the MTs (i.e., maintaining the cell structure and the resistance to cell deformation) under the application of PTX.On the other hand, reduction of the spheroid volume may indicate inhibition of mitotic cell divisions (the second function of the MTs) under the effect of PTX.
On the other hand, the spheroids that were treated with PTX showed reductions in their LIV and OCDS  signals over the treatment time.The LIV and OCDS  reductions may indicate a reduction of the intracellular motility/transport through the MTs (the third function of MTs) and may also indicate PTX-induced cell death.

Drug effects of SN-38 appeared on OCT and dynamics OCT
The dynamics patterns of the TD-3 HT-29 spheroid under all the SN-38 concentrations, except for the control case [Figs.6 and 7] showed a clear low OCDS  layer at the periphery and this layer corresponded well with the peripheral dead cells (red fluorescence) observed in the FLUO images.Among the three concentrations, the 10-μM case showed evident changes in the mean LIV and the mean OCDS  from TD-1 to TD-3 [Fig.8(b) and (c)], and also showed the smallest spheroid volume among all concentrations, even at TD-1 [Fig.8(a)].These results indicate that OCT and dynamic OCT imaging can demonstrate higher drug effects of 10-μM SN-38 than the other lower concentrations.
On TD-6, all the SN-38-treated spheroids at all concentrations (0.1, 1.0 and 10 μM) showed severe structural corruption.In comparison, on TD-3, these shape corruptions were not evident for all concentrations.However, the OCDS  images showed clear low OCDS  layers at the periphery that were not found in the control case.In addition, spheroid volume analysis [Fig.8(a)] showed evidently smaller spheroid volumes, even at TD-3.These findings suggest that dynamic OCT can assess the more moderate drug effect of SN-38 than that can be visualized using non-dynamic OCT.
The alterations of OCT and dynamic OCT signals and the images of the HT-29 spheroids [Figs.8 and 6, respectively] can be understood more by taking the mechanism of SN-38 into account.Irinotecan is a DNA topoisomerase I (Topo-I) inhibitor that is used to treat the advanced stages of colorectal cancer and it is also approved for second-line treatment in metastatic colorectal cancer [57,58].SN-38 is an active metabolite of Irinotecan, which is 100-to 1000-fold more active when compared with the Irinotecan itself [57,59].Because Topo-I is an enzyme involved in DNA transcription and replication, the inhibition of Topo-I caused by SN-38 induces DNA damage and a transient S-phase (the cell cycle in which DNA is replicated) arrest.SN-38 application then leads to irreversible breaks in the single-strand-DNA that are subsequently transformed into double-strand-DNA breaks.As a result of these double-strand-DNA breaks, a series of different apoptotic-related signaling pathways are activated.This then results in apoptotic cell death [57,60].The damage of DNA and subsequent DNA replication may suppress cell division, and this can account for the smaller spheroid size at TD-3.This DNA damage then progresses into apoptosis, and this can lead to the structural corruption observed at TD-6.The thin low-OCDS  (red) peripheral layers observed in all SN-38 treated spheroids on TD-3 may indicate the onset of apoptosis at the spheroid-drug interface.We suspect that the low-OCDS  may indicate the cell death process because the necrosis at the control spheroid center, which is another type of cell death, also exhibits low OCDS  .

LIV and OCDS 𝑙 highlight different aspects of dynamics
LIV is a measure of the magnitude of the temporal fluctuations in the dB-scaled OCT intensity and it may be sensitive to the magnitude of the intracellular motility.On the other hand, the OCDS  quantifies the speed of the OCT intensity fluctuations and it may be sensitive to the speed of the intracellular motility, as discussed elsewhere [38,44].Because of this fundamental difference between them, the LIV and OCDS  may highlight different cellular processes.And this may explain why different image patterns appeared in LIV and OCDS  .For example, in the control MCF-7 spheroids [Fig.3], OCDS  shows clearer boundaries between the central and peripheral regions than LIV.In addition, in some cases, OCDS  shows tessellated domain structures such as those shown in Fig. 4 that are not visible in the corresponding LIV images.
Although the identification of the cellular process that accounts for these LIV and OCDS  findings remains an open issue, we conclude that LIV and OCDS  provide different information and thus are complementary methods.10], it can be found that OCDS  exhibit similar appearances for both cell types.Namely, the center region shows low OCDS  , while the periphery exhibits high OCDS  .In contrast, the LIV patterns of these spheroids are opposite to each other.Namely, the MCF-7 spheroid shows low and high LIV at its central and peripheral regions, respectively, while the HT-29 spheroid showed high and low LIV, respectively, in the corresponding regions [Fig.10 (a) and (f)].
To further investigate this difference, the time sequence of the OCT intensity [Fig.10(b) and (g)] and the autocorrelation curves [Fig.10(c) and (h)] are presented; these figures show the averaged intensity and the autocorrelation curves in the yellow boxed regions indicated in the OCT and dynamic OCT images.Both autocorrelation curves show that the correlation is strongly decayed even at the first computing point (204.8ms), and thus the signal fluctuations in both cases are very fast.In comparison, the time sequence OCT intensities show that the signal fluctuation magnitude is small for MCF-7 but it is large for the HT-29 spheroid.
To provide further validation of this fast autocorrelation decay, a single B-scan location in the same spheroids was scanned 350 times in 4.48 s, i.e., it was scanned with much higher temporal density.This is the protocol that we presented for our original 2D dynamic imaging, and it enables yet another dynamic OCT contrast, the fast/early correlation decay speed (OCDS  ) [38].This contrast is similar to OCDS  , but the slope of the correlation decay is defined at an earlier time range [12.8, 64 ms] than OCDS  , and thus it is more directly sensitive to the fast signal fluctuations.The OCDS  images of the spheroids are presented in Fig. 10 (d) and (i).As shown in these images, the OCDS  is high at the center regions of both spheroids.This represents further proof that the dynamics at the center regions are fast.
It is known that the center part of the spheroid becomes necrotic because of the lack of oxygen and other nutrients [6,49,61,62].In addition, it has been reported previously that cell necrosis shows a fast correlation decay [63].As a result, we suspect that the center regions of both spheroids are in the necrotic process.
It is also known that the necrotic process takes a long time after the cell death starts [64].The difference in the magnitude of the signal fluctuations may be accounted for by the time after the onset of cell death.Namely, we can assume that the onset of cell death for this particular MCF-7 spheroid was early, and hence the necrotic process has been progressed, this results in the low fluctuation magnitude.One the other hand, the cell death onset of the HT-29 spheroid may be late, which means that the signal fluctuation magnitude is still large.
To further investigate these differences in the spheroid core activities, bright field microscopic images acquired using (IX71, Olympus) microscope with an objective with 4x magnification and an NA of 0.13 from the same samples that were presented in Fig. 10 are shown in Fig. 10.
The bright field image of the MCF-7 spheroid [Fig.10 (e)] shows a dark appearance at its center, which is a so-called necrotic core, with very clear boundaries surrounded by a bright appearance (viable cells).In contrast, the HT-29 spheroid [Fig.10 (j)] shows a diffusive border.This may indicate that the necrotic core has not completely formed yet.These bright field images thus support our hypothesis.
In addition, our previous study involving time lapse imaging of the MCF-7 spheroid demon- strated that the LIV at the central region was high at the early time points and it gradually degraded along hours [38].These results may provide further support for our hypothesis.

Other dynamic OCT methods
Including our LIV and OCDS  , there are several dynamic OCT methods.These methods can be classified into two categories: magnitude evaluations and speed evaluations of the OCT signal fluctuations.The latter category can be categorized further into two subcategories that include autocorrelation analysis and time spectrum analysis.The magnitude analysis may include following research works.The cumulative sum (cumsum) method demonstrated by Scholler quantifies the signal fluctuation magnitude using a Brownian bridge model, and it has high tolerance to bulk sample motion [65].The motility amplitude method demonstrated by Oldenberg is also a metric to quantify the signal fluctuation magnitude [66].These modalities are expected to visualize the fluctuation magnitudes of tissue and cellular dynamics.
The correlation analyses can quantify the speeds of the tissue dynamics.Leroux et al. computed the autocorrelation decay of an OCT signal and fitted it with a biexponential function, so that cellular dynamics with multiple speed were quantified [67].
Spectrum analysis is another method that can be used to quantify the temporal characteristics of the OCT signal fluctuations and tissue dynamics.The spectroscopic analysis demonstrated by Oldenburg et al. is one of the earliest demonstrations of this approach [66].Apelian et al. combined a signal frequency analysis with a high-resolution full-field OCT imaging, and enabled high-resolution and functional imaging of ex vivo animal tissues [34].Scholler et al. visualized the activity of retinal organoids through power-spectrum analysis of time-sequential OCT signals [36].Recently, Münter et al. used a time-frequency analysis to enable three-dimensional visualization of tissue dynamics, and visualized the cellular scale dynamics of ex vivo animal tissues [40,42].Both the correlation and spectral analysis methods may contrast the speed of the tissue dynamics.
In the present study, we used two methods: LIV and OCDS  .LIV is a method to quantify the magnitude of the signal fluctuation, quantifies the speed of the fluctuation.As discussed in Section 4.3, the combination of these two methods provides a comprehensive interpretation of the dynamic processes occurring in the tissue.

Conclusion
The spatial patterns of the anti-cancer drug responses of two types of human tumor spheroid, including those derived from breast (MCF-7) and colon (HT-29) cancer cell-lines have been visualized using two types of dynamic OCT algorithms.The response patterns observed in dynamic OCT images were consistent with the corresponding fluorescence microscopy patterns.In addition, the spheroid volume and mean dynamic OCT signal values were computed.The time-course changes in these values revealed different trends among the two types of spheroid and the anti-cancer drugs.
In conclusion, dynamic OCT can be used to highlight the difference in drug response patterns among different tumor spheroids and different drug types.Therefore, dynamic OCT can be a useful tool for anti-cancer drug testing and for the optimal selection of anti-cancer drugs.

Fig. 2 .
Fig. 2. Schematic depictions of two dynamic OCT algorithms:(a) LIV and (b) OCDS  .LIV is defined as the temporal fluctuation (variance) of the dB-scaled OCT intensity.As a result, LIV is expected to be sensitive to the magnitude of the tissue dynamics.On the other hand, OCDS  represents the speed of the dynamics and is defined as the decorrelation rate, i.e., as the slope of the auto-correlation curve [red curve in (b)] of the dB-scaled OCT intensity.The slope is computed here by performing a linear regression of the auto-correlation curve over a specified delay-time range [204.8,1,228.8ms].

Fig. 3 .
Fig. 3. LIV, OCDS  , and FLUO images of MCF-7 spheroid treated with PTX.The images are en face slices at around the center of the spheroids.The B-scan crosssections and the intensity OCT images are available in the Supplementary materials (as Fig. S2 and Fig. S1, respectively.)Although the difference between the control sample and the PTX-treated samples is not obvious with the 1-day treatment, the 3and 6-day-treated spheroids showed significantly different appearances to the control sample.The appearances of the FLUO images are almost consistent with those of the LIV and OCDS  images.The scale bars represent 200 μm in each case.

Fig. 4 . 3 .Fig. 5 .
Fig. 4. Magnified LIV, OCDS  , and FLUO images of representative spheroids from Fig. 3.The upper row shows one treated with 0.1-μM PTX, while the lower row shows one treated with 1-μM PTX.The treatment time in both cases is 3 days.The appearance of the FLUO images is more consistent with the OCDS  images than LIV images.

Fig. 6 .
Fig. 6.En face LIV, OCDS  , and FLUO images of an HT-29 spheroid treated with SN-38.The LIV and OCDS  patterns are very different to those for MCF-7 when treated with PTX (Fig.3).Specifically, the LIV images of the TD-1 spheroids show high LIV (green) at the center surrounded by low LIV (red).After longer treatment times, the LIV images show a granular appearance.The OCDS  images show concentric domain structures.In addition, the OCDS  and FLUO images are found to be highly correlated.The scale bars represent 200 μm in each case.

Fig. 7 .
Fig. 7. Magnified images of HT-29 spheroids with sever volume reduction and shape corruption.These images were extracted from Fig. 6.High levels of pattern correlation between the OCDS  and FLUO patterns are observed, while the LIV patterns are not correlated well with the other images.

Fig. 8 .Fig. 9 .
Fig. 8. Morphological and dynamics analysis of the HT-29 spheroid when treated with SN-38.The plots are presented in the same order as those in Fig. 5.The plotted values represent the average values of five samples (N=5) under each treatment condition and the error bars represent the ± standard-deviation range.

4. 3 . 2 .
Combined interpretation of LIV and OCDS  of tumor cell necrosis By comparing the dynamics OCT images of the TD-1 control spheroids of MCF-7 and HT-29 [Fig.

Fig. 10 .
Fig. 10.Detailed analysis of apparent pattern differences between the control MCF-7 (right) and HT-29 (left) spheroids.(a), (f) Cross-sections of the OCT intensity, LIV, and OCDS  .(b), (g) Time course of the OCT intensity in the core areas of the spheroids (as indicated by the yellow boxes).MCF-7 shows very low fluctuations, while HT-29 shows high levels of fluctuation.(c), (h) Auto-correlation curves for both spheroids, showing rapid decay as the auto-correlation became very low at the first sampling point of the delay time (204.8ms).(d), (i) OCDS  cross-sections of the same spheroids, showing the fast decay (green)at the center parts of the spheroid.These OCDS  images were computed from a cross-sectional frame sequence of 350 frames acquired at a single location.(e), (j) Bright-field microscopy images of the spheroids, which show the necrotic cores as dark central region.

Fig. 12 .
Fig. 12. Cross-sectional LIV and OCDS  images of the MCF-7 spheroid treated with PTX.The images were extracted from the locations indicated by white arrow head on the en face images in Fig. 3 in the full length manuscript, and they show similar tendency to the en face images.Scale bar represent 200 μm.

Fig. 13 .
Fig.13.En face LIV and OCDS  images of additional MCF-7 spheroids treated with PTX.The measurement and treatment were performed with the same protocol to that of Fig.3in the full-length manuscript.The images are presented in the same order as Fig.3in the full-length manuscript.Scale bar is applicable for all the LIV and OCDS  images and it represents 200 μm.

Fig. 14 .Fig. 15 .
Fig. 14.Cross-sectional and en face OCT intensity images of the HT-29 spheroids presented in Fig. 6 in the full-length manuscript.The intensity images show the morphological alteration of the HT-29 tumor spheroid.Scale bars represent 200 μm.

Fig. 16 .
Fig. 16.The en face LIV and OCDS  images of additional HT-29 spheroids treated with SN-38.The images are presented in the same manner of Fig. 6 in the full-length manuscript.The LIV and OCDS  images show similar patterns to those presented in Fig. 6 in the manuscript.Scale bar represents 200 μm.