Abstract

   Patient‐derived cancer organoids (PDOs) hold considerable promise for
   personalizing therapy selection and improving patient outcomes.
   However, it is challenging to generate PDOs in sufficient numbers to
   test therapies in standard culture platforms. This challenge is
   particularly acute for pancreatic ductal adenocarcinoma (PDAC) where
   most patients are diagnosed at an advanced stage with non‐resectable
   tumors and where patient tissue is in the form of needle biopsies. Here
   the development and characterization of microfluidic devices for
   testing therapies using a limited amount of tissue or PDOs available
   from PDAC biopsies is described. It is demonstrated that microfluidic
   PDOs are phenotypically and genotypically similar to the gold‐standard
   Matrigel organoids with the advantages of 1) spheroid uniformity, 2)
   minimal cell number requirement, and 3) not relying on Matrigel. The
   utility of microfluidic PDOs is proven by testing PDO responses to
   several chemotherapies, including an inhibitor of glycogen synthase
   kinase (GSKI). In addition, microfluidic organoid cultures are used to
   test effectiveness of immunotherapy comprised of NK cells in
   combination with a novel biologic. In summary, our microfluidic device
   offers considerable benefits for personalizing oncology based on cancer
   biopsies and may, in the future, be developed into a companion
   diagnostic for chemotherapy or immunotherapy treatments.

   Keywords: chemotherapy, immunotherapy, microfluidic device, pancreatic
   cancer, patient‐derived organoid
     __________________________________________________________________

   Treatment of pancreatic cancer may be made more effective by selecting
   therapies for an individual patient. The authors developed and
   characterized microfluidic devices that may be used to test
   chemotherapies and immunotherapies using needle core biopsies from
   patients with pancreatic cancer.

   graphic file with name ADVS-11-2303088-g005.jpg

1. Introduction

   There is an increasing realization that one‐size‐fits‐most approaches
   to treating cancer patients are ineffective and that treatment regimens
   need to be personalized to an individual patient. This necessitates the
   development of patient‐specific cancer models for therapy testing.^[
   [50]^1 ^] Patient‐derived cell lines and patient‐derived xenograft
   (PDX) models represent some of the established approaches for
   individualizing cancer therapy selection.^[ [51]^2 ^] While incredibly
   useful, these strategies have disadvantages. Cell lines undergo genetic
   and morphological changes over time in culture, resulting in
   inconsistent and inaccurate outcomes.^[ [52]^3 ^] PDX models take
   several months to establish with success rates for tumor engraftment
   varying from 20 to 60% depending on the tumor type.^[ [53]^4 ^] PDX
   models of pancreatic ductal adenocarcinoma (PDAC) are challenging to
   establish given the high degree of fibrosis and low cellularity of PDAC
   tissue.^[ [54]^5 ^] Furthermore, PDAC is an aggressive cancer with 65%
   of diagnoses occurring at stage III or higher with median survival time
   of 8–10 months.^[ [55]^6 ^] The time required to establish a PDX model
   often approaches the survival time which means that a given model may
   not be used for personalized therapy.

   Patient‐derived cancer organoids (PDOs) have the potential of
   addressing limitations of other patient‐specific cancer models.^[
   [56]^2a ^] PDOs are 3D self‐organizing structures that are derived from
   primary tumor cells^[ [57]^7 ^] and provide dimensionality and topology
   similar to that of the native tumor.^[ [58]^2a ^] Compared to other
   cancer models, PDOs are less time‐intensive and costly to establish and
   may therefore be used to rapidly select suitable drugs for aggressive
   cancers. PDAC organoid cultures were first described by Muthuswamy and
   Tuveson labs in 2015.^[ [59]^8 ^] Since then, PDOs derived from PDAC
   and other tumor types have been shown to recapitulate patient‐specific
   abnormalities and to be predictive of patient responses to therapy.^[
   [60]^9 ^]

   Biopsies are routinely collected during diagnosis of PDAC and represent
   starting material for PDO generation.^[ [61]^10 ^] However, needle core
   biopsies typically have a low number of cells (≈100000–500000 cells
   depending on the cancer type and patient),^[ [62]^11 ^] which makes it
   challenging and time‐consuming to generate organoids in sufficient
   numbers to test therapies using standard culture systems.^[ [63]^12 ^]
   The time window between diagnosis of PDAC and chemotherapy
   administration is ≈4 weeks which means that organoid formation,
   expansion, and therapy testing needs to fit within that time frame.

   Microfluidic cell cultures are particularly well‐suited for scenarios
   where testing of multiple drugs needs to be carried out with a limited
   number of cells for cultivating cancer cells and testing responses to
   therapy.^[ [64]^13 ^] For example, Prince et al. developed a
   microfluidic platform integrating gradient generators and hydrodynamic
   traps and used this device to culture clusters of cancer cells and test
   their response to different concentrations of drugs.^[ [65]^13c ^] The
   Takayama lab has developed a microfluidic hanging drop system for
   culturing cancer spheroids and testing chemotherapies,^[ [66]^13d ^]
   while the Kamm lab has pioneered microfluidic cultures integrating
   cancer cells with vasculature to study extravasation.^[ [67]^14 ^]
   There have also been reports of microfluidic cancer cultures for
   testing cancer immunotherapy.^[ [68]^15 ^] For example, Ayuso et al.
   employed a tumor‐on‐chip system to explore anti‐cancer activity of NK
   cells,^[ [69]^15a ^] while Ronteix et al. used a droplet microfluidic
   device to co‐entrap cancer and immune cells for testing
   immunotherapies.^[ [70]^15b ^] Additional approaches employed
   microcapsules comprised of extracellular matrix (ECM) and carrying
   cancer spheroids for therapy testing. For example, Ding et al.
   demonstrated that colon cancer spheroids cultured in Matrigel
   microparticles retained cellular heterogeneity of native tumor tissue,
   could be used to test immunotherapies, and were predictive of patient
   responses to chemotherapy.^[ [71]^3 ^] With a few recent exceptions,^[
   [72]^3 , [73]^16 ^] microfluidic cultures reported to date relied on
   cancer cell lines or PDX models. To the best of our knowledge, there
   has not been a report of microfluidic organoid cultures derived from
   patient biopsies in general and from PDAC biopsies specifically.

   We have previously demonstrated that microfluidic devices are
   well‐suited for rapidly testing therapies based on an input of small
   number of cells.^[ [74]^17 ^] In this prior study, ovarian cancer
   organoids were derived from PDX tissue and shown to better proliferate
   and maintain epithelial cancer phenotype in microfluidic cultures
   compared to traditional culture approaches (standard 3D plate and
   Matrigel cultures).

   In the present study, we take the concept of microfluidic cultures
   farther and describe microfluidic cultivation of PDAC organoids derived
   from patient needle biopsies. As shown in Figure [75] 1 , our devices
   contained arrays of microwells (250 µm diameter) that were made
   low‐binding to promote aggregation of cells into spheroids. To address
   the challenge of minimal number of cells available from biopsies, our
   novel microfluidic device incorporated a port for direct injection of
   organoids or organoid fragments. This design feature enhanced
   efficiency of organoid/tissue utilization. Our study demonstrates that
   while microfluidic organoids were comparable to the gold‐standard
   Matrigel‐based organoids in terms of gene and phenotype expression,
   microfluidic cultures offered multiple advantages including 1) uniform
   size and ease of assessing effects of chemotherapy, 2) the ability to
   add NK cells to create immune cell‐cancer organoid co‐cultures and to
   study cancer‐immune cell interactions, and 3) minimal reliance on
   Matrigel. In addition, to demonstrating cultivation of three different
   PDO lines in the novel device we also assessed novel therapies. These
   included GSK inhibitor therapy which was shown to sensitize PDO
   responses to front‐line chemotherapy (gemcitabine) and also cancer
   immunotherapy consisting of NK cells and a novel immunomodulatory
   biotherapeutic. Overall, our study highlights utility of microfluidic
   organoid cultures for assessing cancer treatment using scant tissue
   available from pancreatic cancer biopsies and represents a step toward
   establishing these cultures as a companion diagnostic in the future.

Figure 1.

   Figure 1
   [76]Open in a new tab

   Schematic describing the concept of this study: Needle core biopsies
   were collected from patients, digested, cultured as organoids, and then
   placed into microfluidic devices. Two types of microfluidic devices
   were used in this study: 1) device type 1 for seeding single cells and
   2) device type 2 for seeding intact organoids or organoid fragments.
   Microfluidic organoid cultures were used to test several chemotherapy
   drugs as well as a novel immunotherapy comprised of NK cells and
   immunomodulator molecule TriKE.

2. Results and Discussions

2.1. Comparing Microfluidic and Standard Organoid Cultures

   One key objective for this study was to evaluate novel microfluidic
   organoid cultures in comparison to more standard culture
   formats—Matrigel and microwell arrays. Throughout this study we used
   two variants of microfluidic devices—device 1 for seeding single cells
   in scenarios with relative abundance of cells and device 2 for seeding
   intact organoids or organoid fragments when cells were scant.
   Comparison of different culture formats reported below was carried out
   with PDO‐001—a poorly differentiated cancer that proliferated in vitro
   producing cells in numbers sufficient for in‐depth analysis.
   Figure [77] 2A and [78]S1 (Supporting Information) shows a workflow for
   the study—PDAC biopsies were dissociated, expanded as organoids on
   Matrigel, and then collected and seeded into microfluidic devices.
   Figure [79]S2 (Supporting Information) shows images of patient biopsies
   used in this study. The extent of expansion on Matrigel was variable.
   Ten to twenty passages were possible for PDO‐001 ensuring supply of
   these organoids while only five to seven passages were possible for
   PDO‐002 and ‐003. Thus, the latter two tissues were of limited supply
   and were cultured in device 2.

Figure 2.

   Figure 2
   [80]Open in a new tab

   Formation of PDAC organoids in microfluidic devices using single cell
   digest as starting material. A) Illustration of single cell seeding in
   microwells from PDOs. Microfluidic device type 1 used to form organoids
   from single‐cell suspension. Inset image shows organoids (PDO‐001) in
   microfluidic device after 5 days of seeding. B) Images of organoids
   from PDO‐001 in Matrigel (upper panel) and microfluidic device (lower
   panel) at different time points during culture (Scale bars = 100 µm).
   C) Quantification of organoid size (area) change in Matrigel and
   microfluidic devices. D and E) Immunofluorescence staining for
   epithelial (E‐cadherin), PDAC (CK19), and proliferation (Ki67) markers
   in organoids cultured in Matrigel D) and microfluidic devices E). The
   scale bars are 100 µm for Matrigel and magnified inset images in
   microwell, and 200 µm for microarray. F) Microwell array culture
   format—cloning cylinders were mounted onto a PDMS based with arrays of
   microwells. Inset image showing organoids (PDO‐001) at 5 days of
   culture. G) Comparing organoid growth in microwell array and
   microfluidic cultures. Merging of organoids made it challenging to
   quantify growth rates in Matrigel cultures. H) Comparing gene
   expression of proliferation marker (Ki67) and cancer markers (CK19 and
   Pdx1) for culture formats. Statistical significance of each group
   indicates ** for p < 0.01 and *** p < 0.001, respectively.

   Figure [81]2A shows device 1 that consisted of five independent
   microfluidic compartments, each with media reservoirs and transport
   channels passively delivering nutrients to the cell culture chamber.
   The latter contained an array of 19 wells, each of 250 µm in diameter.
   The devices were composed of gas‐permeable silicone rubber (PDMS)
   allowing diffusion of atmospheric oxygen and were functionalized with
   nonionic surfactant Pluronic to make microwells low‐binding and to
   promote spheroid formation.^[ [82]^18 ^] As may be appreciated from the
   images in Figure [83]S3 (Supporting Information) and Figure [84]2B,
   spheroid formation occurred 2 days after seeding, with spheroids
   increasing in diameter during subsequent 7 days of culture. Compared to
   Matrigel cultures, spheroids in microfluidic devices were more
   uniformly distributed (Figure [85]2C and Figure [86]S3, Supporting
   Information). This is a key advantage of our culture format, making it
   possible to observe changes in spheroid geometry over time in the
   presence or absence of therapy. Comparable tracking is challenging with
   Matrigel cultures because spheroids are not uniform and have a tendency
   to merge (Figure [87]2B; Matrigel).

   In an effort to benchmark microfluidic organoids against the
   gold‐standard Matrigel cultures, we carried out immunofluorescence
   staining of both culture formats. These results, shown in
   Figure [88]2D,E, demonstrate similarities between the two cultures.
   Both stained strongly for E‐cadherin – a cell adhesion marker
   indicative of epithelial cells comprising pancreatic ducts.^[ [89]^19
   ^] Microfluidic (inset) and Matrigel organoid cultures also had
   comparable expression levels of a PDAC‐specific marker CK19,^[ [90]^20
   ^] and a proliferation marker Ki67.

   There have been several reports of microwell array systems used for
   creating uniformly sized cancer organoids and testing therapies.^[
   [91]^21 ^] Therefore, we wanted to set up a microwell array as another
   comparator culture format and fabricated PDMS insert with an array of
   microwells of the same diameter and pitch as the microwells of the
   microfluidic device. However, unlike the microfluidic device where
   spheroids were confined to a local volume of 3 μl, the microwells in
   the insert were placed into 150 µL – a volumetric equivalent of a well
   from a 96‐well plate. This experiment was designed to address a
   question of whether small local volume of the microfluidic device was
   beneficial for proliferation and phenotype maintenance of pancreatic
   cancer organoids. Previously, we demonstrated this to be the case for
   ovarian cancer organoids that proliferated and expressed epithelial
   cancer phenotype better in microfluidic devices compared to microwell
   arrays.^[ [92]^17 ^] The custom microwell arrays used for comparison
   with microfluidic cultures are shown in Figure [93]2F. Comparison of
   changes in organoid size over time revealed similar proliferation
   dynamics for microfluidic device and microwell array (see
   Figure [94]2G); however, PCR analysis highlighted better gene
   expression in the microfluidic device. As seen from Figure [95]2H, PDAC
   organoids in the microwell array cultures were significantly less
   proliferative (lower Ki67) and expressed lower levels of PDAC‐specific
   markers, CK19 and Pdx1 compared to microfluidic organoid cultures.
   Given better expression of proliferation and phenotypic markers, we
   relied exclusively on the microfluidic organoid cultures for testing
   chemotherapies and immunotherapy going forward.

   Having demonstrated benefit of maintaining organoids in microfluidic
   devices over microwell arrays, we proceeded with a more in‐depth
   comparison of Matrigel and microfluidic organoid cultures. Culturing
   PDAC organoids in Matrigel in WRN‐containing media was first described
   by Boj et al. in 2015 remains the gold standard today.^[ [96]^22 ^] We
   carried out RNA sequencing of organoids derived from the same patient
   biopsy (PDO‐001) after cultivation in either Matrigel or microfluidic
   devices. The study was limited to one patient because of the scarcity
   of organoids available from other patients (PDO‐002 and −003). First,
   samples were compared to each other and to an existing dataset of
   publicly available samples from Hogenson et al.^[ [97]^9c ^] As shown
   in Figure [98] 3A and Figure [99]S4 (Supporting Information), the
   microfluidic device (Device) and Matrigel samples clustered closer to
   each other and were separated by one principal component accounting for
   74% of the distance between the samples (Figures [100]3A and
   Figure [101]S4, Supporting Information). The device and the Matrigel
   organoids were more similar to each other than to other PDAC PDO
   samples generated from primary tumors, PDX and metastases on principal
   component analysis and when comparing sample to sample distances
   (Figure [102]3A,B). This was reflected by similar gene expression of
   classical and basal pancreatic cancer signaling and developmental
   pathways (WNT, NOTCH and TGF‐β) known to be involved in organoid
   formation and maintenance (Figures [103]3C and [104]2D).^[ [105]^23 ^]
   Several genes associated with the Classical PDAC subtype (e.g., LYZ and
   TESC) and with the basal‐like subtype that is linked to worse outcomes
   (e.g., S100A2, ANO1, and DCBLD2) were upregulated in microfluidic
   devices. A volcano plot of the differentially expressed genes enriched
   in Matrigel and device organoid cultures is shown in Figure [106]3E and
   Figure [107]S4C,D (Supporting Information). From this analysis, we
   determined that the majority of genes (99.93%) were similarly expressed
   between the device and Matrigel organoid cultures with only 0.038%
   genes enriched in Matrigel and 0.037% in the microfluidic device
   (Figure [108]3E). The full list of differentially expressed and shared
   genes between matrigel and microfluidic devices is in Table [109]S1
   (Supporting Information). Using gene ontology analysis for biological
   processes (Figure [110]3F) and gene set enrichment analysis against
   curated gene sets (Figure [111]3G), we determined that Matrigel
   organoid cultures were enriched for pathways related to hypoxia, which
   may point to insufficient oxygenation compared to microfluidic organoid
   cultures. Matrigel cultures were also enriched for changes in cell
   metabolism including the regulation of purine synthesis and ATP
   biosynthesis. These changes may suggest increased anaerobic respiration
   due to the hypoxic state of organoids in Matrigel and may negatively
   impact their long‐term viability, phenotype, and response to therapy.
   By contrast, microfluidic organoids appeared to be enriched in pathways
   related to improved communication with the microenvironment
   (Figure [112]3F,G). Of particular note was upregulated expression of
   cytokines known to promote chemotaxis of NK cells, including C─C motif
   chemokine ligand (CCL)‐5, CCL‐7, and C‐X‐C chemokine motif ligand
   (CXCL)‐14.^[ [113]^24 ^] Also enriched was the chemokine CXCL13 which
   has been shown to regulate lymphocyte infiltration within the tumor
   microenvironment.^[ [114]^25 ^] Our findings are significant for
   several reasons. First, microfluidic organoids resemble the
   gold‐standard Matrigel organoids in expression of genes associated with
   pancreatic development, organoid formation, and maintenance. This
   observation is significant because Matrigel is a key component of stem
   cell or cancer organoid culures^[ [115]^26 ^] and yet it was used very
   sparingly (1% in media for 2 days during spheroid formation) in the
   microfluidic cultures. Second, upregulated expression of genes
   associated with chemotaxis of leukocytes in general and NK cells
   specifically supports our hypothesis that small volumes of microfluidic
   devices amplify autocrine/paracrine signaling and points to the
   benefits of using microfluidic organoid cultures for testing cellular
   immunotherapy such as NK cell therapy described later. We remind a
   reader that practical benefits of microfluidic devices over Matrigel
   cultures extend beyond gene expression and include 1) uniform
   distribution of spheroid sizes, 2) ease of tracking individual
   organoids over time, and 3) the ability to create co‐cultures with
   immune cells.

Figure 3.

   Figure 3
   [116]Open in a new tab

   Transcriptomic comparison of Matrigel‐based and microfluidic organoid
   cultures. A) Multidimensional scaling (MDS) plot comparing PDOs in
   microfluidic devices and Matrigel cultures, and with publicly available
   organoid sequencing data from primary PDAC tumors or P (n = 6), PDXs (n
   = 21), and metastases or M (n = 4). Five hundred top expressed genes
   from each sample were used to calculate pairwise distances between
   samples. B) Heatmap of sample‐to‐sample distances via Pearson
   correlation between microfluidic culture compared to Matrigel culture
   and to PDOs. C) Heatmaps of normalized gene expression counts (reads
   per kilobase per million mapped reads or rpkm) for Basal and classical
   PDAC subtype markers in PDOs cultured in devices versus Matrigel. D)
   Heatmaps of normalized gene expression counts (rpkm) for WNT, NOTCH,
   and TGF Beta signaling pathways. E) Volcano plot with genes
   significantly upregulated and downregulated between PDOs in devices
   compared to Matrigel culture highlighted. The horizontal line
   corresponds to ‐log[10](p‐value) with p‐value = 0.05 and the two
   vertical lines represent log[2](fold change) of 2 and −2. Hits were
   selected to have an ‐log[10](0.05) and effect size > 2 or < −2. F) Gene
   Ontology of biological processes enriched in the Matrigel culture and
   Microfluidic culture differentially expressed genes. G) Selected GSEA
   enrichment plots highlighting key pathways enriched in Matrigel and
   microfluidic culture platforms. Top 20 GSEA enrichment plots for
   Matrigel and microfluidic culture format may be found in
   Figures [117]S5,S6 (Supporting Information) respectively.

2.2. Culturing Intact Organoids in a Microfluidic Device

   As noted earlier, PDAC biopsies often have low cellularity and present
   a distinct challenge for organoid formation and expansion. In the case
   of two patient biopsies, we found it is difficult to generate organoids
   in numbers sufficient for a microfluidic device type 1 described in
   Figure [118]2. Given significant losses of cellular material during
   organoid dissociation and filtration to produce single‐cell suspension,
   we wanted to modify the design of the microfluidic device to introduce
   intact organoids or organoid fragments (≈100‐ to 150‐µm diameter). This
   novel device (see Figure [119] 4 A) had a side injection port protected
   by a normally closed valve. Negative pressure was applied to open the
   valve and a needle was introduced to transfer intact organoids directly
   into the culture chamber. We compared organoids from three different
   patients, denoted as PDO‐001, PDO‐002, and PDO‐003, in microfluidic
   devices and Matrigel cultures. Images of Matrigel cultures for all
   three PDOs are provided in Figure [120]4B. It may be appreciated that
   PDOs formed into uniform spheroids in the microfluidic device but were
   more randomly distributed in Matrigel cultures. In fact, no formation
   of distinct spheroids was observed for PDO‐002. Different patients were
   morphologically distinct. The organoids in PDO‐001 were compact, solid
   clusters with clear edges, whereas PDO‐003 had solid spherical
   structures with loose cells or debris accumulating around organoids
   over time. Cultures of PDO‐002 presented two distinct morphologies in
   Matrigel: spherical and elongated/stretched. Some of the organoids from
   PDO‐001 and ‐002 had luminal structures, while no such structures were
   observed for PDO‐003. Microfluidic organoid cultures exhibited features
   observed in Matrigel, with PDO‐001 and ‐002 exhibiting luminal/ductal
   structures (Figure [121]4C,D) and PDO‐003 forming compact spheroids
   (Figure [122]4E).

Figure 4.

   Figure 4
   [123]Open in a new tab

   Microfluidic cultures of intact PDAC organoids A) A novel microfluidic
   device with an injection port used for direct placement of intact
   organoids into the culture chamber. Inset shows intact organoids
   (PDO‐001) after 5 days of culture. B) Representative images of
   organoids in Matrigel derived from three different patients. C–E)
   Time‐course images of organoids in Matrigel and microfluidic cultures.
   Live/Dead staining was carried out for microfluidic organoid cultures
   at day 7. F–H) Immunofluorescence staining for E‐cadherin (Red) for all
   three organoid lines in Matrigel and microfluidic devices. Green and
   blue staining is for actin (phalloidin) and nuclei (DAPI) respectively.

   Next, we compared proliferation and phenotype expression for all three
   patient organoids in microfluidic and Matrigel cultures. As may be seen
   from Figure [124]4C to E, all three organoid types proliferated well in
   microfluidic devices albeit with different growth rates. These images
   also highlight challenges of characterizing organoids in Matrigel
   culture where individual clusters fused over time contributing to a
   wide distribution of organoid sizes. In contrast, organoids in
   microfluidic cultures had a much narrower size distribution. Growth
   rates and size distribution may be better appreciated in
   Figure [125]4C–E. The size changes over time for PDO‐001 cultured as
   intact spheroids in microfluidic devices were comparable to that of
   organoids in single‐cell devices (Figures [126]2C and [127]4C),
   demonstrating that these devices may be used interchangeably depending
   on cellularity and organoid expansion dynamics of individual tumor
   biopsies.

   Importantly, microfluidic organoid cultures for all three patients
   stained strongly for epithelial marker E‐cadherin, PDAC marker CK19,
   and proliferation marker Ki67 (Figure [128]4F–H and Figures [129]S7,S8,
   Supporting Information). Levels of expression for these markers were
   visually comparable to organoids in Matrigel. In summary, we
   demonstrated intact organoids derived from three different PDAC
   biopsies could be cultured in the novel microfluidic device with
   markers of proliferation and epithelial/cancer phenotype expression
   similar to the gold‐standard Matrigel organoid cultures.

2.3. Assessing Interactions of PDMS‐Based Microfluidic Devices with
Chemotherapy Drugs

   Our microfluidic devices are comprised of PDMS—material that has the
   potential to take up small hydrophobic molecules that are typically
   used as chemotherapy drugs.^[ [130]^27 ^] We wanted to assess the
   extent to which absorption of chemotherapy drugs occurred in our
   devices. To carry out this assessment, microfluidic devices were
   prepared in the way that was identical to cell culture experiments. The
   devices were functionalized with Pluronic (nonionic surfactant that
   imparts low‐binding property) and then incubated with serum‐containing
   media for 3 days. The latter step was designed to mimic 3‐day period
   allotted for organoid formation prior to drug testing. The devices did
   not contain cells to eliminate the possibility of drugs being taken up
   by cells and to focus squarely on PDMS uptake. Chemotherapy drugs
   (Encorafenib (En), Binimetinib (Bi), Gemcitabine (GEM), and GSKI) were
   prepared at concentrations ranging from 10 to 100 μM in media and were
   incubated in devices at 37 °C for 2 days. Identical protocol was
   followed for drug testing with PDOs. UV–Vis spectroscopy was used to
   construct absorbance versus concentration calibration curves for
   individual drugs (see Figure [131]S9A, Supporting Information) and to
   determine drug concentration. As highlighted by the data in
   Figure [132]S9B (Supporting Information), the uptake of chemotherapy
   drugs in PDMS devices was negligible. This observation may be
   attributed to a combination of factors including 1) hydrophilic
   functionalization of the surface with nonionic surfactant Pluronic and
   2) potential deposition of serum proteins from the media. Modest
   increases in concentration for En and Bi are likely due to media
   evaporation during 2 days of incubation in microfluidic devices. Thus,
   our analysis shows that the chemotherapy drugs of interest to this
   study were not taken up by the PDMS‐based microfluidic devices and that
   drug concentrations experienced by PDOs in the devices were expected to
   be similar to the original concentrations.

2.4. Testing Molecular Therapies Targeting RAF and MEK Kinases Using
Microfluidic Organoid Cultures

   After characterizing organoid cultures in a microfluidic device, we
   proceeded with testing chemotherapies. Exome sequencing of PDO‐001
   revealed the presence of CADPS2‐BRAF chromosomal rearrangement
   (mutation) (Table [133] 1 ).^[ [134]^28 ^] This mutation upregulates
   signaling along the RAS‐MEK‐ERK axis, promoting cancer cell
   proliferation, survival, and transformation.^[ [135]^29 ^] Based on
   these considerations, En and Bi, inhibitors of RAF and MEK
   respectively, were chosen as targeted therapies for treating PDO‐001.
   Responses of PDAC organoids to these molecularly targeted therapies
   were compared to GEM – a front‐line chemotherapy that inhibits DNA
   synthesis and leads to apoptosis of cancer cells (see Figure [136] 5 A
   for chemical structures).^[ [137]^30 ^] Five experimental groups were
   compared: 1) control group without drug, 2) En treatment, 3) Bi
   treatment, 4) combination treatment of En+Bi, and 5) GEM treatment.

Table 1.

   Clinical information for 3 PDAC patients.
   Patient Molecular profiling Stage Grade
   PDO‐001 CADPS2‐BRAF chromosomal rearrangement, TP53 (R282W), PTPN13
   (4258+2T>C), TMB (2.6), MSS (RNAseq: Overexpressed MET, EZH2, HRAS and
   NRAS, and Underexpressed CDKN2A) Stage IV Moderate to poorly
   differentiated
   PDO‐002 KRAS (G12R), TP53 (R175H), CNL in CDKN2A, CDKN2B, MTAP and
   SMAD4, TMB (2.6), MSS (RNAseq: MRC2‐PRKCA chromosomal rearrangement)
   Stage IV Moderately differentiated
   PDO‐003 KRAS (G12C), ATM (3712_3716del), TMB (4.69) Stage IV Moderately
   differentiated
   [138]Open in a new tab

   TMB: Tumor mutational burden

   MSS: Microsatellite stable

   CNL: Copy number loss

Figure 5.

   Figure 5
   [139]Open in a new tab

   Chemotherapy targeting MEK and RAF signaling tested in microfluidic
   cultures. A) (Left) Experimental timeline for seeding cells and
   administering drugs. (Right) Chemical structures of drugs used in this
   experiment: En, Bi, and GEM. B) Representative images of microfluidic
   organoid cultures at day 0 and day 7 (end of treatment) for each group.
   The fluorescence images represent Live/Dead staining at day 7 (Green:
   Live, Red: Dead). C) Relative drug response graph for four drug
   treatment groups after 7 days of treatment (viability %). D) Spheroid
   size changes at day 7 compared to day 0 for all drug treatment groups.
   Statistically significant differences compared to control condition (0
   μm) were *p < 0.05, ** p < 0.01, and *** p < 0.001. NS is
   nonsignificant statistically (p ≥ 0.05). E) In situ monitoring
   apoptosis in organoids using Caspase 3/7 green assay. Upper images are
   green fluorescence signal at 48 h and 72 h. Bottom graph is
   fluorescence mean intensity over time.

   Given the proliferative nature and relative abundance of cells
   available from PDO‐001, we seeded single cells into the microfluidic
   device type 1 (see Figure [140]2). As mentioned previously, single
   cells were given 2 days to form spheroids in a microfluidic device
   after which drugs were administered for 7 additional days (see
   Figure [141]5A for experimental workflow). Figure [142]5B shows
   brightfield and live/dead staining images of microfluidic organoid
   cultures before (Day 0) and after (Day 7) drug treatment. While
   organoids without therapy proliferated and retained high viability
   (≈96%) after 7 days of culture, different treatment groups and drug
   concentrations resulted in varying levels of cancer toxicity. As
   quantified in Figure [143]5C,D, organoids were more sensitive to Bi
   with ≈55% viability observed for 5 µM Bi, compared to En where loss in
   viability was minimal (≈90.14%) after exposure to 10 µM En. We next
   tested whether En and Bi have synergistic effect and found that 2.5 µM
   En in combination with 2.5 µM Bi had a similar effect on tumor
   regression as 5 µM Bi alone, but not for higher concentrations (see
   Figure [144]5B–D and Figure [145]S10A, Supporting Information). In
   addition to carrying out Live/Dead staining and monitoring spheroid
   size change, we also used Caspase‐3/7 apoptosis assay to assess
   cytotoxic effects of drugs. These results, summarized in Figure [146]5E
   and Figure [147]S11 (Supporting Information), support our observations
   of Bi being a more effective therapy compared to En as well as high
   sensitivity of these organoids to GEM (Figure [148]S10B, Supporting
   Information). The results described here illuminate a common clinical
   scenario where actionable mutations are identified by sequencing but
   functional responses to these drugs still need to be confirmed
   experimentally for a given patient. In our case, BRAF‐mutated
   patient‐specific organoids were significantly more sensitive to MEK
   inhibitor Bi compared to RAF inhibitor En.

   The response of BRAF‐mutated PDO‐001 was compared to BRAF wild‐type
   PDAC cells, PDO‐002 and PANC‐1 cell line.^[ [149]^31 ^] After 7 days of
   treatment with either En or Bi or the combination of the two drugs,
   PDO‐002 had only minimal loss in viability at concentrations reaching
   10 µM (Figure [150]S12A,B, Supporting Information). PANC‐1 spheroids
   were similarly unaffected by individual treatment with En or Bi but did
   exhibit a modest loss in viability (≈10%) and more pronounced decrease
   in proliferation for combination treatment at concentrations exceeding
   2.5 μM (Figure [151]S13, Supporting Information). In summary, we
   demonstrate that BRAF‐mutant PDAC organoids cultured in a microfluidic
   device responded to molecularly targeted therapy drugs En and Bi while
   BRAF‐wild type cells did not. This is another indication that
   microfluidic organoid cultures allow to accurately assess treatment
   responses.

2.5. Testing Efficacy of a Selective GSK‐3β Inhibitor in Combination with GEM

   We have previously shown that GSK‐3β is overexpressed in PDAC and that
   its knockdown or pharmacologic inhibition leads to decreased
   proliferation and induction of apoptosis.^[ [152]^32 ^] Furthermore, we
   used a GSKI, 9‐ING‐41, to demonstrate that GSK‐3β plays an essential
   role in sensitizing PDAC cell lines to GEM.^[ [153]^33 ^] As the next
   step, we wanted to assess the effects of GSKI and GEM using more
   physiological PDO cultures. We seeded intact organoids into the device
   type 2 as described in Section [154]2.2, allowed organoids to acclimate
   for 2 days and then introduced drugs.

   The timeline for administering combination chemotherapy is described in
   Figure [155] 6A. Prior to testing combination therapy, we performed
   dose‐response study for GEM alone for all three PDOs. As shown in
   Figure [156]S14 (Supporting Information), the IC50 values were 129 nM,
   67 nM, and 500 nM for PDO‐001, ‐002, and ‐003 respectively. These
   results revealed that GEM caused minimal cytotoxicity at concentrations
   of 10 nM and lower. We therefore chose to carry out synergy analysis
   using 10 nM GEM and varying concentrations of GSKI. As may be
   appreciated from Figure [157]6B–G, patient‐specific differences in
   combination treatment were observed. Limited cytotoxicity was observed
   in PDO‐001 organoids exposed to drugs (Figure [158]6B,C and
   Figure [159]S15A, Supporting Information) whereas PDO‐002 and ‐003 did
   show enhanced response to combination treatment in all concentration
   ranges (Figure [160]6D–G and Figure [161]S15B,C, Supporting
   Information). Synergism scores, which indicate synergistic effect when
   less than 1, were calculated using Calcusyn software and were
   determined to be >2.193 for PDO‐001, 0.347 for PDO‐002, and 0.178 for
   PDO‐003, in all concentrations.^[ [162]^33 , [163]^34 ^]

Figure 6.

   Figure 6
   [164]Open in a new tab

   Combination therapy of GSKI and GEM testing using three
   patient‐specific microfluidic organoid cultures. A) (Left) Experimental
   timeline for cell seeding and drug treatment. (Right) Chemical
   structures and molecular weight of 9‐ING‐41 (GSKI) and GEM. B, D, and
   F) Images of organoids after 1‐day or 7‐day treatment with individual
   drugs or combination of drugs. Live/Dead staining performed on day 7 to
   quantify response to therapy. C, E, and G) Relative drug response of
   single (GSKI) and dual (GSKI+GEM) treatment quantified based on
   Live/Dead staining at day 7 (Viability %). H) Relative change in
   organoid size after treatment. Area of organoids at day 7 of treatment
   was normalized to day 1 (before treatment). Statistical significance of
   combination treatment (GSKI+GEM) when compared to GSKI treatment alone
   (NS p > 0.05, * p < 0.05, ** p < 0.01, and *** p < 0.001).

   The results described in Figure [165]6 are exciting for several
   reasons. First, to the best of our knowledge, this is the first
   assessment of a novel GSKI using PDOs. Second, responses of organoids
   parallel previous pre‐clinical and clinical observations suggesting
   that pre‐treatment with GSK‐3β inhibitor sensitizes patients with PDAC
   to front‐line therapy of GEM.^[ [166]^35 ^] Third, we observed
   patient‐to‐patient variability of organoid responses to combination
   therapy. Lack of response in PDO‐001 may be attributed to BRAF mutation
   and MET overexpression which have been reported to decrease sensitivity
   to GSKI by activating nuclear factor kappa‐B (NF‐κB) pathway downstream
   of GSK‐3β.^[ [167]^29 , [168]^35 , [169]^36 ^] A modest synergistic
   effect in PDO‐002 and ‐003 may be attributed to mutation of KRAS‐G12
   (Table [170]1), which has been shown to increase GSK‐3β gene expression
   and render cancers resistant to GSKI treatment.^[ [171]^37 ^] We note
   that the novel GSKI tested here is not being used clinically and
   therefore it was not possible to correlate response of microfluidic
   organoids in patients. However, evidence presented in our study—1)
   selective response of BRAF mutant PDO‐001 but not BRAF wild‐type
   PDO‐002 to MEK and RAF inhibitor and 2) difference in responses to GSKI
   based on molecular profile of each PDO—point to excellent utility of
   microfluidic organoid cultures for assessing molecularly targeted
   therapy.

2.6. Assessing Combination Cancer Immunotherapy of NK Cells and TriKE
Immunomodulators Using Microfluidic Organoid Cultures

   Microfluidic devices allow us to position cancer organoids in an array
   format and maintain their phenotype for at least 9 days while
   monitoring changes in dimensions of individual organoids. These
   capabilities make our platform particularly useful for testing
   immunotherapies by quantifying interactions between immune cells and
   cancer organoids. Cancer immunotherapy is a novel type of treatment
   that may entail stimulation of patient's own immune cells to enhance
   anti‐tumor activity or an adoptive transfer of immune cells.^[ [172]^38
   ^] While immune checkpoint inhibitor and CAR T‐cell therapies have been
   used more extensively,^[ [173]^39 ^] NK cells are emerging as an
   exciting alternative immunotherapy strategy because they can attack
   tumors without prior activation or sensitization.

   Unlike T cells that are restricted to recognizing specific human
   leukocyte antigen (HLA) class I in the context of antigen peptide, NK
   cells are not HLA restricted. This means that NK cells need not be
   autologous and are increasingly seen as a potential “off‐the‐shelf”
   product to be used in the allogenic setting for cancer immunotherapy.^[
   [174]^40 ^]

   Anti‐tumor activity of NK cells may be further enhanced by using
   therapeutics, such as Tri‐specific Killer Engagers (TriKEs).^[ [175]^41
   ^] TriKEs are engineered immunomodulators that create a synapse between
   immune cells and cancer cells, mediating tumor suppression.^[ [176]^42
   ^] Several recent reports highlight the therapeutic potential of TriKEs
   targeting immune checkpoint B7‐H3 on cancer cells.^[ [177]^42 ,
   [178]^43 ^] Vallera et al. developed a TriKE that is formed by a single
   domain nanobody specific for B7‐H3 linked to recombinant IL‐15, and a
   scFv that binds to the activating receptor CD16 on NK cells.^[ [179]^42
   ^] B7‐H3 (or CD276) is a surface marker expressed in several cancer
   types including PDAC,^[ [180]^43b,c ^] while CD16 receptor triggers
   antibody‐directed cell‐mediated cytotoxicity (ADCC) in NK cells.^[
   [181]^41 , [182]^44 ^] While described for the first time 2 years ago,
   anti‐B7‐H3 TriKE has not yet been tested with patient‐derived cancer
   organoids. Thus, we employed microfluidic cultures of PDAC organoids to
   test effects of NK cells and anti‐B7‐H3 TriKE.

   Three treatment groups were compared: (1) a control group consisting of
   cancer organoids only, (2) a co‐culture group of NK cells and cancer
   organoids, and (3) a group where cancer organoids were co‐cultured with
   NK cells in the presence of the anti‐B7‐H3 TriKE. Cancer toxicity in
   microfluidic devices type 2 (direct organoid seeding device) was
   monitored over the course of 72 h to assess therapeutic efficacy. The
   experimental timeline is shown in Figure [183] 7 A and a schematic of
   the mechanism by which anti‐B7‐H3 TriKE enhances killing of tumor cells
   by NK cells is shown in Figure [184]7B. We supplemented the organoid
   media with IL‐2 for NK cell maintenance and added SYTOX orange to
   visualize dead cells.

Figure 7.

   Figure 7
   [185]Open in a new tab

   Testing NK cell‐based immunotherapy using microfluidic organoid
   cultures. A) Timeline and workflow of immunotherapy testing. “D”
   denotes the day at which each step was completed. B) Schematic
   describing structure and function of immunomodulator B7‐H3 TriKE. This
   molecule contains two Ab fragments, anti‐CD16 targeting NK cell marker,
   and anti‐B7‐H3 specific to immune checkpoint on cancer cells, as well
   as recombinant IL‐15 to stimulate NK cells. (C) Representative images
   of microfluidic organoid cultures with and without NK cells/ B7‐H3
   TriKEs at day 0 and day 3 of culture. Scale bars = 500 µm. Organoids
   from three patients were tested. D) Normalized changes in organoid
   diameter during 3 days of treatment (NS p > 0.05 and * p < 0.05).

   For all 3 PDOs, the monocultures of cancer organoids showed sustained
   growth inside microfluidic devices over 3 days of culture
   (Figure [186]7C, 72 h). The observed growth was on par with our control
   experiments during chemotherapy testing, suggesting that addition of
   IL‐2 and SYTOX orange to the culture media did not adversely affect
   organoid proliferation. To assess organoid growth in a quantitative
   manner, we tracked the diameter of individual organoids every 24 h,
   normalized the diameter at 24, 48, and 72 h to the initial size at 0 h
   and plotted the average organoid diameter per condition for all PDAC
   samples (Figure [187]7D). In the absence of B7‐H3 TriKE, NK cells
   showed a sustained killing of cancer cells, decreasing the size of the
   cancer organoids by 35–45% at 24 h, and by 85% at 72 h for PDO‐002,
   with PDO‐001 and ‐003 decreasing in size by 70% at 72 h. In comparison,
   the presence of the anti‐B7‐H3 TriKE increased cytotoxicity in all
   PDOs. All cancer organoids showed a decrease in size of ≈60% at 24 h, a
   15–25% improvement compared to non‐TriKE conditions, with a 90–95% size
   decrease for all PDOs (Figure [188]7D) at 72 h timepoint. Our results
   suggest that anti‐B7‐H3 TriKE improved cytolytic activity of NK cells
   toward cancer organoids. These results also highlight utility of
   microfluidic cancer organoid cultures for testing immunotherapies, in
   addition to the chemotherapy studies discussed above.

   As noted above, we and others have shown that culturing cells in a
   microfluidic device in the absence of convection enhances autocrine and
   paracrine signaling compared to standard (large volume) cultures. We
   therefore hypothesized that signaling of such inflammatory cytokines as
   interferon‐gamma (IFN‐γ) may be enhanced in the microfluidic device. To
   test this hypothesis, we compared PDO‐NK cell co‐cultures in the
   microfluidic device (type 1) and in the microwell array described in
   Figure [189]2F. Immunofluorescence staining for STAT3, an element in
   the interferon response pathway, revealed higher levels of
   phosphorylation (pSTAT3) in the microfluidic device compared to
   microwell array after 2 days of culture (Figure [190]S16, Supporting
   Information). In addition, more NK cells were observed to infiltrate
   PDOs in the microfluidic device. These results highlight an additional
   advantage of microfluidic cultures for testing immunotherapy—amplified
   autocrine/paracrine inflammatory signaling. Interestingly, the
   observation of enhanced infiltration of NK cells in the microfluidic
   device is aligned with sequencing results pointing to the upregulated
   expression of CCL5, CCL7, and CXCL14—cytokines associated with
   chemotaxis of NK cells.^[ [191]^24 ^]

3. Conclusion

   In this study, we developed and characterized novel microfluidic
   cultures of PDAC organoids. We demonstrated microfluidic devices to be
   superior to microwell array format in terms of organoid proliferation
   and cancer phenotype expression as well as in activation and
   infiltration of immune cells. RNA sequencing analysis revealed
   microfluidic organoid cultures to be similar to the gold‐standard
   Matrigel cultures, with key differences of better oxygenation and
   priming for chemokine signaling and immune cells infiltration in the
   microfluidic format. Compared to Matrigel cultures, microfluidic
   organoids were advantageous for therapy testing because of 1) uniformly
   sized organoids that may be easily monitored over time to assess
   response to therapy, 2) the ability to introduce immune cells for
   testing immunotherapy, and 3) minimal reliance on Matrigel for organoid
   maintenance. Importantly, microfluidic organoids accurately responded
   to molecularly targeted therapy. BRAF‐mutant PDOs were shown to be
   responsive to MEK and RAF inhibitors while BRAF wild‐type PDOs were
   resistant. Our study also confirmed previous pre‐clinical and clinical
   observations that GSKI may be used to sensitize tumor to front‐line
   therapy of GEM. Importantly, we evaluated a novel, recently developed
   GSKI, 9‐ING‐41, that has not yet been tested with PDOs. In addition to
   testing chemotherapy or molecularly targeted therapy, our platform is
   also well‐suited for testing immunotherapy. To demonstrate this point,
   we assessed a novel immunotherapy comprised of NK cells and an
   immunomodulatory biologic, TriKE. We demonstrated that cytolytic
   activity of NK cells may be further enhanced by the presence of TriKE.

   In summary, we described a novel microfluidic organoid culture platform
   for testing patient‐specific responses. This platform helps maintain
   cancer organoids in a manner similar to gold‐standard Matrigel
   cultures, enables drug testing with minimal amount of organoids
   available from a needle biopsy and is well‐suited for testing
   immunotherapies. In the future, our microfluidic platform may be
   integrated with biosensing capabilities for quantitative assessment of
   individual organoids and may be developed into a companion diagnostic
   for personalized cancer therapy.

4. Experimental Section

Materials

   Advanced DMEM/F‐12, DMEM/F‐12 no phenol red,
   4‐(2‐hydroxyethyl)−1‐piperazineethanesulfonic acid (HEPES), Glutamax,
   penicillin/streptomycin (P/S), Fetal bovine serum (FBS), RPMI 1640,
   LIVE/DEAD viability/cytotoxicity Kit for mammalian cells, Goat
   anti‐Mouse IgG (H+L) Cross‐Adsorbed Secondary Antibody Alexa Fluor 546,
   Monoclonal E‐cadherin Mouse Antibody, and CellEvent Caspase‐3/7 Green
   Detection Reagent were purchased from Thermo Fisher Scientific
   (Waltham, MA, USA). N2 supplement, B27 supplement, epidermal growth
   factor (EGF), fibroblast growth factor 10 (FGF10), Cultrex PathClear
   Reduced Growth Factor BME, Type 2 (Matrigel), and Human Cytokeratin 19
   Antibody were obtained from R&D systems (Minneapolis, MN, USA).
   SB202190 (p38 MAP) and recombinant human IL‐2 were acquired from
   Peprotech Inc. (Rochy Hill, NJ, USA). N‐acetyl‐L‐cysteine (NAC),
   Nicotinamide, Gastrin I human, Collagenase type IV, Pluronic F‐127,
   CryoStor cell cryopreservation media, and Bovine serum albumin (BSA)
   were purchased from Sigma‐Aldrich (St. Louis, MO, USA). Gentle cell
   dissociation reagent (GCDR) and Rosette Separation NK cell isolation
   kit were purchased from STEMCELL Technologies (Vancouver, Canada).
   Gemcitabine solution (38 mg mL^−1) was purchased from the Mayo Clinic
   pharmacy and Y‐27632, Encorafenib, Binimetinib, and 9‐ING‐41 (GSK‐3β
   inhibitor; GSKI) were acquired from MedChemExpress (Monmouth Junction,
   NJ, USA). Monoclonal Ki‐67 Mouse Antibody, Phalloidin‐iFluor 488, and
   Phospho‐STAT3 Rabbit Antibody were purchased Cell Signaling Technology
   (Danvers, MA, USA) and Abcam (Cambridge, UK), respectively. STAT3
   Monoclonal antibody was acquired from Proteintech (Rosemont, IL, USA).
   The PANC‐1 and L‐WRN cell lines were purchased from the ATCC (Manassas,
   VA, USA).

   For PDAC organoid media, conditioned media from L‐WRN is prepared.
   L‐WRN conditioned media (Wnt‐3A, R‐spondin 3, and noggin conditioned
   medium) was collected by following previously reported protocol^[
   [192]^10 , [193]^45 ^] and stored at −30 °C prior to the preparation of
   organoid media. The organoid media for organoid culture was based on
   50% Advance DMEM/F‐12 and 50% L‐WRN conditioned media, supplemented
   with 1× HEPES, 1× Glutamax, 1× N2 supplement, 1× B27 supplement,
   500 ng mL^−1 EGF, 1 µg mL^−1 FGF10, 3 µM SB202190, 0.5 µM A83‐01, 1 mM
   NAC, 10.25 mM Nicotinamide, 10 nM Gastrin, 1× P/S (100 U mL^−1
   penicillin and 100 µg mL^−1 streptomycin), and 10 µM Y‐27632.

Collection of PDAC Biopsies

   The patient PDAC biopsy collection was approved by the Mayo Clinic
   under IRB #17‐003174 and were carried out with the full, informed
   consent of the subjects. The samples were collected by
   ultrasound‐guided biopsies using an 18‐gauge needle and were ≈1 mm
   thick × 10–20 mm long. The procedure was performed at the Mayo Clinic
   in Rochester, MN. Two to four needle biopsies were made available for
   research from each biopsy collection. The needle biopsies collected
   from three different patients were shown in Figure [194]S2 (Supporting
   Information).

Organoid Formation and Culture in Matrigel Domes

   Needle core biopsies were placed in the tissue storage solution on ice
   in the procedure room and delivered to the research lab within 30 min
   of collection. The workflow from biopsy digestion to seeding steps is
   illustrated in Figure [195]S1 (Supporting Information). Fresh biopsies
   were placed on a 6 mm‐culture dish with Krebs‐Ringer Bicarbonate (KRB)
   buffer on ice and minced to fragments of <1 mm in length and width
   using a sterile disposable blade. The fragments were digested with
   2.5 mg mL^−1 collagenase type IV under mild agitation in a water bath
   (37 °C) for 5 min. The dissociated supernatant was filtered through a
   100 µm cell strainer to remove ECM, and then centrifuged at 1500 rpm
   for 5 min. To remove red blood cells (RBCs), the dissociated cells were
   immersed in RBC lysis buffer for 3 min at 25 °C. The effect of the
   lysis buffer was stopped by adding 10% FBS‐containing PBS. Dissociated
   cells were then seeded onto Matrigel domes in a 12‐well plate using
   organoid media. For passaging organoids, Matrigel domes were broken by
   treating them with GCDR for 1 min, followed by organoids collection.
   The organoids were resuspended in 50% Matrigel/50% organoid media and
   dispersed onto a 12‐well plate to make Matrigel domes. After gelation
   of the Matrigel domes, organoid media was added carefully. The media
   was changed every 3–5 days. The organoids from patient 1 (PDO‐001) were
   passaged 10–20 times while PDO‐002 and ‐003 were passaged five to seven
   times due to the reduced proliferation by successive passages of
   PDO‐002 and ‐003 after passage No. 8. For later use, the PDOs were
   cryopreserved using CryoStor following manufacturer's instructions.

Fabrication of Microfluidic Devices

   Two types of microfluidic devices (see Figures [196]2A and [197]4A)
   were fabricated using standard soft lithography techniques.^[ [198]^17
   , [199]^46 ^] Device 1 was designed to organize single cells into
   organoids whereas device 2 was used to seed intact organoids. The
   detailed protocol for fabricating molds and PDMS devices is described
   in the Supplementary Information.

   Before use, microfluidic devices were degassed in a vacuum desiccator
   and treated overnight with 1% Pluronic F‐127 in 1× PBS. This treatment
   minimizes cell attachment to the surface of PDMS and promotes cell
   aggregation into spheroids.^[ [200]^47 ^] The devices were then washed
   with 1× PBS to remove the Pluronic solution and exposed to UV light for
   30 min for sterilization. Prior to cell seeding, PBS solution was
   removed from media reservoirs.

Assessing Interactions of Chemotherapy Drugs with PDMS‐Based Microfluidic
Devices

   The concentrations of each drug (En, Bi, GSKI, and GEM) after
   incubation in microfluidic device were quantified by UV–Vis
   spectroscope (NanoDrop One‐C; Thermo Fisher Scientific). Four different
   drugs were dissolved in 10% FBS‐containing phenol red‐free DMEM/F‐12
   media at 10, 50, and 100 μM, excepting for 100 mM GSKI due to the
   severe sedimentation during incubation. Then, the drug solutions were
   incubated in the microfluidic devices and stored at 37 °C incubator for
   2 days. The solution was recirculated every day. To mimic chemotherapy
   testing with PDOs, the device was pretreated with Pluronic F‐127 for 1
   day and 10% FBS‐containing phenol red‐free DMEM/F‐12 media for 2 days
   before adding drugs. After 2 days of incubation, the drug solutions
   were collected and analyzed by UV–Vis spectroscopy. Calibration curves
   were created by measuring absorbance values for drug concentrations
   ranging from 0 to 1250 μM, unknown samples from drugs after incubation
   were then determined using these calibration curves.

Characterization of PDAC Organoids by Immunofluorescence and RT‐PCR

   Size (area) changes of PDAC organoids in Matrigel or wells were
   measured using ImageJ software. For immunostaining, whole PDAC
   organoids were fixed and permeabilized with 4% Paraformaldehyde (PFA;
   Electron Microscopy Sciences) and 0.1% Triton X‐100 (Sigma‐Aldrich) for
   30 min, respectively, and blocked with 2% BSA for 1.5 h at 25 °C. Then,
   the primary antibodies and fluorophore‐conjugated secondary antibodies
   with DAPI (BD Biosciences) and Phalloidin‐Alexa 488 in 1% BSA solution
   were sequentially treated for 1–1.5 h, at 25 °C. For staining in
   microwells, 200 µL of each solution was dispensed into the media
   reservoirs and recirculated every 15 min. the organoids were imaged
   using confocal microscopy (LSM 780; Carl Zeiss, Germany).

   For RT‐PCR analysis, PDO‐001 was cultured in microwells format and
   microfluidic device for 5 days. Then, total RNA was isolated using a
   commercial kit (RNeasy kit, Qiagen (Valencia, CA, USA)) following the
   manufacture's protocol. Approximately 20–30 ng of total extracted RNA
   was used for synthesis of cDNA, following the reverse transcription kit
   instructions (Roche). The primer sequences used for RT‐PCR were listed
   in Table [201]S2 (Supporting Information). Gene expression was tested
   using the QuantStudio 5 System (Thermofisher Scientific) with SYBR
   Green qPCR Master Mix and was normalized to glyceraldehyde 3‐phosphate
   dehydrogenase (GAPDH). The amplification procedure for Real‐time PCR
   consists of 40 cycles of denaturation at 95 °C for 5 s, annealing at
   55 °C for 15 s, and extension at 69 °C for 20 s. The final evaluation
   was carried out based on the threshold cycles using the ΔΔCT method and
   normalized to expression level of PDAC organoids in Matrigel.

Preparation of PDAC Organoids for Seeding into Microfluidic Devices

   Organoids embedded in Matrigel were expanded for 1–1.5 weeks. Then
   Matrigel domes were dissociated by treatment with GCDR and gentle
   pipetting. This digest was centrifuged at 1500 rpm for 4 min at 4 °C
   conditions to remove Matrigel debris and washed with cold DMEM for
   removing residual Matrigel thoroughly. At this point in the protocol,
   we produced intact organoids. These organoids were either seeded intact
   into a microfluidic device 2 or were broken down into single cells.
   When dissociating organoids into single cells, digest was exposed to
   2 mL of 0.25% Trypsin‐EDTA (Gibco) for 4 min at 37 °C (in water bath)
   under mild agitation after which DMEM with 10% FBS was added to stop
   digestion. Clumps of cells or debris were filtered using 100 µm
   strainer prior to seeding.

Forming and Maintaining Organoids in Microfluidic Devices

   When seeding into the microfluidic device 1, single cell digest was
   resuspended to a concentration of 4 × 10^6 cell·mL^−1 in organoid media
   with 1% DNase (see Section 4.1. for media description). Then, 100 µL of
   cell suspension was placed in one of the media reservoirs to create a
   difference in hydrostatic pressure and drive cells into the device.
   Positive pressure was applied to one of the reservoirs, by covering the
   top with a finger and pushing gently, to increase the flow inside the
   chamber and equilibrating media levels between the cylinders. At this
   point, the flow in the device stops and cells settle down into
   microwells. The microfluidic devices were washed twice with fresh media
   to remove cells that did not land in microwells. Organoid media with 1%
   Matrigel was added, and devices were moved to an incubator (at 37 °C,
   5% CO[2]). The media was recirculated after 24 h of culture by
   aspirating 50 µL from one media reservoir and dispensing it into
   another reservoir. Organoid formation was typically observed after
   48 h. From this point onwards, organoids were cultured in media without
   Matrigel.

   The microfluidic device type 2 contained a side injection port
   protected by the normally closed valve to allow opening the port,
   introduce intact organoids directly into the culture chamber (see
   Figure [202]4A), then reseal the port and use the device as a standard
   petri dish without connection lines. Operation of a device to open the
   valve was described recently.^[ [203]^46 ^] Briefly, a device was
   connected to a vacuum line and placed on a microscope in a biosafety
   hood. The valve was opened by negative pressure to allow insertion of
   27‐gauge needle and transfer of organoids into the culture chamber
   containing in microwells. Each device was typically populated with 13
   organoids. These organoids settled down into microwells after 20 s
   under static conditions. Once organoids were confined to microwells,
   the vacuum line was disconnected, and valve/injection port closed. No
   external sources of pressure were needed to keep this normally closed
   valve sealed during a multi‐day experiment. Microfluidic organoids were
   cultured in organoid media supplemented with 1% Matrigel for the first
   48 h. No Matrigel was used in media thereafter.

Bulk RNA Sequencing

   Organoids were cultured in Matrigel as described in Section 4.3 and
   seeded into microfluidic devices as described in Section 4.7. Three
   sample replicates from one patient for each type of organoid culture
   platform (Device and Matrigel) were processed. Culture media was
   removed and organoids were washed with PBS. Cells were lysed and
   homogenized for RNA extraction following the Qiazol‐based miRNeasy
   Micro Kit (Qiagen). Isolated RNA was assessed for quality using RNA
   fragment analysis and all samples with RNA Integrity Scores (RIN) >7
   were processed. RNA samples were cleaned up to ensure any contaminating
   salts and solvents were removed. A standard input stranded mRNA library
   prepration was used to focus on the poly‐adenylated transcripts.
   Subsequently, bulk RNA sequencing data was obtained using paired‐end
   next‐generation sequencing IlluminaNovaSeqSP (100 cycles).

Differential Gene Expression Analysis

   RNA‐seq FASTQ file pairs were trimmed using TrimGalore version 0.6.4
   and Cutadapt version 4.4. FASTQC files from trimming were used to
   assess RNA quality. Subsequently, trimmed FASTQ files were aligned
   using STAR version 2.7.1a to the human reference genome GRCh38. Counts
   for each gene were obtained from STAR output and differential gene
   expression analysis was performed using the EdgeR version 3.40.2 R
   package. EdgeR allowed for assessment of similarity between samples
   using hierarchical clustering of sample distances and principal
   component analysis to determine if the Matrigel and Device groups had
   significantly different overall gene expression. Heatmaps were created
   using the R package pheatmap version 1.0.12 and the volcano plot was
   generated using ggplot2 version 3.4.3. Genes with a log[2](fold change)
   >2 or <−2, and an adjusted p‐value of <0.05 were considered
   significantly differentially expressed.

Pathway Enrichment Analysis and Gene Set Enrichment Analysis

   The R package ClusterProfiler version 4.6.2 was used to identify Gene
   Ontology (GO) biological processes enriched in the significantly
   upregulated and downregulated genes. Gene set enrichment analysis (GSEA
   version 4.3.2) was performed with default settings, specifically, 1000
   permutations using the c2.all.v2023.1.Hs.symbols.gmt gene set database
   and the Human_Ensembl_Gene_ID_MSigDB.v2023.1.Hs.chip Chip platform
   ([204]http://software.broadinstitute.org/gsea/index.jsp). Gene sets for
   NOTCH (KEGG_NOTCH_SIGNALING_PATHWAY M7946), WNT
   (KEGG_WNT_SIGNALING_PATHWAY [205]M19428), and TGF BETA
   (KEGG_TGF_BETA_SIGNALING_PATHWAY M2642) were obtained from the GSEA
   Molecular Signatures Database.

Additional Pancreatic Cancer Datasets

   Publicly available Bulk RNA sequencing data of patient‐derived
   organoids derived from Pancreatic Ductal Adenocarcinoma were retrieved
   from the NCBI SRA Run Selector under BioProject accession number
   PRJNA873279 ([206]https://www.ncbi. nlm.nih.gov/geo/). Fastq files were
   trimmed, aligned, and differentially expressed genes were analyzed as
   described in Section 4.10.

Administering Chemotherapy to Microfluidic Organoid Cultures

   Chemotherapy testing was carried out in the following manner. First,
   all the drugs (with the exception of Gemcitabine) were dissolved in
   DMSO to create 10 mM drug stock solution. Then, chemotherapy drugs were
   dissolved in organoid media at concentrations ranging from 0.1 nM to 10
   µM. DMSO concentration ranged from 10^−7 to 0.1% for this concentration
   range. Previous studies have shown that DMSO is nontoxic when used at
   concentrations below 0.5%.^[ [207]^48 ^]

   The same concentration was used for three treatment arms: 1)
   Encorafenib (En), 2) Binimetinib (Bi), and 3) Gemcitabine (GEM). Media
   with drug was added at the 48 h time point once organoids had formed
   and/or have acclimated to culture conditions. Drug treatment lasted for
   7 days with fresh media (containing drugs) changed every 48 h. Media in
   the devices was recirculated using the method described above where
   50 µL from one media reservoir was collected and placed into the other
   reservoir. Images of microfluidic organoid cultures were acquired every
   2 days using an inverted fluorescence microscope (BZ‐X800, Keyence). On
   day 7 of treatment, the media was removed, and organoids were stained
   using a Live/Dead staining kit following the manufacture's protocol.
   The viability was calculated following the equation: (Viability% = 100
   × A[green] / (A[green] + A[red])) with area measurements performed
   using ImageJ software.

   For apoptosis assay, 2 µM of Caspase‐3/7 reagent solution in organoid
   media was applied to each group after 2 days of seeding. The media with
   reagent was recirculated every day and changed every 2 days. Apoptotic
   cells emitted green fluorescence and were imaged using a fluorescence
   microscope (BZ‐X800, Keyence). The fluorescence intensity was
   calculated using the ImageJ software.

   When testing a combination of GSKI and GEM therapy, different
   concentrations of individual drugs or a combination of drugs were
   dissolved in organoid media and added to each microfluidic device at
   day 2 after seeding. Cultivation of organoids and assessment of drug
   response was carried out in the same manner as described in the
   previous paragraph.

PANC‐1 Cell Culture and BRAF‐Targeted Chemotherapy Testing in the
Microfluidic Device

   The PANC‐1 cells were cultured at 37 °C incubator in 75 T flask until
   the confluency reaches ≈95%. The medium for PANC‐1 is 90% high glucose
   DMEM, 10% FBS, and 1% P/S. The culture media was changed every 2–3
   days. To perform BRAF‐targeted chemotherapy testing, the PANC‐1 cells
   were washed with 1x PBS and treated TripLE at incubator for 3 min to
   detach cells from the flask. Sequentially, cells were collected by
   centrifugation at 1500 rpm for 4 min and seeded them into the
   microfluidic device type 1 (single cell seeding device). The process
   for seeding cells and chemotherapy testing followed previously
   described in Sections 4.8 and 4.13.

Isolation and Expansion of NK Cells

   Primary human NK cells were isolated from apheresis cones from
   anonymized healthy donors obtained from the Mayo Clinic Blood Bank
   (Rochester, MN, USA) as previously described.^[ [208]^49 ^] Briefly,
   human peripheral blood mononuclear cells (PBMCs) were isolated by
   Ficoll‐Hypaque (GE Healthcare, Chicago, IL, USA) density gradient
   centrifugation. The resulting PBMCs were then mixed with RBCs at a
   ratio of 1:100 and incubated with antibodies from Rosette Separation NK
   cell isolation kit for 20 min at room temperature as per the
   manufacturer instructions. Ficoll‐Hypaque density gradient
   centrifugation was then repeated, and the resulting NK cells were
   analyzed for purity by flow cytometry using anti‐CD3 and anti‐CD56
   (BioLegend, San Diego, CA, USA) and Ghost Dye (Tonbo, San Diego, CA,
   USA).

   Primary human NK cells were expanded using the modified mbIL‐21 K562
   line, CSTX002, which was kindly supplied by Dean Lee (Ohio State
   University). Expansion was accomplished as described in Somanchi
   et al.^[ [209]^44 ^] Briefly, mbIL‐21 K562 cells were irradiated at
   100 Gy and mixed with NK cells at a E:T ratio of 1:2 for 7 days with
   fresh medium and IL‐2 supplied on days 3, 5, and 7. This cycle of
   expansion was repeated three times for a total of 21 days. NK cells
   were expanded and maintained in RPMI 1640 supplemented with 50 U mL^−1
   recombinant human IL‐2, 10% FBS, and 1% each of Pen Strep, sodium
   pyruvate, MEM nonessential amino acids, and L‐glutamine from Corning
   (Corning, NY, USA).

   The Tri‐Killer Engager (TriKE) molecule contained an anti‐CD16 antibody
   fragment, an IL‐15 moiety, and an anti‐B7‐H3 single‐chain variable
   fragment (scFv). Detailed description of this biologic is provided in a
   recent paper.^[ [210]^42 ^]

Testing NK Cell‐Based Immunotherapy Using Microfluidic Organoid Culture

   Organoids were seeded into microfluidic devices as described in Section
   4.7. Three experimental conditions were tested: (1) cancer organoids,
   (2) cancer organoids + NK cells, and (3) cancer organoids + NK cells +
   TriKE treatment. When working with experimental groups 2 and 3 using NK
   cells, the following protocol was followed. PDAC cells were given ≈48 h
   in microfluidic devices for organoid formation before adding NK cells.
   Immune cells were labeled with Vibrant DiD cell tracker dye and
   resuspended at a density of 10×10^6 cell mL^−1. A 100 µL of cell
   suspension was infused into each microfluidic device. Excess cells were
   washed out, and all devices were filled with fresh organoid media
   supplemented with 50 U mL^−1 recombinant human IL‐2, and SYTOX orange
   at a concentration of 5 µM. For experimental group 3, media was
   supplemented with B7‐H3 TriKE at 3 nM. This concentration was shown to
   be effective previously. Three microfluidic devices were used for each
   experimental group. The media was recirculated every 24 h by gently
   collecting contents of one media reservoir with a pipette and placing
   these contents into the opposite reservoir. Images of the organoids in
   microwells were acquired every 24 h using an inverted fluorescence
   microscope. The protocol described above was used for all three patient
   samples/PDOs.

Analysis of Cancer Organoid Killing by NK Cells and STAT3 Activation in NK
Cells

   For immunotherapy studies, images were acquired with a fluorescence
   microscope using a 10× objective. A z‐stack was acquired for each
   microfluidic chamber, obtaining up to 15 slices with a 10 µm pitch,
   covering all 19 microwells in every chamber. Images were acquired in
   brightfield, GFP channel for GFP‐transfected PDAC (PDO‐001), TRITC
   channel for SYTOX orange staining of dead cells, and Cy5 channel for
   DiD labeling of NK cells. For non‐GFP‐expressing cancer cells,
   Calcein‐AM (viability marker) was added at the end of the experiment
   (72 h) to help visualize live organoids. All the images acquired for
   each microfluidic chamber were processed to obtain a single full‐focus
   image per channel, by using the microscope software.

   Image analysis was done using Fiji software.^[ [211]^50 ^] The area for
   each organoid in all images was tracked over time for every tested
   condition and PDAC sample. The diameter of each organoid at 24, 48, and
   72 h were normalized against 0 h to determine organoid growth or
   shrinking. For immunofluorescence staining of STAT3 and p‐STAT3, NK
   cells and PDO‐001 were co‐cultured in microfluidic device (single cell
   seeding device) and PDMS microwell for 1–3 days and stained for
   expression of STAT3 and p‐STAT3 as described above. ImageJ was used to
   quantify fluorescence due to p‐STAT3 and STAT3 staining.

Statistical Analysis

   T‐test was performed for all data sets to determine significance
   between conditions for immunotherapy data. The significance was denoted
   as *, and NS, which means p < 0.05, and nonsignificant (p > 0.05),
   respectively. All the organoids in the three chambers per condition
   were averaged (n ≥ 40). Error bars denote standard deviation. One‐way
   ANOVA analysis followed by Tukey's post hoc was used to analyze
   significance when comparing more than two groups. Origin and GraphPad
   Prism software were used to plot and statistically analyze the data.

Conflict of Interest

   The authors declare no conflict of interest.

Supporting information

   Supporting Information
   [212]Click here for additional data file.^ (3MB, pdf)

   Supplemental Table 1
   [213]Click here for additional data file.^ (1MB, csv)

Acknowledgements