Bi-Phenotypic Differentiation of Pancreatic Cancer in 3D Culture

Abstract

Objectives:

Pancreatic ductal adenocarcinoma (PDAC) is the third most common cause of cancer death in the US. Improved characterized models of PDAC are needed for drug screening.

Methods:

We grew four established pancreatic cancer cell lines in hanging drop cultures to produce spheroids. We also grew organoids from explanted xenografted PDAC and surgically resected primary PDAC. We performed transmission and scanning electron microscopy and compared findings to those of the normal pancreatic duct. We also performed single cell cloning to determine the potential options for differentiation.

Results:

Spheroids contained tight junctions and desmosomes but lacked zymogen granules, as expected. The former features were present in normal pancreatic duct, but absent from PDAC cell lines grown in standard 2D culture. Spheroids functionally excluded macromolecules in whole mounts. Cells on the surface of PDAC spheroids were carpeted by microvilli except for rare cells with prominent stereocilia. Carpets of microvilli were also seen in low passage organoids produced from xenografts, and surgically resected human PDAC, in addition to normal human pancreatic duct. We performed single cell cloning and resulting spheroids produced both cell phenotypes at the same approximate ratios as those from bulk cultures.

Conclusions:

Pancreatic cancer spheroids/organoids are capable of bi-phenotypic differentiation.

Introduction

Pancreatic cancer is a notoriously lethal malignancy, while only the 11th most common cancer in men and the 9th most common in females in incidence, it is the third highest cause of cancer deaths in the US.¹ Pancreatic cancer has a 5-year survival of only 8.5% and has barely risen over the last 50 years.² The poor prognosis for patients with pancreatic cancer is explained by the fact that it is typically diagnosed late in the disease process and the malignancy is widely chemorefractory. Several first and second line therapies do exist however, and there are 457 recruiting interventional trials for pancreatic cancer (clinicaltrials.gov, last accessed 2/5/2019).^{3, 4}

The pancreatic duct is responsible for collecting and safely delivering enzyme-rich fluid to the duodenum. Of the 15 digestive enzymes produced in the acini, 9 of them, including trypsin and chymotrypsin, are secreted in inactive forms that are later activated in the duodenum by the enzyme enterokinase. To avoid autolysis of the pancreas, the pancreatic duct must be a professional tight junction producer, and pancreatitis can result if the duct is damaged.^{5, 6} Pancreatic duct cells produce bicarbonate in response to secretin, a hormone produced in the small intestine. This bicarbonate rich fluid is secreted into the duodenum where it neutralizes the acidic material coming from the stomach. Microvilli are a prominent feature of pancreatic duct epithelia where they project into the lumen, while they are scantly present on the apical surface of acinar cells.⁷

Brush cells (also known as tuft cells) have also been described in pancreatic ducts of the rat. These cells contains apical tufts of long microvilli with prominent rootlets, and are generally considered to be chemosensory cells.^{8, 9} However, the existence and roles of tuft cells in human pancreatic duct has been less evident, although there are some indications of tuft cell existence in human pancreatic duct and PDAC.^{10, 11}

Monolayer cell culture has been used to study a variety of fields of cell biology. However, the flat and hard plastic substrates used to create monolayers are clearly not representative of an in vivo environment and do not faithfully reflect the essential physiological properties of live tissues.¹² In vivo, cells in tissues interact with their neighbors and with the extracellular matrix (ECM), which is important for differentiation, proliferation, and other cellular functions. Three-dimensional (3D) culture forces cells to directly interact with each other. Indeed, 3D culture has been shown to be more in vivo-like than 2D culture in terms of differentiation and proliferation.^{13, 14} A number of 3D cell culture methods have been established, including both scaffold or non-scaffold cultures. Scaffolds can be made from biological materials such as proteins consisting of ECM, a classic one being matrigel. Among the non-scaffold 3D cultures is a hanging drop culture, where cells are forced to interact with their neighbors by gravity at the base of the droplet.¹⁵

In this study, we examined the differentiation of human PDAC grown in 3D culture as spheroids or organoids. We found that PDAC cells in spheroids produced tight junctions by electron microscopy and functionally excluded macromolecules. The surfaces of spheroids were carpeted by microvilli, similar to that seen at the surface of resected pancreatic cancers and normal human pancreatic ducts. Stereocilia were also found on rare cells on the surface of spheroids. Single cell cloning experiments demonstrated that both phenotypes could arise from single malignant cells (ie, bi-phenotypic differentiation).

MATERIALS AND METHODS

Samples

All samples were obtained with IRB approval and informed consent. Low passage (<20 passages) cell lines were produced from rapid autopsies (A6L [Pa02C] and A10.7 [Pa03C]) or directly from surgical resection tissues (Panc215 [Pa09C] and Panc 10.05 [Pa16C]) as previously reported.^16-20 Organoids were also produced from xenografts that had been derived from surgically resected human PDACs.²¹ The cancer associated fibroblast cell line (CAF35) was previously reported to be smooth muscle actin positive and diploid by SNP microarray.²² Photographs of all of the cell lines are included as Supplemental Figure 1.

Spheroid and Organoid Production

We designated spheres of cells produced in hanging drop plates from cell lines as “spheroids.” Spheroids were produced by plating cells into 384 well Perfecta3D Hanging Drop Plates (3D Biomatrix, Ann Arbor, Mich) at 3000 cells/30 μl. Cells were cultured for three days to form spheroids that were subsequently used for further analysis. Mixed spheroids were produced by co-plating 3000 A6L with 3000 CAF35 cells.

We designated aggregates of cells produced in matrigel from resection samples as “organoids” and those from surgically resected pancreata and normal pancreatic duct cells as “organoids” as previously described.²³ Organoids for scanning electron microscopy were resuspended in 10 mL of ice-cold complete media, washed twice with 10 mL of ice-cold complete media, before a final resuspension in 1 mL of complete media. We confirmed that the cells growing as organoids were derived from neoplastic PDAC cells by sequencing of GNAS and KRAS using PCR and Sanger sequencing.

Patient derived PDAC xenografts were raised as previously described.²⁴ Briefly, surgically resected specimen of PDAC patients without chemotherapy or radiotherapy were subcutaneously implanted into 5- to 6-week-old female nu+/nu+ mice. Human PDACs were harvested from xenografts, tumors minced, and digested with 200 U/ml collagenase IV (Invitrogen, Waltham, Mass) and 0.6 U/ml dispase I (Sigma, St. Louis, Mo). The tumors were mechanically dissociated using a GentleMacs dissociator (Miltenyi, Bergisch Gladbach, Germany) on the h_tumor_01 setting for 3 dissociation cycles. Each cycle was followed by vortexing for 10 seconds and incubated for 30 minutes at 37°C and 5% CO₂ on a MACSMix mini tube rotator. Following ficoll-plaque density centrifugation, mouse cells were removed using biotinylated anti-mouse CD31 (Becton Dickenson, Franklin Lakes, NJ), anti-mouse H2Kd (Becton Dickenson) and anti-mouse lineage cocktail (Miltenyi) using anti-biotin beads (Miltenyi) and QuadroMACS separator. Cells were resuspended in DMEM:F12 (Life Technologies, Carlsbad, Calif), 1x B27 (Life Technologies), 20 ng/ml bovine fibroblast growth factor (Sigma) and plated on low attachment plates for 7 days.

Primary pancreatic cancer cells were cultured as organoids as previously described.^{23, 25} Briefly, tissue was minced and digested in advanced DMEM/F12 media (Life Technologies) supplemented with 2.5% fetal bovine serum (Life Technologies), 5 mg/mL collagenase type II (Thermo Fisher, Waltham, Mass), and 1.25 mg/mL dispase (Thermo Fisher). Dissociated cells were washed with PBS before plating in Matrigel (Becton Dickenson). Organoids were cultured with Human Feeding Media (HFM) of AdvDMEM/F12 (Thermo Fisher) supplemented with B27 (Thermo Fisher), 1.25 mM N-Acetylcysteine (Sigma), 10 nM gastrin (Sigma), 50 ng/mL EGF (Peprotech, Rocky Hill, NJ), 10% RSPO1-conditioned media, 10% Noggin-conditioned media, 100 ng/mL FGF10 (Peprotech) and 10 mM Nicotinamide (Sigma). When organoids reached confluency, extra cellular matrix was depolymerized with ice-cold media, cells were dissociated with TrypLE (Thermo Fisher), and plated.

IHC and Phalloidin Staining

Spheroids were fixed with 4% paraformaldehyde at 4°C overnight. After embedding, 5 ¼m sections were labeled with anti-E-cadherin (Roche, Basel, Switzerland, Cat# 760–4440) or anti-claudin-4 (Thermo Fisher, Cat#32–9400). For phalloidin staining, PDAC spheroids were frozen in optimal cutting temperature compound (Sakura, Torrance, Calif), and 5 ¼m sections cut using Leica CM 1850 cryostat (Leica, Wetzlar, Germany). These sections containing spheroids were stained with phalloidin (Alexa Fluor™ 488 Phalloidin, Thermo Fisher) and 4’,6-Diamidino-2-Phenylindole, Dilactate (Thermo Fisher, Cat#D3571).

Transmission Electron Microscopy (TEM)

Samples were fixed in 2.5% glutaraldehyde, 3 mM MgCl₂, in 0.1 M sodium cacodylate buffer, pH 7.2 at room temperature for four hours, and then overnight at 4°C. Samples were then rinsed in buffer and post-fixed in 1% osmium tetroxide in 0.1 M sodium cacodylate buffer (1 hr) on ice in the dark. Following a DH₂O rinse, samples were en bloc stained with 2% aqueous uranyl acetate (0.22 ¼m filtered, 1 hr in the dark), dehydrated in a graded series of ethanol, propylene oxide and embedded in Eponate 12 (Ted Pella, Redding, Calif) resin. They were placed into EPON (Ted Pella) overnight followed by three changes of pure EPON the next day before being placed into the oven at 60°C overnight for curing. Thin sections, 60 to 90 nm, were cut with a diamond knife on the Reichert-Jung Ultracut E ultramicrotome (Leica, Buffalo Grove, Ill) and picked up with 2×1 mm formvar coated copper slot grids. Grids were stained with 2% uranyl acetate in 50% methanol and 0.4% lead citrate before imaging on a Philips CM120 electron microscope at 80 kV. Images were captured with an AMT (Rockville, Md) XR80 CCD (8 megapixel) camera. The length of the microvilli were measured using Fiji (NIH, Bethesda, Md, version download date 5–30-2017).²⁶

Scanning Electron Microscopy (SEM)

Samples were fixed in 2.5% glutaraldehyde, 3 mM MgCl₂, in 0.1 M sodium cacodylate buffer, pH 7.2 for overnight at 4°C. After buffer rinse, samples were post-fixed in 1% osmium tetroxide in 0.1 M sodium cacodylate buffer (1 hr) on ice in the dark. Following a deionized H₂O rinse, samples were dehydrated in a graded series of ethanol and left to dry overnight with hexamethyldisilazane (HMDS). Samples were mounted on carbon-coated stubs and imaged on the Zeiss Leo FESEM (Field Emission Scanning Electron Microscope, Oberkochen, Germany) at 1 kV.

Exclusion of Mmacromolecules

Pure PDAC spheroids were grown for three days in hanging drop culture. Dextran, Texas Red®, 3,000 or 70,000 MW, Neutral (Thermo Fisher) or F(ab’)₂ Alexa Fluor 594 conjugated goat anti-rabbit IgG (Cell Signaling, Danvers, Mass, Cat#8889) was added at the final concentration of 1 mg/ml or 1:25 dilution, respectively, and incubated for three hours, placed on a slide and a cover slip applied. The live whole mount slides were then visualized by using fluorescent microscope (Nikon, Tokyo, Japan).

Single Cell Cloning

Pancreatic ductal adenocarcinoma cells were subcultured and single cells were sorted by flow cytometry (core facility, Johns Hopkins Bloomberg School of Public Health, Baltimore, Md) into individual wells of 96-well plates. Conditioned media was prepared by harvesting media (DMEM with 20% fetal calf serum) used to culture respective PDAC cell lines after 24 hours incubation and followed by centrifugation and filtration with 0.22 μm filter. Conditioned media was added directly to the cells without washing. When cells were grown at 50–70% confluent (approximately 14 days), cells were harvested and subcultured for further analysis.

Structural Comparisons

Ultrastructure observations of tight junctions and desmosomes were performed using TEM on established PDAC cell lines and comparing 2D monolayer culture to spheroids produced in 3D culture. Microvilli were examined in spheroids from the established PDAC cell lines and compared to those on the surface of a normal human pancreatic duct sample and a single primary human PDAC organoid from a surgical resection. They were compared to PDAC cells grown in 2D and a CAF cell line grown in 3D as controls. Stereocilia were discovered in PDAC spheroids, and were compared to PDAC xenograft organoids and a single primary PDAC tissue. Single cell cloning was only performed on established PDAC cell lines.

RESULTS

Pancreatic Cancer Cells Grown as 3D Spheroids Have Tight Junctions

We grew human pancreatic cancers in 3D culture and investigated whether they might recapitulate normal duct or normal acinar cell morphology. First, we asked if they produced tight junctions in vitro, a feature of both acinar and duct cells. To test this, we placed spheroids in fluorescent high-molecular weight macromolecules, made live whole mounts, and photographed the spheroids using fluorescent microscopy. Spheroids excluded both 70 kD molecular weight dextran (Fig. 1A) and 110 kD fluorescent IgG (Fab’)₂ fragments (Fig. 1B). Here we produced mixed spheroids by co-plating A6L and CAF35 cells. Mixed spheroids displayed prominent expression of the cell adhesion molecule E-cadherin (Fig. 1C), which was localized to the three surfaces of the PDAC cells and absent from the outer surface (arrows). E-cadherin was not expressed in the CAF cells, as expected. Consistent with the functional exclusion of macromolecules, PDAC cells in spheroids also produced the tight-junction protein claudin-4 (Fig. 1D). Expression of both proteins was absent, as expected, in the CAF cells. By transmission electron microscopy (TEM), PDAC spheroids produced from only A6L cells (no CAF cells) contained tight junctions and desmosomes at every cell-cell contact at the surface of the spheroids (Fig. 1E). This was consistent with the morphology obtained from the normal human pancreatic duct (Fig. 1F). Microvilli could also be seen in these photomicrographs (Figs. 1E, F), however zymogen granules were not found. The lack of zymogen granules, which are specific to acinar cells, was expected and is consistent with a duct cell origin of human PDAC.

FIGURE 1.

A6L and A6L/CAF35 spheroids form tight junctions. Dye exclusion of Texas Red labeled 70,000 molecular weight dextran (A) and Alexa Fluor 594B labeled goat anti-rabbit IgG F(ab')2 fragments, approximate molecular weight 110 kD (B) in pure A6L spheroids. Expression of E-cadherin (C) in mixed spheroids demonstrating polarization. Note the basolateral expression with lack of expression on the apical surface of some cells with basally located nuclei (arrows) and the lack of staining of the CAF core. Expression of the tight junction protein claudin-4 (D). Note the lack of staining of the CAF core. Transmission EM (2% uranyl acetate) of pure A6L spheroids (E) and normal human pancreatic duct (F). Note the presence of tight junctions (TJ) and desmosomes (DES). Microvilli can also be seen in these micrographs. Size bars: 100 μm (A-D), 1 μm (E, F).

Microvilli Coat the Surface of PDAC Cells Grown in 3D

To further investigate the presence of microvilli we performed scanning electron microscopy (SEM), where we found the surface of spheroids produced from A6L cells to be carpeted with microvilli (Fig. 2A). We documented microvilli in spheroids from additional cell lines, A10.7, using TEM (Fig. 2B) and Panc 10.05 (data not shown). As microvilli are known to be F-actin containing structures, we fixed, sectioned and stained whole spheroids with phalloidin, which showed intense staining around the periphery, consistent with F-actin localization at the spheroid surface (Fig. 2C). On higher power TEM, we noted prominent vesicles near the tips of microvilli suggesting a role in active secretion (Fig. 2D). Using TEM, normal human pancreatic duct also displayed prominent microvilli (Fig. 2E). Microvilli were also prominent on the surface of a low-passage organoid from a resected human PDAC (Fig. 2F). We measured the length of the microvilli on the PDAC spheroids and compared them to that of the normal pancreatic duct (Table 1). The microvilli on spheroids of two of the four PDAC cell lines were shorter (100–150 nm shorter), while those from one cell line was somewhat longer (~200 nm). The microvilli lengths had higher standard deviations on the spheroids from PDAC cell lines compared to the normal pancreatic duct sample. Despite these differences, microvilli length varied between the samples by less than 2-fold. In contrast to these findings, PDAC cells grown in 2D showed only scant microvilli and spheroids produced from the CAF35 cell line completely lacked microvilli (Supplemental Fig. 2). We also produced mixed spheroids (from mixtures of PDAC and CAF cells), and these were qualitatively similar, so detailed analysis was not performed (data not shown).

FIGURE 2.

Microvilli coat the surface of spheroids, normal pancreatic duct, and primary PDAC. Scanning electron microscopy (1% osmium tetroxide) of pure A6L spheroids showing a carpet of microvilli decorating the surface of the spheroid (A). Transmission electron microscopy of A10.7 cell line spheroids (B). Arrows indicate microvilli (MV). Phalloidin Alexa Fluor 488 staining of A6L PDAC spheroid section with DAPI staining (C). Note the intense staining at the surface of the spheroid. Spheroid structure was distorted from its spherical shape during the process of embedding and sectioning. Transmission electron microscopy of A6L spheroids demonstrating vesicles (arrows) being produced from the tips of the microvilli (D). Normal human pancreatic duct demonstrating prominent microvilli (E). Arrows indicate microvilli. One electron dense cell is noted. Scanning electron microscopy of freshly resected PDAC (H32) tissue (F). Size bars: 1 μm (A-B), 50 μm (C), 100 nm (D), 1 μm (E-F)

TABLE 1.

Length of Microvilli on PDAC Spheroids and Normal Pancreatic Duct

PDAC Spheroids/Normal Pancreatic Duct	Length, Mean (SD), nm	No. of Microvilli Measured	P
A6L	361.1 (142.6)	131	<0.0001
Panc 215	432.4 (174.3)	25	0.0365
Panc 10.05	395.7 (202.2)	61	<0.0001
A10.7	702.7 (264.1)	57	<0.0001
Normal pancreatic duct	501.5 (141.2)	110	NA

NA indicates not available.

Stereocilia Are Also Prominent on the Surface of PDAC Cell Spheroids

While confirming the presence of microvilli and determining their density, we discovered that some cells on the surface of the spheroids also had prominent stereocilia (A6L, Fig. 3A; Panc 10.05, Fig 3B). The length of the stereocilia ranged from 1.5 to 4 microns. We also found stereocilia containing cells on the surface of organoids produced from patient derived xenografts (Figs. 3C, D). Stereocilia containing cells were relatively rare, and we quantified them relative to the area of the spheroid surface that was investigated (Table 2). The number of stereocilia containing cells for a given surface area was surprisingly consistent (1.6–4.1 cells/10,000 μm², with only 2.5× variation), and not obviously different between the cell lines and xenografts. Stereocilia were also observed in low passage organoids from surgically resected PDAC (Fig. 3E) and freshly harvested PDAC tissue from surgical pathology (Fig. 3F). Attempts to visualize stereocilia by TEM were unsuccessful, presumably because stereocilia are relatively uncommon. Stereocilia were much less frequently seen in PDAC cells grown in standard 2D culture conditions.

FIGURE 3.

Stereocilia on the surface of spheroids, organoids and primary PDAC. Scanning electron microscopy (1% osmium tetroxide) of spheroids produced from the PDAC cell lines A6L (A) and Panc 10.05 (B), where the length of the stereocilia is noted. Scanning electron microscopy of organoids from xenografts JH102 PX4 (C) and Panc253 PX24 (D). Scanning electron microscopy of passage 6 PDAC organoids (H42) from freshly resected PDAC tissue (E), and PDAC tissue without culture (F).

TABLE 2.

Number of Stereocilia Observed on Spheroids

Cell Type	Name	Stereocilia Containing Cells Per 10,000 μm2, Mean (SD), n
Cell lines*	A6L	2.2 (2.0)
	A10.7	1.9 (0.6)
Explanted xenografts	JH102 (PX4)	2.2 (2.4)
	Panc 265 (PX10)	4.1 (4.2)
	Panc 253 (PX24)	1.6 (3.6)

^*Density not calculated for Panc215 or Panc 10.05.

Single PDAC Cells Are Capable of Bi-phenotypic Differentiation

We generated a model that shows cells from bulk culture, when plated in hanging drop plates, produce spheroids with mostly microvilli and rare stereocilia (Fig. 4A). We wondered whether each individual PDAC cell might be capable of producing cells with both distinct morphologies, namely “bi-phenotypic differentiation” (Fig. 4B, Outcome #1). Alternately, PDAC cell lines might be mixtures of two distinct cell types where the more common cell type contains microvilli and the less common cells produce stereocilia (Outcome #2). We reasoned that single cell cloning could be used to distinguish between these alternate hypotheses.

FIGURE 4.

Stereocilia on spheroids from single cells. Model showing spheroid formation after plating cells from bulk culture (A). Short "hairs" represent the microvilli and the long "hairs" represent stereocilia. After single cell cloning, expansion and plating in 3D culture, there are two potential outcomes (B). If the cells are fundamentally equivalent, then the progeny of each clone should be equivalent and similar to the spheroids from bulk culture (Outcome #1). Alternatively, if there are two fundamentally distinct cell types, then progeny from single cells would produce spheroids coated purely with stereocilia (left) and others coated purely with microvilli (right) (Outcome #2). SEM (1% osmium tetroxide) of spheroids produced from single cell clones from A6L (C), A10.7 (D), and 10.05 (E). Note that the stereocilia in panel D appear to be coating the entire cell surface. Number of stereocilia containing cells per 10,000 μm² from bulk culture or single cell clones (F). The A6L Clone #1 is the average of only two biologic replicates, resulting in small error bars, whereas the others are from 3-5 replicates.

Accordingly, we flow sorted PDAC cell lines into single cells, which were then expanded into clones for each of the three of the cell lines (n = 3, A6L; n = 2, Panc 10.05; n = 2, A10.7). After cloning, we plated the expanded cells in hanging drop cultures to produce spheroids, which were fixed, collected and subjected to SEM. Spheroids produced from single cells (Figs. 4C-E) were qualitatively identical to those produced from bulk culture. This supports the first model (Fig. 4B, Outcome #1), namely that individual cells within PDAC cell cultures contain the inherent ability to differentiate into cells coated with pure microvilli as well as cells containing stereocilia. The number of stereocilia per surface area was graphed and clones were compared to bulk culture (statistically not-significant, Fig. 4F).

DISCUSSION

We studied the differentiation of PDAC cells grown in vitro under 3D culture conditions. Tight junctions were demonstrated both by TEM, claudin immunohistochemistry, and the functional exclusion of macromolecules. We also showed that carpets of microvilli decorate the surface of these cells when grown in 3D. Stereocilia are also seen in these differentiated 3D cultures, but stereocilia containing cells while relatively uncommon, were at relatively consistent numbers. We demonstrated that clones from single cells were capable of differentiating into both microvilli and stereocilia coated cells, supporting the concept of bi-phenotypic differentiation of PDAC.

Tight junctions are critical to block the passage of molecules between adjacent cells. Tight junctions were observed at virtually every cell-cell contact point in the PDAC spheroids, consistent with macromolecule exclusion in whole mounts, and claudin positive immunohistochemistry. The presence of tight junctions between adjacent epithelial cells is consistent with the notion that in three-dimensional growth, forced cell-to-cell contact allows cells to exhibit complex physiological. We also demonstrate their presence in normal human pancreatic duct by TEM. Literature containing TEM demonstrating tight junctions in human duct samples is relatively scant, although the elegant work of Kodama clearly demonstrates them.²⁷ The presence of tight junctions in 3D culture has been reported by other investigators.^28-30

Microvilli are small (0.5 to 1 micron in length) finger-like F-actin containing projections that are especially prominent on epithelial cells associated with absorption, such as the small intestine, and with secretion, such as the pancreatic duct.^{8, 27, 31} In PDAC spheroids from different cell lines, we found them to be carpeted with microvilli (Figs. 2A, B), the length of which was fairly consistent, varying by less than 2× from one cell line to the next. We also found vesicles near the microvilli (Fig. 2B) and emanating from them (Fig. 2D). Others have also shown microvilli on the surface of organoids from normal pancreatic duct.³¹ Further, Gagliano and colleagues showed prominent microvilli on the surface of PDAC organoids,²⁹ but Ishiwata and colleagues only showed scant microvilli, possibly related to high passage number.³²

Stereocilia bearing cells were present on the surface of spheroids from all four PDAC cell lines. Their surface density was relatively consistent in both PDAC derived spheroids and the three PDX derived organoids (Table 2). Because of their consistent density, we hypothesize that there is likely a feedback mechanism regulating their expression. We also documented stereocilia containing cells on organoids isolated directly from surgically resected PDACs (Fig. 3E). We hypothesize that these cells are tuft cells, known to be decorated with stereocilia, and generally believed to be chemosensory cells. Tuft cells are well-documented to be present on the apical surface of the main pancreatic duct and its first-order branches in rat and possibly in humans.^8-10 We were able to find only one report suggesting stereocilia bearing tuft cells in human PanIN and possibly PDAC.¹¹ It is unclear why chemosensory cells might be present in the pancreatic duct, although one reasonable hypothesis is that they might be sensing one of the enzymes produced in acini.

Biphenotypic differention (the presence of microvilli and stereocilia containing cells) of PDAC organoids can be explained by competing hypotheses: are PDAC culture fundamentally mixtures of the two cell types (thereby being manifest on organoids) or are they a single “progenitor” type of cell capable of differentiating into two cell types? To address this, we performed single cell cloning, expansion of these single cells and spheroid formation. Our data definitively support the concept of bi-phenotypic differentiation (Fig. 4B, Outcome #1). The results were qualitatively similar across the experimental systems used, and were not affected by the presence or absence of CAFs or whether the malignant cells were grown as hanging drop cultures from cell lines or organoids in matrigel from xenografted tumors.

It is unlikely that the stereocilia we report are related to invadopodia. Invadopodia are subcellular structures that are typically located on the ventral surface of cells grown in 2D culture and play a role in local invasion and metastasis.³³ At the ultrastructural level, they are actin containing projections that are associated with invaginations.³⁴ The stereocilia we describe are different from invadopodia since they decorate the whole cell surface and are not associated with invaginations.

One shortcoming of our work is that we did not perform detailed grading as described by Sipos and colleagues.³⁵ However, the microvilli and stereocilia are qualitatively similar between the cell lines, so that any potential differences that might exist by grading, would not be causative.

In summary, PDAC cells forced to grow in intimate contact with one another in 3D cultures display bi-phenotypic differentiation, where both cell types can be produced from a single cell. The presence of tight-junctions, dense microvilli carpeted surfaces and stereocilia bearing cells further support a ductal origin of human PDAC.^{27, 36}