Graphical abstract

   graphic file with name fx1.jpg
   [51]Open in a new tab

Highlights

     * •
       Human primary cholangiocyte organoids (PCOs) upregulate HLA-I/II in
       inflammatory conditions
     * •
       Allogeneic immune response is substantially ameliorated by
       donor-recipient HLA matching
     * •
       Autologous PCOs induce a low-level immune infiltration into the
       graft site
     * •
       Humanized mice distinguish early/vigorous and late/resolved stages
       of allogeneic response
     __________________________________________________________________

   Petrus-Reurer et al. use in vitro co-cultures and in vivo humanized
   mice to examine autologous and allogeneic immune response to human
   primary cholangiocyte organoids (PCOs). The authors show that
   autologous cells induce a low-level immune infiltration into the graft
   and allogeneic cells display HLA-driven evolving stages of immune
   rejection.

Introduction

   Regenerative cellular therapies have emerged as a promising approach
   for the repair or replacement of diseased or damaged tissues and
   cells.[52]^1^,[53]^2^,[54]^3 The predominant strategy for the
   generation of therapies is differentiating embryonic stem cells (ESCs)
   or genetically modified induced pluripotent cells (iPSCs) into the
   desired cell type.[55]^1^,[56]^4 Cells cultured in 3D organoid format
   have also recently been developed as an alternative strategy for the
   generation of cellular therapies.[57]^5^,[58]^6^,[59]^7 An important
   barrier for the clinical translation of cellular therapies is a
   thorough understanding of their immunogenicity.[60]^8^,[61]^9 This
   knowledge is critical for informing the use of the clinical strategies
   for the reduction of the immune response after transplantation,
   including the use of immunosuppressive drugs, donor-recipient matching,
   gene editing, and encapsulation of the cells.[62]^10

   It is generally expected that “autologous” cellular therapies, derived
   from cells obtained from the intended recipient, are unlikely to induce
   an immune response. However, early studies suggested that syngeneic
   (autologous) mouse iPSCs may or may not be
   immunogenic[63]^11^,[64]^12^,[65]^13^,[66]^14^,[67]^15^,[68]^16 and
   that immunogenicity of autologous iPSC-derived cellular therapies may
   be dependent on the target cell type and the immune
   microenvironment.[69]^17^,[70]^18^,[71]^19^,[72]^20 These conflicting
   findings may be in part because generation of ESC- and iPSC-derived
   cellular therapies necessarily involves genetic modification that may
   affect the immunogenicity of the derivative cells. Furthermore, ESC-
   and iPSC-derived cells may lack full differentiation into an adult
   phenotype, which may also impact their immunogenicity.[73]^21
   Additionally, other molecular factors independent of cell origin, but
   instead derived from in vitro culturing (e.g., acquired mutations by
   passaging, cell viability, and culture matrix), may also have a
   significant impact in triggering immune responses.

   Development of autologous cellular therapies represents significant
   logistic and economic barriers that are likely to limit their
   widespread clinical use for the foreseeable future.[74]^4^,[75]^22
   “Allogeneic” cellular therapies, derived from a different donor than
   the intended recipient, are thus being explored as potentially cheaper
   “off the shelf” treatments.[76]^8 Clinical data from decades of solid
   organ transplant experience demonstrate that the immune response to
   most allogeneic solid organs is driven predominantly by human leukocyte
   antigen (HLA) molecules. Consequently, when permitted by a large pool
   of donors and recipients (such as in kidney transplantation), HLA
   matching between the donor and the recipient is an effective strategy
   for reducing (but not eliminating) the likelihood of immune-mediated
   rejection and for prolonging graft survival.[77]^23^,[78]^24^,[79]^25
   Allogeneic cellular therapies are similarly expected to induce an
   immune response via direct and indirect allorecognition mechanisms
   driven in large part by HLA. However, it is important to note that most
   cellular therapies in advanced stages of development consist of a pure
   population of a single cell type,[80]^26^,[81]^27^,[82]^28 in stark
   contrast to solid organs that are composed of numerous cell types with
   diverse phenotypes and functions, including antigen-presenting cells
   that are highly adapted to driving an immune response.

   A key challenge to understanding the immune response to human cellular
   therapies has been the availability of appropriately refined
   experimental models and access to primary human tissues. Most studies
   to date have been performed exclusively in vitro, have used
   mouse-derived cellular therapies transplanted into wild-type
   immunocompetent mice, or have used ESC- or iPSC-derived cells
   transplanted into mice reconstituted with an allogenic human immune
   system.[83]^10^,[84]^11^,[85]^12^,[86]^13^,[87]^14^,[88]^15^,[89]^16^,[
   90]^17^,[91]^18^,[92]^19 Therefore, a detailed profiling of the human
   immune response under different HLA-matching scenarios using relevant
   humanized models and cutting-edge analytical technologies is critical
   for the advancement of these therapies.

   We have previously shown that human primary cholangiocyte organoids
   (PCOs) can be derived from human cholangiocytes without genetic editing
   and with long-term expansion potential for use as cellular and
   bioengineered therapies for bile duct
   disorders.[93]^29^,[94]^30^,[95]^31 In this study, we use a
   comprehensive panel of in vitro and in vivo experimental models to
   unprecedentedly examine the response of autologous and allogeneic
   (including Partially Matched and Fully Mismatched) immune cells to
   human PCOs, as an exemplar human primary organoid cellular therapy. We
   demonstrate that autologous cells induce a detectable immune response,
   which may be influenced by new mutations acquired during in vitro
   culture, cell viability, and cell matrix. We also show that the
   allogeneic immune response to PCOs is driven by the level of HLA
   mismatch between the donor and recipient, showing evolving stages of
   immune rejection in a humanized mouse model. Our findings provide
   high-resolution analysis of immunogenicity of autologous and allogeneic
   organoid cellular therapies, highlighting important implications for
   their clinical translation.

Results

PCOs upregulate expression of HLA-I and HLA-II in inflammatory conditions

   Human PCOs were generated and characterized as described previously
   from gallbladder or bile duct biopsies[96]^29^,[97]^30^,[98]^31 taken
   from more than 70 HLA-typed deceased transplant organ donors. For this
   work, we selected one bile duct-derived and long-term expanded PCO line
   (PCO Line 1), which showed characteristic PCO spheroid morphology
   ([99]Figure S1A) and expression of specific cholangiocyte markers
   including SOX17, SOX4, TFF2, FGF2, KRT-7, and KRT-19 ([100]Figure S1B),
   in addition to alkaline phosphatase and gamma-glutamyl transferase
   enzymatic activity comparable to human primary cholangiocytes
   ([101]Figures S1C and S1D).

   Flow cytometric analysis revealed high bimodal expression of HLA-I and
   low expression of HLA-II (<40%) by primary cholangiocytes. PCOs
   cultured under normal conditions abundantly expressed HLA-I but no
   HLA-II. Both HLA-I and HLA-II were upregulated on PCOs when cultured in
   inflammatory conditions, simulated by 2 days of co-culture with
   interferon (IFN)-ɣ. Human primary cholangiocytes and PCOs all expressed
   the cholangiocyte marker CD326/Epcam as expected ([102]Figure 1A). The
   relative expression of HLA-I (A, B, and C) and HLA-II (DP, DQ, and DR)
   genes was quantified using RT-qPCR ([103]Figure 1B), demonstrating
   upregulation by PCOs in inflammatory conditions for all loci and
   comparable to primary cholangiocytes. To examine whether the
   upregulation of HLA-II by PCOs upon co-culture with IFN-ɣ is
   physiologically relevant, we transplanted PCOs under the kidney capsule
   of mice reconstituted with an allogeneic human immune compartment
   (humanized mice). There was robust in vivo upregulation of expression
   of HLA-II by PCOs that were transplanted into humanized mice compared
   to immunodeficient mice ([104]Figure 1C). These data suggest that both
   HLA-I and HLA-II are likely to drive the immune response to PCOs in the
   clinical setting.

Figure 1.

   [105]Figure 1
   [106]Open in a new tab

   HLA expression by human primary cholangiocytes and PCOs

   (A) Representative flow cytometry charts (upper row) and bar graphs
   showing percentage of live cells (7-Aminoactinomycin D [7-AAD]
   negative) and mean fluorescence intensity (MFI, lower row) of CD326,
   HLA-I and HLA-II surface expression of human primary cholangiocytes,
   and PCOs cultured with or without IFN-ɣ stimulation for 2 days from
   three primary gallbladder/bile duct-derived lines (shown in different
   symbols). Fluorescence minus one (FMO) controls are used for gating.
   Error bars represent mean ± SEM from three technical replicates per
   line in three PCO lines.

   (B) Quantitative reverse-transcription PCR (RT-qPCR) of HLA-I (HLA-A,
   HLA-B, HLA-C, and HLA-E) and HLA-II (HLA-DP, HLA-DQ, and HLA-DR)
   expression by primary cholangiocytes and PCOs cultured with or without
   IFN-ɣ stimulation for 2 days from three primary gallbladder/bile
   duct-derived lines (shown in different symbols). Error bars represent
   mean ± SEM from three technical replicates per lines in three PCO
   lines.

   (C) Immunofluorescence images showing expression of human HLA-II and
   KRT-7 in PCO grafts under the kidney capsule in immunodeficient NSG
   mice with and without reconstitution with allogeneic human immune cells
   (humanized). Scale bars: 100 μm. See also [107]Figure S1.

Donor-recipient HLA mismatch determines in vitro immune response to PCOs

   We next assessed the level of immune activation as measured by cytokine
   secretion in vitro by co-culturing PCOs with spleen-derived mononuclear
   cells (SPMCs) derived from same deceased donors used to generate the
   PCO lines ([108]Table S1), thus enabling immunological experiments with
   autologous or allogeneic donors with known HLA mismatch. In autologous
   experiments, PCOs and SPMCs were obtained from the same deceased
   transplant organ donor. In Partial Match experiments, PCOs and SPMCs
   were matched at HLA-I only (A, B, and C loci). In Full Mismatch
   experiments, PCOs and SPMCs were mismatched at both HLA-I and HLA-II
   loci (A, B, C, DP, DQ, and DR) ([109]Table S2). We used SPMCs (rather
   than peripheral blood mononuclear cells; PBMCs) for these experiments
   as it was possible to obtain much larger numbers of SPMCs per donor to
   enable both in vitro and in vivo experiments. Of note, we have
   previously shown that SPMCs can be used for immunological assays in
   vitro and in vivo.[110]^32 Fully Mismatched PCOs induced the strongest
   and significantly higher immune response as measured by the levels of
   IFN-ɣ, tumor necrosis factor alpha (TNF-α), and interleukin (IL)-6
   secretion compared to other groups ([111]Figures 2A–2C; [112]Table S3).
   Except for IFN-ɣ, Autologous and Partial Match groups induced a similar
   level of immune activation that was significantly higher than the
   negative controls. Note that, interestingly, the levels of IFN-ɣ
   production induced by Autologous and Partial Match PCOs were not
   different from negative controls. Secretion of the anti-inflammatory
   IL-10 was also significantly higher in the Full Mismatch group compared
   to other groups, potentially suggestive of simultaneous activation of
   inhibitory pathways ([113]Figure 2D). There were no changes in
   IL-12p70, IL-13, IL-1β, IL-2, IL-4, and IL-8 secretion levels
   ([114]Table S3).

Figure 2.

   [115]Figure 2
   [116]Open in a new tab

   Cytokine secretion of autologous and allogeneic lymphocytes co-cultured
   with PCOs

   (A) Bar graphs showing concentration of IFN-ɣ secretion by SPMCs when
   co-cultured (1:15 ratio) with Autologous, Partial Match, and Full
   Mismatch PCOs (passage 10) for 5 days. Negative controls are PCOs only
   and SPMCs only, and positive control is SPMC Mixed Lymphocyte Reaction
   (MLR) from two different donors. Same conditions were analyzed for
   TNF-α (B), IL-6 (C), and IL-10 (D). Error bars represent mean ± SEM
   from three independent experiments with technical duplicates. ∗p <
   0.0001; ∗∗p < 0.001; ∗p < 0.01. See also [117]Tables S1, [118]S2, and
   [119]S3.

Magnitude of in vivo immune infiltration to PCOs is driven by the level of
HLA mismatch

   We next examined the in vivo immune response to PCOs using a humanized
   mouse model. PCO fragments were injected under the kidney capsule of
   immunodeficient non-obese diabetic severe combined immunodeficiency
   gamma (NSG) mice. At 3 weeks post-injection, animals were humanized by
   intraperitoneal injection of Autologous, Partial Match, or Full
   Mismatch SPMCs obtained from the same donors as used for the in vitro
   co-culture experiments. Importantly, in each animal, one kidney was
   transplanted with PCOs in Matrigel, and the contralateral kidney was
   injected with Matrigel only, thus serving as control for the effect of
   surgery and possible immune response to the Matrigel matrix. 6 weeks
   post-injection, animals were culled and their spleens were recovered to
   assess the human immune cell composition ([120]Figure 3A). Flow
   cytometry analysis demonstrated 20%–40% human CD45^+ cells in the
   spleen, confirming successful humanization ([121]Figures S2A and S2B;
   [122]Table S4). Cytomety by Time-Of-Flight (CyTOF) analysis revealed
   reconstitution with predominantly lymphoid cells as expected, with
   paucity of myeloid cells. Of note, the immune profile in the spleen of
   the mice was different in animals transplanted with Autologous, Partial
   Match, or Full Mismatch PCOs ([123]Figures 3B–3D; [124]Table S4). The
   Full Mismatch group had a decreased abundance of CD8 and a relatively
   higher proportion of B cells and CD57^− effector memory CD4 T cells.
   Original donor SPMCs also showed differences in abundance of immune
   cell types (CD4, CD8, natural killer [NK], monocytes/dendritic cells
   [DC], and B cells), but with overall good representation of all cell
   types in all groups ([125]Figure S2C; [126]Table S4).

Figure 3.

   [127]Figure 3
   [128]Open in a new tab

   Immune profile of humanized mice transplanted with PCOs

   (A) (Upper panel) Schematics of the PCO derivation, PCO injection, and
   subsequent humanization with SPMCs after 3 weeks (n = 4–5 per group).
   Animals were culled, and samples (kidneys, spleens, and blood)
   collected after 3 weeks of humanization. (Lower panel) Representation
   of the groups included in the study: PCOs were injected in Matrigel
   under the left kidney capsule. Matrigel alone was injected under the
   contralateral kidney capsule to control for Matrigel-driven immune
   infiltration. Humanization was performed using Autologous, Partial
   Match, or Full Mismatched human SPMCs. A Non-Humanized group was added
   as control.

   (B) T-Distributed Stochastic Neighbor Embedding (t-SNE) visualization
   and Flow Self-Organizing Map (FlowSOM) clustering show the immune
   profile of human CD45^+ cells with the distribution of immune cell
   subpopulations in different colors. Mice with the same donor
   engraftment have been overlaid.

   (C) Stacked bar graph representing the frequency of immune subsets in
   each type of engraftment.

   (D) Heatmap illustrating the level of expression of each marker for
   each cluster defined by FlowSOM analysis (all groups combined). See
   also [129]Figure S2 and [130]Tables S1, [131]S2, and [132]S4.

   Assessment of local immune infiltration to the graft site was performed
   by immunofluorescence staining with specific human immune and
   cholangiocytes markers (CD45 and Keratin-7 [KRT-7], respectively). PCO
   grafts had the characteristic morphology and KRT-7 surface marker
   expression in all groups ([133]Figure 4A), including upregulation of
   HLA-II, indicative of alive PCOs responding to surrounding inflammatory
   response ([134]Figure S3). The entire engrafted areas were scanned, and
   hCD45+ cells were segmented and counted with a custom-made pipeline
   ([135]Figure S4; [136]Table S5). The “background” hCD45+ immune
   infiltration in the Matrigel-only area of the contralateral control
   kidney was subtracted to correct for non-PCO-driven immune
   infiltration. The total PCO-specific immune infiltration was
   significantly greater in the Full Mismatch group, followed by Partial
   Match and Autologous groups ([137]Figure 4B). Specifically, hCD45+
   immune infiltration was 55.3% ± 2.9% in the Autologous group
   (respective Matrigel-only: 47.9% ± 3.1%), 30.1% ± 1.6% in the Partial
   Match group (respective Matrigel-only: 20.1% ± 2.6%), and 30.9% ± 2.7%
   in the Full Mismatch group (respective Matrigel-only: 14.1% ± 5.4%)
   ([138]Table S5). Importantly, autologous PCOs induced an immune
   infiltration (5%–10% of total hCD45+ cells) above that seen in the
   contralateral Matrigel-only control kidneys from the same mice. This
   apparent immune response to autologous cells was consistent with the
   low-level immune activation suggested by the in vitro co-cultures
   ([139]Figure 2).

Figure 4.

   [140]Figure 4
   [141]Open in a new tab

   Infiltration of hCD45+ cells into PCO graft sites

   (A) Hematoxylin/eosin and immunofluorescence images of injected PCOs
   (passage 10) under the kidney capsule showing expression of human KRT-7
   and human CD45 markers in Autologous, Partial Match, Full Mismatch, and
   Non-Humanized groups. Matrigel-only images, from the same animals, are
   shown as controls for background immune infiltration induced by
   surgical procedure and injection of Matrigel. Note that, despite the
   dense CD45^+ cell infiltration into the graft site in the Partial Match
   representative image shown (PCO), this was not induced by the
   cholangiocytes present in the graft given the similarly dense
   infiltration also observed in the Matrigel-only control from the
   contralateral kidney of the same animal. Scale bars: 100 μm.

   (B) Percentage of hCD45+ cells infiltrating into the graft site
   quantified in each group relative to the respective Matrigel-only
   controls (by subtraction). Error bars represent mean ± SEM from 1 to 4
   quantified areas per mouse per group. ∗∗∗p < 0.0001; ∗∗p < 0.01. See
   also [142]Figures S3 and [143]S4 and [144]Table S5.

HLA mismatch determines the phenotype and stage of immune response into the
graft site in vivo

   To determine the phenotype of the immune infiltrate into the graft
   site, we performed spatial transcriptomics using the GeoMx platform
   (NanoString, Bruker), which enabled the retrieval of RNA from hCD45+
   and hKRT-7+ cells. Deconvolution of the hCD45+ compartment showed that
   the infiltrate in the Full Mismatch group had a decrease in CD4^+
   memory T cells and an increase in T regulatory cells (Tregs), B cells
   (naive and memory), and monocytes relative to Matrigel-only controls
   ([145]Figures 5A, 5B, and [146]S5A). Difference in immune composition
   was less prominent between the Autologous and Partial Match groups;
   however, the former showed a shift from CD4 memory to CD8 memory; and
   the latter a change from CD4 native to CD8 memory. Overall, there was
   limited overlap in up-regulated or down-regulated genes between the
   groups, and different immune-related pathways were enriched in the
   three groups ([147]Figures 5C, [148]S5B, and S5C; [149]Table S6). In
   particular, hCD45+ cells in the Autologous group did not show
   enrichment in any particular immune responses compared to controls
   (Matrigel-only). Conversely, the Partial Match group depicted general
   cellular immune activation, while the Full Mismatch group showed an
   antigen-driven, humoral response in addition to B cell and Treg
   pathways ([150]Figures 5B and 5C), which is consistent with the cell
   types described by the deconvolution. Interestingly, the hKRT-7+
   compartment shared significantly more up- and down-regulated genes
   between groups ([151]Figures S6A and S6B; [152]Table S6). Pathway
   enrichment analysis suggested that PCOs in the Partial Match group had
   upregulated pathways related to cell death state (e.g., apoptotic
   pathways), while Full Mismatch PCOs had upregulated cell stress (but
   not death; e.g., ubiquitin) pathways ([153]Figure S6C). Collectively,
   the data are consistent with the hypothesis that immune-related
   pathways are relatively quiescent in the Autologous group but that
   there is ongoing activation of the immune pathways in the Partial Match
   and the Full Mismatch groups that have evolved to different stages of
   the immune response.

Figure 5.

   [154]Figure 5
   [155]Open in a new tab

   Phenotype of infiltrating immune cells into PCO graft sites

   (A) Histogram showing abundance of different immune cells types
   infiltrating into PCO graft sites (left) and their respective fold
   change compared to Matrigel-only controls (right). Matrix (SafeTME) was
   extracted from Danaher et al.[156]^33

   (B) Immunofluorescence images of injected PCOs (passage 10) under the
   kidney capsule for human immune markers CD3, CD20, and FoxP3 in
   Autologous, Partial Match, and Full Mismatch groups. Matrigel-only
   images, from the same animals, are shown as controls for background
   immune infiltration induced by surgical procedure and injection of
   Matrigel. Scale bars: 100 μm.

   (C) Pathway enrichment analysis of Autologous, Partial Match, and Full
   Mismatch groups for general, humoral, antigen, cellular, B cell, and
   Treg pathways.

   (D) Dot plot showing the association of Autologous, Partial Match, and
   Full Mismatch groups with rejection categories extracted from human
   kidney rejection datasets (TCMR, ABMR, chronic TCMR, acute TCMR, and
   allograft rejection).

   (E) Schematic of the evolution of the stages of the human immune
   response to PCOs with different levels of matching in a humanized mouse
   model. See also [157]Figures S5 and [158]S6 and [159]Table S6.

Immune pathways upregulated in humanized mice correlate with those activated
by human organs undergoing rejection in transplant recipients

   Our data suggest that, compared to the Partial Match group, the immune
   response in the Full Mismatch group is at a later stage of its
   evolution, with a dominance of B and regulatory T cells infiltrating
   the graft site and activation of the associated pathways. Conversely,
   Partial Match grafts appeared to have induced an earlier “active” stage
   of the immune response. To examine this hypothesis, we assessed how the
   patterns of immune activation in the humanized mice compared to those
   seen in human transplant recipients of solid organs with biopsy-proven
   active clinical rejection. We thus compared the spatial transcriptomics
   data from the humanized mice to publicly available human datasets: two
   kidney rejection microarray datasets[160]^34^,[161]^35; Banff-Human
   Organ Transplant (B-HOT) dataset[162]^36 with classified T
   cell-mediated rejection (TCMR), antibody-mediated rejection (ABMR),
   acute TCMR, and chronic TCMR; and allogeneic rejection pathways
   extracted from Kyoto Encyclopedia of Genes and Genomes (KEGG) and
   PathCards (allograft rejection). Consistent with our hypothesis, we
   noted that the gene expression profile in human kidneys undergoing
   rejection correlated best with the Partial Match group and less so with
   the Full Mismatch group ([163]Figure 5D). Overall, spatial
   transcriptomics in this SPMC-humanized mouse model captured at a single
   snapshot in time (6 weeks post-cell therapy injection) suggest that
   Partial Match immune response corresponds to early stages of response
   with potent T cell activation and with PCOs showing transcriptional
   signs of apoptosis. Conversely, Full Mismatch induced an immune
   response that has evolved to a later stage involving B cells, Tregs,
   and humoral response, with PCOs showing more transcriptional signs of
   stress than apoptosis ([164]Figure 5E).

Mutational burden, viability, and culture matrix composition influence
autologous PCO response

   We next examined potential mechanisms by which autologous PCOs can
   induce an immune response. To assess the extent of in vitro
   culture-related mutations in PCOs, we sequenced the genomes of
   organoids at different passages from our PCO line using nanorate
   sequencing (NanoSeq),[165]^37 an advanced single-molecule duplex
   sequencing method allowing ultra-accurate detection of mutations in
   polyclonal cell populations ([166]Figure 6A). Estimates of mutation
   burdens per cell for each PCO sample provided evidence for roughly
   constant acquisition of mutations during culture ([167]Figure 6B).
   Organoids acquired an average of 1,248 mutations per cell after five
   passages (49 days), with an estimated average of ∼14 new protein-coding
   mutations per cell ([168]Figure 6B). This observation is consistent
   with previous studies reporting accumulation of in vitro DNA damage
   during cell culture in other systems.[169]^38^,[170]^39^,[171]^40
   Because at least some of these mutations may lead to changes in
   immunogenic epitopes, it is plausible that some of the immune response
   to autologous cells could be related to presentation of neoantigens by
   autologous PCOs.

Figure 6.

   [172]Figure 6
   [173]Open in a new tab

   Potential conditions influencing immune response to autologous PCOs in
   vitro

   (A) Schematic diagram of the experimental design for PCO culture and
   sequencing.

   (B) Mutation burdens per cell as estimated through NanoSeq in the
   primary tissue and PCO samples. Mean burdens per cell are indicated
   above each bar, with error bars denoting 95% confidence intervals.
   Mutations in excess of the mutation burden for the primary tissue
   sample (horizontal line) are colored in dark blue. Passage number (P)
   and days in culture (d) are indicated for each PCO sample (top).
   Estimated numbers of mutations acquired during in vitro culture for
   each PCO sample (P0, P1, and P5). Total mutations per cell are
   calculated as the difference between the mean burdens per cell for the
   PCO sample and the primary tissue sample (dark blue portion of the bars
   in B). Coding mutations per cell are calculated by scaling the total
   mutations per cell by the ratio between whole genome size and
   protein-coding genome size (bottom).

   (C) Heatmap showing cytokine secretion of IFN-ɣ, TNF-α, IL-6, and IL-10
   by SPMCs when co-cultured (1:15 ratio) with autologous PCOs in normal
   or killed (80°C, 1.5 min) conditions for 5 days. Values are normalized
   to No Kill:SPMC co-culture condition. Negative controls are PCOs only
   and SPMCs only, and positive control is SPMC Mixed Lymphocyte Reaction
   (MLR). ∗∗p < 0.01; ∗p < 0.05.

   (D) Bright-field pictures showing PCOs cultured in Matrigel or
   synthetic hydrogel for 3 days prior to assay. Scale bars: 200 μm.

   (E) Bar graphs showing concentration of IFN-ɣ, TNF-α, IL-6, and IL-10
   secretion by SPMCs when co-cultured (1:15 ratio) with autologous PCOs
   for 5 days. Negative controls are PCOs only and SPMCs only, and
   positive control is SPMC MLR. Error bars represent mean ± SEM from
   three independent experiments. ∗∗p < 0.01; ∗p < 0.05.

   We also hypothesized that cell viability could impact the immune
   response to autologous PCOs. To test this, control PCOs (82% viable)
   and PCOs warmed to 80°C for 1.5 min to induce cell death (40% viable)
   were co-cultured with autologous SPMCs. The cytokine secretion profile
   was significantly impacted by the level of viable PCOs in the
   co-cultures ([174]Figure 6C), therefore indicating that as cells die
   they may release and express a variety of molecular factors that
   influence immune response.

   Finally, we sought to investigate the impact of the cell culture
   matrix. PCOs grown in regular Matrigel or in a commercially available
   synthetic hydrogel (chemically defined) were co-cultured with
   autologous SMPCs ([175]Figure 6D). As hypothesized, cytokine secretion
   profile was significantly impacted by the cell matrix composition in
   which PCOs were cultured in ([176]Figure 6E). Specifically, PCOs in the
   test synthetic hydrogel triggered a stronger immune response, possibly
   due to altered cell growth, phenotype, and antigen expression. Overall,
   our data suggest that in vitro culture, cell viability, and cell
   culture matrix can affect immune response to autologous cells.

Discussion

   This study provides a distinctive and comprehensive comparison of in
   vitro and in vivo immune responses to autologous and allogeneic
   long-term expanded healthy organoids derived from human primary cells.
   We selected PCOs as our exemplar genetically unedited organoid cellular
   therapy, as they have previously been shown to have potential utility
   as a cellular therapy.[177]^29^,[178]^30^,[179]^31 We show that human
   PCOs upregulate HLA-I and HLA-II molecules in inflammatory conditions
   in vitro and in vivo. The immunogenicity of PCOs in vitro, as
   quantified by release of inflammatory cytokines by co-cultured
   lymphocytes, was dependent on the level of donor-recipient HLA
   matching. Consistent with this finding, immune infiltration into the
   graft site in humanized mice was also dependent on the level of
   donor-recipient HLA matching. Autologous PCOs also induced low-level
   immune infiltration into the graft site compared to controls, maybe
   influenced by new mutations acquired during in vitro culturing, cell
   viability, and the cell culture matrix. Spatial transcriptomics data
   indicate evolving stages of immune rejection dependent on
   donor-recipient HLA matching. Specifically, the immune infiltrate
   profile in the Partial Match graft sites was consistent with an early
   and prominent T cell-driven response, compared to a later-stage immune
   response driven by B cells and Tregs in the Full Mismatch group.
   Finally, we were able to correlate the immune profile of the
   graft-infiltrating cells with previously published human kidney
   allograft rejection profiles.

   Here, we focused on the impact of HLA on immunogenicity of the PCOs,
   because decades of evidence from solid organ transplantation have
   established HLA as the predominant driver of the alloimmune response
   and demonstrated the beneficial impact of donor-recipient HLA matching
   on graft immune rejection and survival.[180]^23^,[181]^24^,[182]^25 To
   enable this study, we obtained paired human primary immune cells and
   primary cholangiocytes from numerous deceased transplant organ donors.
   By obtaining large numbers of spleen-derived lymphocytes from every
   donor, we were able to replicate the same experiments in vitro and in
   vivo. Thus, unprecedentedly in a human primary-derived organoid system,
   we were able to compare the immune response to autologous PCOs, PCOs
   matched at HLA-I (A, B, and C) but mismatched at HLA-II (DP, DQ, and
   DR) loci (Partial Match group), and PCOs mismatched at both HLA-I and
   HLA-II loci (A, B, C, DP, DQ, and DR).

   It is noteworthy that previous studies assessing immunogenicity of
   human organoids have either used aggregated non-expandable primary
   cells[183]^41 or organoids derived from pluripotent stem cells (thus
   being substantially genetically modified to differentiate them into
   [not-fully] functional cells).[184]^42^,[185]^43 The PCOs we present in
   this study were human primary-derived cholangiocyte cells with
   comparable 3D morphology and function to primary tissue and with very
   high expansion potential. It cannot be assumed, therefore, that they
   will elicit the same immune responses as non-expanding cells or
   organoids derived from genetically modified pluripotent cells.

   Our data demonstrate that PCOs, similar to human primary
   cholangiocytes, upregulate both HLA-I (A, B, and C) and HLA-II (DP, DQ,
   and DR) in inflammatory conditions in vitro and in vivo. Consistent
   with our findings, upregulation of HLA in inflammatory in vitro
   conditions has been previously reported for other putative cellular
   therapies derived from ESCs or
   iPSCs.[186]^44^,[187]^45^,[188]^46^,[189]^47 These experiments confirm
   that PCOs have the capacity to induce immune responses to both HLA-I
   and HLA-II after transplantation (i.e., inflammatory conditions), even
   if they do not express high levels of HLA-II under ideal culture
   conditions. The magnitude of the in vitro and in vivo allogeneic immune
   responses (as measured by the total number of hCD45+ infiltrating
   immune cells) was significantly reduced by donor-recipient matching at
   HLA-I but was nonetheless greater than autologous response, thus being
   driven by mismatch at HLA-II. These findings are indicative that
   allogeneic PCOs would induce both anti-HLA-I and anti-HLA-II immune
   responses in a clinical setting and that donor-recipient HLA matching,
   at both HLA-I and HLA-II loci, would be an effective strategy for
   amelioration of the alloimmune response in patients.

   We performed engraftment of the cholangiocyte organoids first and then
   humanization, to (1) ensure we have surviving graft cells and (2) allow
   time for resorption of the hydrogel and recovery from surgery, before
   introducing the immune system. This approach also maximizes the
   experimental window by minimizing the likelihood of graft vs. host
   disease in the model. Additionally, by injecting Matrigel alone under
   the contralateral kidney capsule in every humanized mouse, we
   controlled for both the immune response induced by surgery and Matrigel
   itself and the inter-animal variance in immune engraftment that is
   inevitable with humanized mouse models.[190]^48^,[191]^49^,[192]^50
   Remarkably, the immune profile of the lymphocytes infiltrating into the
   graft site in vivo was different depending on the level of HLA
   mismatch. Given that initial immune subsets from original donors have
   similar composition, and levels of humanization at endpoint were around
   60–70% in all analyzed mice, the only difference between study groups
   resides in the level of HLA matching in the PCOs. This confirms an
   association between local immune cell infiltration and HLA mismatch of
   the grafts. Our findings are consistent with the clinical observation
   that donor-recipient HLA mismatch influences likelihood of immune
   rejection in liver transplant recipients[193]^51 and the known
   mechanisms of allorecognition including the role of HLA in
   orchestrating T and B cell responses.[194]^52

   Spatial transcriptomics analysis revealed differences in immune cell
   populations and cell enrichment pathways present at the graft site
   depending on the level of HLA mismatch. The immune infiltrate in the
   Partial Match group was dominated by CD8^+ and CD4^+ (CD3^+) T cells,
   but with B cells and Tregs in the Full Mismatch group. The adaptive
   immune response to alloantigens is known to evolve over time after
   transplantation, being initially dominated by effective CD4^+ and CD8^+
   (CD3^+) T cells, while at later time points there is a gradual
   emergence of B cell responses as well as regulatory T
   cells.[195]^52^,[196]^53^,[197]^54 Our findings are consistent with
   this view of the kinetics of the alloimmune response and suggest the
   dominance of two different phases of immune response, possibly
   corresponding to an early/prominent vs. a late/resolved response, in
   Partial Match and Full Mismatch groups, respectively. In fact, the
   prominent immune profile captured in the Partial Match group was highly
   associated with human rejection signatures extracted from datasets of
   human kidneys. Although PCO grafts for all groups demonstrated
   characteristic morphology and maker expression (KRT-7+) as well as
   expected response to an inflammatory milieu (HLA-II upregulation),
   spatial transcriptomics suggested a more apoptotic-like state in the
   Partial Match group. These results are consistent with an evolving
   immune response, comprising a more vigorous early immune response
   captured in the Partial Match group and a later quiescent stage in the
   Full Mismatch group.

   Interestingly, we also observed greater production of IL-10, a
   predominantly anti-inflammatory cytokine, in lymphocyte in vitro
   co-cultures with Full Mismatch PCOs. Release of IL-10 and other
   inhibitory cytokines has indeed been postulated as one of the negative
   feedback mechanisms to curtail ongoing immune responses to
   alloantigens, including from NK cells.[198]^55 However, we were unable
   to experimentally test how donor-recipient HLA mismatch influences the
   kinetics of the evolution of the immune response to PCOs, because all
   animals were culled at a single fixed time point after immune
   reconstitution. It is also possible that differences in the donor
   immune compartment or immune engraftment in the animals contributed to
   some of the differences in the immune responses captured in our study.
   Of note, however, the composition of the immune compartment of the
   donors was broadly similar.

   The immune response to autologous human cellular therapies has been
   reported in previous studies with conflicting
   findings.[199]^17^,[200]^18^,[201]^19^,[202]^56^,[203]^57^,[204]^58 Our
   data suggest that autologous PCOs may also induce a low-level immune
   response. Autologous PCOs induced in vitro production of higher levels
   of the inflammatory cytokines TNF-α and IL-6, but not IFN-ɣ. Autologous
   PCOs also induced greater infiltration of immune cells into the graft
   site compared to control (Matrigel-only) injection sites. Following up
   on previous studies demonstrating the mutational impact of in vitro
   culturing,[205]^59^,[206]^60 we hypothesized that some low-level
   autologous response could be caused by mutations acquired with PCO
   culture. Here we show that mutations are acquired by passage, with a
   proportion affecting coding regions, which could plausibly lead to the
   expression of neoantigens and drive immune response in in
   vitro-expanded autologous cells. Additionally, we interrogated if cell
   viability (inevitable when cells are cultured in vitro and after
   manipulation for transplantation in vivo) and cell matrix composition
   could influence autologous immune response. Our results indicate that
   autologous co-cultures are significantly impacted by both cell
   viability and culture matrix, and possibly other factors. Although our
   data confirm that culture conditions affect expression of HLA on PCOs,
   the specific molecules and factors that are up- or down-regulated to
   induce these responses are yet to be identified. Despite this, it is
   not possible to definitively conclude whether autologous PCOs would
   induce a similar immune response in a clinical setting or if such weak
   immunogenicity would be clinically relevant and lead to significant
   cell loss or require short-term immunosuppression. Overall, these
   analyses allude to the importance of characterizing the immunogenicity
   of any culture prior to its use as cellular therapy at both genetic and
   phenotypic levels despite their autologous origin.

   In summary, our study confirms that both HLA-I and HLA-II are likely to
   drive T and B cell-dependent immune response to organoid cellular
   therapies in a clinical setting. Donor-recipient HLA matching and
   elimination or HLA expression by genetic editing are likely to be
   effective strategies to substantially ameliorate, but not eliminate,
   the immune response to allogeneic cellular therapies in a clinical
   setting. This is consistent with transplant studies that have
   identified minor histocompatibility (and other) molecules as making a
   small contribution to the alloimmune response.[207]^61 Our findings
   suggest that some immunosuppressive therapies are still likely to be
   required in patients receiving optimally HLA-matched cellular
   therapies. Moreover, our data also do not exclude the possibility that
   autologous cellular therapies may similarly induce a low-level immune
   response in patients. However, it is well established that minimizing
   the overall burden of immunosuppression reduces its adverse
   effects.[208]^62 Elimination of the alloimmune response against HLA-I
   and HLA-II, through the use of autologous cellular therapies,
   HLA-matching, or genetic editing, would thus be expected to
   substantially reduce the required dose of immunosuppression and
   associated adverse effects in patients. Taken together, this work
   provides a high-resolution perspective to guide future efforts focused
   on deeper understanding of the immunogenicity of cellular therapies and
   help accelerate the clinical translation of cell-based products for
   regenerative medicine applications.

Limitations of the study

   In our in vivo experiments, we transplanted PCOs under the kidney
   capsule rather than into the liver. We chose the kidney capsule because
   it is a highly vascular niche[209]^63^,[210]^64 that supports the
   survival of PCOs, enables localization of the graft at the experimental
   endpoint, and represents a well-characterized niche for assessment of
   immunogenicity. It is possible, however, that the immune response to
   PCOs may differ qualitatively or quantitatively from our findings when
   transplanted into the liver in patients. While we used unselected SPMCs
   (consisting of both lymphoid and myeloid cells) in our in vitro
   experiments, our humanized mouse models primarily recapitulated the
   lymphoid human immune compartment. It is thus possible that recipient
   myeloid cells may also influence the immune response to PCOs in a
   clinical setting. Future studies utilizing other experimental models
   would be necessary to examine the contribution of myeloid cells to the
   alloimmune response to PCOs. Notwithstanding these caveats, it is
   important to note that our in vitro and in vivo experimental models
   were sufficiently refined to distinguish between the immune response to
   allogeneic PCOs with different levels of donor-recipient HLA mismatch.

Resource availability

Lead contact

   Requests for further information should be directed to and will be
   fulfilled by the lead contact, Kourosh Saeb-Parsy (ks10014@cam.ac.uk).

Materials availability

   Cholangiocyte organoids described in this publication will be made
   available on request from the University of Cambridge under a materials
   transfer agreement with the university.

Data and code availability

     * •
       Raw mass cytometry data are available on Zenodo
       ([211]https://doi.org/10.5281/zenodo.10054742). Spatial
       transcriptomics data (FASTQ files, processed count matrices,
       annotation files, and metadata) are available on the Gene
       Expression Omnibus (GEO: [212]GSE253679
       ) and Dryad ([213]https://doi.org/10.5061/dryad.g4f4qrfx9). DNA
       sequencing data are available on the European Genome-phenome
       Archive (EGA: EGAD00001015456).
     * •
       Code for spatial transcriptomics analyses has been deposited on
       Zenodo ([214]https://doi.org/10.5281/zenodo.10054742). Custom
       computer code for DNA sequencing data has been deposited on GitHub
       ([215]github.com/baezortega/PCO2024).
     * •
       Any additional information required to reanalyze the data reported
       in this paper is available from the [216]lead contact upon request.

Acknowledgments