Rationale
The broad goal of our research is the generation of organs-on-chip models of human lymphoid tissue to study human immunology and immunotherapy. In this proposal, we use design principles from the human lymph node and the clinically targeted receptor CTLA-4 as a case study to identify the impact of providing relevant 3D organization, matrix and cell types on human immunology R&D. The lymphoid tissue module will be invaluable in
  1. Mimicking polarization of the T and B cell receptor machinery and the presence of Fc-receptor expressing cells to analyze the true efficacy and cytokine release induced on cross-linking of therapeutic antibodies. This includes immediate antibody mediated cytokine release syndrome (e.g. rituximab) and delayed immune related adverse events seen with checkpoint blockade (4).
  2. Recapitulating 3D interactions between human immune cells, cancer cells and human endothelial cells-important for tumor induced immunosuppression, trafficking of anti-tumor lymphocytes (including CAR T cells) and antibodies (e.g rituximab efficacy on malignant B cells resident in the LN) and their function.
  3. Using the prototype human lymphoid tissue on chip (hLToC) to design in vitro trials for some of the staggering number of combinatorial immunotherapies being considered for cancer treatment and identification of the best combinations for clinical trial.
  4. Using the hLToC to uncover personalized tumor-immune biology and test the impact of genetic diversity on anti-tumor immunity.
    There are two published 3D models of human lymphoid tissue, both trying to recapitulate the human lymph node: VaxDesign’s MIMIC system and Probiogen’s HuALN (8). In the HuALN, the T and B cells are disorganized. The HuALN shows long-term survival and generation of IgM in response to antigen but no IgG, suggesting that the immune response does not mature. In the MIMIC system, T and B cells cultured on separate microcarriers are mixed to simulate T cell dependent antibody responses. There is no reticular network and immune response is generated by immigrant dendritic cells. The model doesn’t recapture LN anatomy and has not been tested for immunooncology.
    The design principles we have followed derive from a
    Preliminary data:
  1. Use of matrix and cell populations based on published data: The extracellular matrix (ECM) composition of human LNs is unknown but there are preliminary studies on primates. We compared existing literature about the primate LN ECM (9) and Matrigel (10). Matrigel is a commercially available basement membrane-like gel that contains a fibrillar network of collagen IV, as well as laminin and heparin sulfate proteoglycan, which is reminiscent of the primate LN ECM. To improve longevity of the gel, we added 1.5mg/ml bovine collagen I. PBMC preparations have 40-50% T cells and only 3-15% B cells. Whole lymph nodes, however, contain ~50-60% T cells and ~40-50% B cells. The previously published MIMIC system has shown that cultures of T and B cells in a LN-like ratio produce higher titers of Ig (8). Thus in preliminary studies we cultured an approx. 1:1 mix of T and B cells in 60%Matrigel and 1.5mg/ml collagen I.
  2. Polarization of CD3 and CTLA4 upon 3D culture: We found that 4-7 days of culture, a significant fraction of T cells were polarized with CD3 and CTLA-4 accumulating in cap like structures.
  3. Lymph-tissue-on-chip design: We engineer microfluidic devices of different geometries with polydimethylsiloxane (PDMS), an optically clear, biocompatible and stable polymer. These devices can be perfused for weeks via engineered vessels lined with human vascular or lymphatic endothelium and visualized in-situ by microscopy; it is also possible to perfuse these devices with whole human blood. We can engineer ring-like (Fig 2A) or parallel channels (Fig 2B) to mimic separated T and B zones as seen in the LN and perfuse them via lymphatic and blood vessels (ideal for studying trafficking into the LN). Yet, in many tertiary lymphoid structures, this segregation is not as evident and we started our experiments with a thin mixed lymphocyte gel (T+ B, 200uM thickness) separated from a perfusion channel by a porous membrane (Fig. 2C).
  4. Controlling for donor and experimental variability: Tertiary lymphoid structures can be quite variable. Thus for establishing donor and experimental variability, we relied on design principles from the LN and basic lymphocyte activation assays. A LN-like T:B ratio of 60:40 in standard 2D cultures was activated with heat-killed formalin fixed Staph. aureus Cowan I (SAC) (Fig. 3). SAC is commonly used to model T-cell dependent polyclonal immunoglobulin (Ig) production, and it is typically performed using purified B cells provided with T cell produced cytokine or using whole PBMC preparations that have 40-50% T cells and only 3-15% B cells. Whole lymph nodes, however, contain ~50-60% T cells and ~40-50% B cells. The previously published MIMIC system has shown that cultures of T and B cells in a LN-like ratio produce higher titers of Ig (8).
  5. Use of Matrigel based on published data: The extracellular matrix (ECM) composition of human LNs is unknown but there are preliminary studies on primates. We compared existing literature about the primate LN ECM (9) and Matrigel (10). Matrigel is a commercially available basement membrane-like gel that contains a fibrillar network of collagen IV, as well as laminin and heparin sulfate proteoglycan, which is reminiscent of the primate LN ECM. Thus, in preliminary studies, we cultured lymphocytes in Matrigel. In future studies several biological and synthetic polymers will be tested.
  6. Organoid formation in Matrigel: The T and B cells isolated from human PBMC were labeled with two different colors of CellTrace or CellTracker dyes to enable discrimination. Over multiple donors, we defined labeling conditions and the right consistency of Matrigel. We expected that uniform distribution of lymphocytes would be maintained in the absence of organizing cues from stromal cells. Serendipitously we found that in 2-4 days, the cells organize into clusters, in particular the B cells form clusters in a bed of T cells (Fig. 4A and B), reminiscent of B cell follicles in the LN. Cluster formation was density dependent. 1-2 million cells in 10ul formed clusters, but 0.5 million cells did not (Fig. 4C), suggesting close range or contact-dependent interactions. Clustering occurred in experiments performed both on chip and in Transwell membrane cultures. Our data suggest that T and B cells can drive their own 3D organization. This phenomenon is especially relevant to the formation of tertiary lymphoid organs which develop without the presence of lymphoid tissue inducer (LTi) cells known to important in embryonic and neonatal lymphoid organization. Previous data indeed suggest that B cells can perform LTi like functions in tertiary lymphoid tissue (11).
  7. IL-2 production by hLN-on-chip: To simulate antigen arrival at the organized structures, we will fabricate microfluidic chips containing channels lined by human lymphatic endothelial cells. However, we first confirmed that antigen can be delivered under flow to the chip. We perfused 1:14,000 SAC in the perfusion channel and were able to stimulate IL-2 production (Fig. 5).
These data demonstrate that an organ-on-chip strategy is feasible to create an in vitro models of human lymph tissues. Further we were able to establish the survival and function of key lymphocyte populations. We have also optimized labeling of the immune cell recovered from PBMC such that they can be visualized for 2 weeks by imaging.
C. Specific aims: Cancer immunotherapy can be best designed by using human tissue provided with the right biomechanical cues. Yet, peripheral blood mononuclear cells (PBMC) are the most commonly available resource. Our preliminary data provides evidence that development of a prototype human lymphoid module using engineered polydimethylsiloxane microfluidic devices and human peripheral blood mononuclear cells is feasible. The human lymphoid tissue on chip (hLToC) will enable the study cancer immunotherapy in an entirely human, patient specific paradigm. To focus our studies on a discrete lymph node function and benchmark the impact of increasing matrix and cellular complexity on immunotherapy design, we chose to study the expression pattern of CTLA-4, a clinically relevant target and response to CTLA-4 blockade therapy. The matrix composition that we have defined in preliminary experiments induces CD3 polarization as seen in the human lymph node where T cells are primed for response. As predicted, we find that CTLA-4 is also polarized. In doing so, our preliminary data identified an intriguing phenomenon: the self-organization of lymphocytes into CTLA-4+ clusters. Thus our specific aims are:
Normal donors:
  1. Test standard immune responses in in a 3D microfluidic model of human lymphoid tissue (hLT-on-chip) and describe the efficacy and toxicity of CTLA-4 blockade
  2. Perfusion of the hLT-on-chip with lymphatic and blood endothelial cells via microengineered vessels and impact on function and response to CTLA-4 blockade
    Cancer patients:
  3. Modeling the impact of cancer on human lymphoid tissue, specifically lung cancer, lymphoma and chronic lymphocytic leukemia
    D. Research Design and Methods
    Aim 1: Test standard immune responses in in a 3D microfluidic model of human lymphoid tissue (hLT-on-chip) and describe the safety and efficacy of CTLA-4 blockade
  1. T and B cell stimulation will be studied using anti-CD3 antibody OKT3, which is known to stimulate T cells in PBMC cultures and with heat-killed formalin fixed Staph. aureus Cowan I (SAC) which induces polyclonal Ig production to test if the 3D organization improves reactivity to immunostimulation. OKT3 is particularly relevant as it is an immunomodulator used to prevent graft rejection and induces cytokine release as seen with anti-cancer antibodies. In addition to providing the ability to activate B cells, SAC serves as positive control due to the abundance of pathogen associated molecular patterns (PAMPs, eg: LPS). The parameters tested will be IL-2 and IgM production by ELISA, dilution of the fluorophore as cells divide and lastly multiplex cytokine assays to perform unbiased assessment of functional states via differences in cytokine profile.
  2. Are the CTLA4+ clusters germinal centers-AID staining and IgG production: The clusters that we observe in response to SAC stimulation have densely packed B cells along with CTLA4+ T cells. This is reminiscent of germinal center staining in human lymph nodes. This presents the intriguing possibility that we have modeled the germinal center in vitro. The key distinguishing factor between 2D stimulation of peripheral B cells vs. in-vivo responses is the appearance of germinal centers is enhanced class switching and somatic hypermutation. We will test these by checking the expression of AID (expressed during class switching and somatic hypermutation), ….(expressed only during somatic hypermutation) and checking the level of IgG production. If we can detect…, we might further perform immunosequencing i.e. testing the sequences of the antibodies where divergence of genomically encoded sequences can be detected. In case we do NOT observe somatic hypermutation, our model is still valuable to assess the impact of T cell polarization, matrix and tissue like density on immunotherapeutic responses.
  3. Effect of anti-CTLA-4 therapy on T and B cell responses: hLN-on-chip prototype will be activated with OKT3 and SAC with or without monoclonal antibodies blocking CTLA-4 and these data will be compared to T cell activation in 2D culture with bead bound antibodies. Readouts defined in (b) will be tested. To our knowledge, these data would be the first entirely human, in vitro model of CTLA-4 blockade in previously untreated tissue resident T-cells.
  4. Providing a reticular network of antigen capturing cells: The lymph node stroma is composed of several types of immune and stromal cells that form a reticular network continuously scanned by lymphocytes for antigen and costimulatory signals (12). Further subcapsular sinus macrophages provide a layer of retained antigen for lymphocytes to access. To model retained antigen and costimulatory signals provided by non-lymphocytes within organized tissue, we have 3 accessible cell types in monocytes (CD14+ magnetic bead selection before isolation of lymphocytes) and monocyte derived macrophages and dendritic cells (moDC).
    In preliminary experiments we will first use monocytes and monocyte derived dendritic cells (GM-CSF/IL-4 cultured monocytes) to establish an antigen presenting stroma for lymphocyte scanning. Depending on the success of these experiments and if time permits, the other cellular sources we can explore are monocyte derived macrophages, macrophages, dendritic and stromal cells from the human lymph node. We have already tested that a 5% monocyte frequency can be successfully introduced in the gel. We expect to test a range within 5-20% and ask if self-clustering, polarization and response to stimulation and checkpoint blockade changes in single cell suspension recovered from the hLN-on chip. Using the assays described above, we will test the impact of Mo/Mo-DC inclusion on the response to SAC, OKT3 and anti-CTLA-4, alone or in combination.
    Caveats: If organization of PBMC into lymphoid tissue like structures needs to be optimized, we can use bioengineering approaches to create cytokine microdomains or use lentiviral overexpression of cytokines. It is possible that the clusters we observe do not show AID positivity and somatic hypermutation-we will again learn from in-vivo design and ask what other cell types, secreted signals and matrix components can be provided to induce somatic hypermutation in response to activation. Further, not all tertiary lymphoid structures show germinal center activity. Thus our model could still be used to test the impact of immunotherapy of activation of lymphocytes within tumor associated tertiary lymphoid structures.