Reconstructing a human immune system in mice for immuno-oncology testing
Building a human immune system inside a mouse begins with clearing space. The recipient's own hematopoietic stem cell niches must be emptied — otherwise resident mouse cells outcompete the incoming human cells for the limited bone marrow real estate needed to engraft.
A "humanized" mouse begins from the same severely immunodeficient genetic background used in PDX tumor models — most commonly NSG or NOG mice (NOD-scid IL2 receptor gamma chain knockout). These animals already lack mature T cells, B cells, and functional NK cells, so a foreign human hematopoietic graft is not immediately rejected the way it would be in an immunocompetent mouse.
But the absence of an adaptive immune system alone is not enough — the mouse's own hematopoietic stem cells still occupy the bone marrow niches that a transplanted human stem cell would need to colonize. Conditioning is the step that opens up that physical and signaling space.
Neonatal mice are preferred recipients because their thymus is still developing and more receptive to seeding by incoming human T-cell progenitors — a window that narrows substantially in adult mice.
Total body irradiation at a carefully calibrated sub-lethal dose (commonly 1–4 Gy depending on strain and pup age) damages and depletes the recipient's own bone marrow hematopoietic stem and progenitor cells without killing the animal outright. This frees up the specialized marrow "niche" microenvironments — osteoblastic and vascular niches — that stem cells require for engraftment, survival, and self-renewal signaling.
Overdosing causes lethal marrow aplasia or radiation sickness; underdosing leaves too much resident competition and yields poor human chimerism. Facilities running humanized mouse colonies at scale titrate dose carefully by pup weight and age, and irradiators are calibrated and dosimetry-verified on a strict schedule since even modest dose drift measurably shifts engraftment outcomes across a cohort.
Because irradiation carries welfare and reproducibility concerns (variable dosimetry, radiation-induced tissue damage confounding downstream tumor or drug studies), several alternative or adjunct conditioning approaches are used:
• Busulfan chemotherapeutic conditioning: an alkylating agent that ablates marrow stem cells without whole-body radiation exposure, increasingly used for adult mouse conditioning • Anti-CD117 (c-Kit) antibody conditioning: selectively depletes host hematopoietic stem cells by blocking the stem cell factor receptor needed for their survival, sparing other tissues entirely • Genetic niche-clearing strains: some newer strains carry mutations that intrinsically weaken resident mouse HSC fitness, reducing or eliminating the need for irradiation altogether
Each approach trades off engraftment efficiency, cost, technical complexity, and confounding effects on the eventual experimental readout.
With marrow niches cleared, purified human CD34-positive hematopoietic stem and progenitor cells are introduced into the neonatal mouse. These are the master cells from which every lineage of the human immune system — T, B, NK, and myeloid — will ultimately derive.
CD34 is a surface glycoprotein expressed on hematopoietic stem and progenitor cells but lost as cells differentiate into mature lineages, making it a reliable marker for stem/progenitor isolation. Common tissue sources include:
• Umbilical cord blood: rich in CD34+ cells, ethically accessible via consented donation programs, and the most widely used source for humanized mouse colonies • Fetal liver tissue: historically important, very high CD34+ content, but access is more restricted • G-CSF-mobilized adult peripheral blood: CD34+ cells mobilized from marrow into circulation, collected by apheresis, used less commonly due to lower reconstitution efficiency
Cells are purified using immunomagnetic bead selection or fluorescence-activated cell sorting (FACS) targeting the CD34 surface marker, typically achieving greater than 90% purity before transplant. Purity and viability are checked by flow cytometry immediately before injection.
Once injected — most commonly directly into the liver of a neonatal pup (a site rich in hematopoietic activity during early development) — CD34+ cells must migrate to and colonize the bone marrow. This homing is guided principally by the CXCR4 receptor on stem cells binding SDF-1 (CXCL12) chemokine gradients produced by marrow stromal cells, drawing cells out of circulation and into the niche.
Once lodged in the marrow niche, successfully engrafting stem cells begin self-renewing divisions (maintaining the stem cell pool) alongside differentiating divisions that generate the earliest committed progenitors — the first branch point in what will become the full human hematopoietic tree.
Species-crossed signaling is imperfect: several mouse cytokines (e.g., mouse thrombopoietin, mouse IL-3) bind human receptors poorly, which is precisely why enhanced strains expressing human cytokine knock-ins (see Stage 3) dramatically improve engraftment over standard NSG mice.
Engraftment efficiency is dose-dependent but with diminishing returns — beyond a certain CD34+ cell number, additional cells do not proportionally increase long-term chimerism because niche space, not cell number, becomes the limiting factor. Typical practice injects on the order of 0.5–2×10⁵ purified CD34+ cells per neonatal pup.
Human chimerism, measured as the percentage of human CD45+ (pan-leukocyte marker) cells among total CD45+ cells in peripheral blood, is essentially undetectable in the first 1–2 weeks, then begins rising as the first differentiated progeny emerge from the engrafted stem cell pool — the true "reconstitution" phase that follows in Stage 3.
Over the following weeks and months, the engrafted stem cell pool differentiates into every branch of the human immune system, each with its own maturation site and timeline inside the mouse — recreating, in miniature, the architecture of human hematopoiesis.
Different human immune lineages emerge from the engrafted CD34+ pool on distinct schedules, reflecting the different maturation pathways each cell type requires:
• B lymphocytes: the earliest lineage to appear in circulation (around 4–6 weeks), maturing directly within the mouse bone marrow much as they would in a human, since B cell maturation does not require a specialized organ beyond marrow • T lymphocytes: appear later (8–12 weeks) because T cell progenitors must first migrate to the mouse thymus, undergo positive and negative selection against mouse thymic epithelium, and only then emigrate as mature, but mouse-thymus-educated, human T cells • NK cells and myeloid cells (monocytes, dendritic cells): develop throughout but are typically under-represented relative to a real human immune system in standard NSG-based humanized mice, because mouse-derived cytokines supporting these lineages (IL-15, GM-CSF, M-CSF) act poorly on human receptors
By 16 weeks post-engraftment, human CD45+ chimerism in peripheral blood commonly reaches 40–70% in well-humanized animals.
Because standard NSG/NOG mice express only mouse cytokines, several lineages — particularly myeloid cells and NK cells — reconstitute poorly, skewing the resulting immune system toward an unrepresentative lymphoid-heavy composition. Newer engineered strains address this directly by replacing mouse cytokine genes with their human orthologs (knock-in), so human progenitor cells receive human-compatible growth signals:
• MISTRG mice: human M-CSF, IL-3, GM-CSF, and Thrombopoietin knock-ins, dramatically improving human myeloid and NK cell development • NSG-SGM3: human SCF, GM-CSF, and IL-3 transgenes, improving myeloid lineage output • Human IL-15 knock-in strains: specifically boost human NK cell reconstitution, otherwise poorly supported by mouse IL-15
Each engineered strain trades increased biological fidelity for higher cost, more complex breeding, and sometimes reduced overall litter viability, so the choice of strain depends on which immune compartment a given study needs to be well represented.
Cytokine knock-in humanized strains can raise functional human NK and myeloid cell frequencies several-fold compared to standard NSG-based humanization, which is often decisive for immuno-oncology studies where innate immune effector cells matter.
Reconstitution is tracked longitudinally by drawing small peripheral blood samples and running multicolor flow cytometry panels against human lineage markers: CD45 (pan-human-leukocyte), CD3/CD4/CD8 (T cells), CD19/CD20 (B cells), CD56 (NK cells), and CD14/CD33 (myeloid cells). Plotting these percentages over successive weeks generates a lineage reconstitution curve unique to each animal, since engraftment efficiency and lineage balance vary meaningfully even within a genetically identical cohort due to biological variability in the original CD34+ graft and individual animal factors.
Only animals crossing pre-defined chimerism and lineage-diversity thresholds are typically enrolled into downstream immuno-oncology studies, since under-reconstituted animals would give an incomplete or misleading picture of human immune-drug interactions.
The presence of human immune cell surface markers is necessary but not sufficient — a truly useful humanized mouse model must demonstrate that its reconstituted human cells actually function: recognizing antigens, mounting coordinated responses, and communicating through human cytokine networks.
Surface marker expression confirms a cell is present and differentiated, but functional assays are required to confirm the reconstituted immune system can actually respond to a challenge:
• Antigen-specific T-cell proliferation: immunizing humanized mice with a defined antigen (e.g., a viral peptide) and measuring T-cell expansion and cytokine production (IFN-γ, IL-2) confirms functional antigen recognition and T-cell receptor signaling • Mixed lymphocyte reactions (in vitro): co-culturing isolated human T cells from the mouse with allogeneic stimulator cells tests proliferative and cytotoxic capacity • Antibody responses: immunization followed by measuring human antigen-specific IgM-to-IgG class switching tests whether B cells are receiving appropriate T-cell help and undergoing germinal center-like maturation • Cytokine release assays: stimulating whole blood or splenocytes and measuring human cytokine panels (IFN-γ, TNF-α, IL-6, IL-2) confirms coordinated innate-adaptive crosstalk
A common validation failure mode is a mouse with high overall human chimerism but a poor functional T-cell response — usually because T cells were educated on mouse (not human) thymic epithelium, producing a T-cell repertoire imperfectly restricted to human antigen-presentation machinery.
Because T cells in standard humanized mice mature against mouse (not human) MHC molecules in the thymus, their T-cell receptor repertoire is only partially optimized for recognizing antigen presented by human HLA molecules — a mismatch that can blunt antigen-specific responses. Strains additionally engineered to express human HLA class I and/or class II molecules (sometimes alongside implanted human thymic tissue fragments, as in the BLT — bone marrow/liver/thymus — model) allow T-cell education to occur, at least partly, on human antigen-presenting machinery, producing a more physiologically realistic and functionally competent T-cell compartment.
The BLT model in particular — co-transplanting human fetal thymus and liver fragments under the kidney capsule alongside CD34+ marrow reconstitution — produces some of the most robust and functional human T-cell responses of any current humanized mouse platform, at the cost of substantially greater surgical and tissue-sourcing complexity.
Before enrolling humanized mice into an immuno-oncology efficacy study, most programs apply hard cutoffs on both chimerism level and functional readouts — for example, requiring greater than 25% human CD45+ chimerism and a positive antigen-specific proliferative response before an animal is used. Animals failing these criteria are typically excluded, since including a poorly reconstituted or non-functional immune system would introduce noise that could mask or falsely suggest a drug effect.
This validation gate is analogous to quality control steps used across other biotech model systems — the parallel articles on PDX genomic drift monitoring and PDX biobank annotation in this collection describe similar QC philosophies applied to tumor rather than immune models.
The entire purpose of building a human immune system inside a mouse is realized at this final stage: engrafting a human tumor alongside the reconstituted immune system creates a preclinical platform capable of testing therapies that work by engaging the immune system itself — something conventional PDX models in immunodeficient mice cannot do.
Standard PDX models rely on severely immunodeficient mice specifically because they lack the T cells, B cells, and NK cell function that would otherwise reject the implanted human tumor. This is exactly the machinery that immune checkpoint inhibitors, CAR-T cell therapies, bispecific T-cell engagers, and cancer vaccines are designed to engage — meaning these drug classes are fundamentally untestable for efficacy in a standard immunodeficient PDX mouse, since there is no immune system present to activate.
Humanized mice solve this by co-engrafting a functional (if imperfect) human immune system alongside the tumor, either by combining the humanization protocol described in Stages 1–4 with tumor implantation, or by directly infusing human peripheral blood mononuclear cells (PBMCs) into a tumor-bearing immunodeficient mouse for shorter-duration studies.
Common immuno-oncology study designs using humanized mice include:
• Checkpoint inhibitor efficacy: treating tumor-bearing humanized mice with anti-PD-1, anti-PD-L1, or anti-CTLA-4 antibodies and measuring tumor growth inhibition alongside intratumoral T-cell infiltration and activation markers • CAR-T and TCR-engineered cell therapy: infusing human CAR-T cells into a humanized or PBMC-humanized tumor-bearing mouse tests both tumor-killing efficacy and on-target/off-tumor toxicity in a system with human immune cell interactions present • Bispecific T-cell engagers: molecules that physically bridge a tumor antigen and CD3 on T cells require a human T-cell compartment to demonstrate their mechanism, making humanized models essential for their preclinical evaluation • Combination immunotherapy regimens: testing checkpoint inhibitors alongside chemotherapy, radiation, or targeted agents to identify synergistic or antagonistic interactions with the immune compartment
The PBMC-humanized mouse (direct injection of human PBMCs into an immunodeficient tumor-bearing mouse, bypassing the multi-week HSC engraftment process) is a faster, cheaper alternative for short studies — but is limited by graft-versus-host disease that typically develops within 3–5 weeks as the infused human T cells begin reacting against mouse tissue antigens.
A persistent challenge in humanized mouse immuno-oncology studies is xenogeneic graft-versus-host disease (GvHD): because human T cells recognize mouse tissue as foreign, they can eventually mount a systemic attack against the mouse host itself, causing weight loss, skin changes, and organ damage that both harms the animal and confounds interpretation of an experimental therapy's effect on tumor growth.
Study designs mitigate this by keeping study windows short relative to typical GvHD onset (often completing efficacy readouts by 4–8 weeks post-tumor-engraftment), using HLA-matched or partially MHC-compatible strains that reduce xenoreactivity, and closely monitoring body weight and clinical GvHD scoring throughout the study so that animals developing significant GvHD can be censored from efficacy analysis rather than misread as a treatment effect.