The Price of “Humanization”

Immunodeficient mice engrafted with functional human cells and tissues, also known as “humanized” mice, represent an innovative preclinical bridge for in vivo studies of human infectious disease, cancer, regenerative medicine, graft-versus-host disease, allergies, and immunity. ¹⁵ The NSG mouse (NOD.Cg-Prkdc^scidIl2rg^tm1Wjl /SzJ) and less frequently the NRG (NOD.Cg-Rag1^tm1MomIl2rg^tm1Wjl/SzJ) are commonly used to generate mice with a “humanized” immune system. The most common approach is to engraft human CD34+ hematopoietic stem cells (human CD34+ HSC), which migrate to the bone marrow of the engrafted mice and differentiate to all lineages of the mature immune system. ¹³ Humanization, however, has unintended consequences that complicate and confound results and may pose significant hurdles by causing illness and euthanasia of study animals before study completion. In the current issue of Veterinary Pathology, 2 articles ^2,6 document the clinicopathological consequences of the humanization of NSG/NRG mice.

In this issue of the journal, Blümich and colleagues from the University of Zurich examined a large cohort of young humanized and nonhumanized NSG mice and correlated clinical, morphologic, flow cytometry, and molecular data to provide insights into the variability of humanization of the immune system and its potential research impact. ² They characterized the morphological phenotype of NSG mice and their humanized counterpart, the CD34+ hu-NSG murine model generated via injection of human CD34+ HSC isloated from fetal liver cells into the liver of perinatal NSG mice. They reported a multisystemic granulomatous inflammatory lesion that was presumed to be mediated by T helper cells. Intriguingly, the authors observed variability and a general lack of correlation between engraftment levels in the blood and lymphoid tissues as measured by the abundance of human CD45+ in circulation versus tissue. Notably, the authors observed that a high immune reconstitution (humanization), measured by human CD45+ cells in circulation, does not necessarily predict a high repopulation at the tissue level. This variability and disparity between the blood and tissue engraftment levels raises a question on the reliability of blood engraftment levels as a measure of the humanization of the immune system.

Blümich et al also made an interesting observation that the proportion of human lymphocytes in the blood were significantly reduced in mice that exhibited a granulomatous inflammatory phenotype compared with those that did not. They speculated that this might result from enhanced migration of human leukocytes from the bloodstream to the sites of tissue injury (akin to infections and sepsis) or due to compromised hematopoietic function secondary to lesions in the bone marrow. In either case, an interesting proposition from the authors is that an increase in T cells, primarily activated helper T cells in the circulation. Blümich et al propose that this might help spot individual mice affected by subclinical inflammatory lesions, which could then be confirmed via the histological examination of organs such as the bone marrow, liver, and kidneys that are typically affected by these inflammatory lesions. Their finding is consistent with previous studies that implicate aggressive expansion of human CD4+ T cells in driving the inflammatory phenotype. ^3,19 Blümich et al ² also documented morphologic variation in control humanized NSG and showed that even those background lesions are exacerbated by humanization. They also documented lesions in the hu-NSG mice attributed to pre-engraftment total body irradiation, which will be of value to pathologists using these models.

If humanization has a cost, then more humanization comes with a steeper price, as Janke and coauthors showed in this issue of the journal. ⁶ In this notably collaborative effort, Janke and colleagues from St. Jude Children’s Hospital and collaborators across multiple institutes documented the consequence of “enhanced” humanization in NSG-SGM3 or NRG-SGM3 mice, NSG/NRG mice that express human stem cell factor (SCF), granulocyte-macrophage colony-stimulating factor (GM-CSF), and interleukin-3 (IL-3). These cytokines enable stable and enhanced engraftment of human myeloid-derived cells and regulatory T cells to support improved multilineage human hematopoietic reconstitution and function which may be inadequate in NSG mice. ^1,17 Janke et al ⁶ characterized NSG-SGM3/NRG-SGM3 mice that were naive (unmanipulated) or engrafted with human CD34+ HSC or patient-derived leukemia xenografts. The authors made 3 fascinating observations that were consistent with the predominance of myeloid-derived proliferations: histiocytic (granulomatous) lesions reminiscent of secondary hemophagocytic lymphohistiocytosis/macrophage activation syndrome (HLH/MAS), mast cell hyperplasia, and eosinophil hyperplasia. The authors speculate that these lesions were driven by the cytokine transgenes, especially by GM-CSF, which is likely behind the exuberant proliferation and activation of macrophages and the ensuing the HLH/MAS-like clinical syndrome. Janke et al ⁶ based their diagnosis on clinical features (anemia) and microscopic observations of hemophagocytosis. It is worth pointing out that Blümich et al ² noted an indirect evidence of hemophagocytosis (occasional hemosiderin-containing macrophages) in a subset of their humanized mice. There is a lack of consensus on the utility of hemophagocytosis as a reliable marker of HLH in human medicine as occasional/incidental hemophagocytosis may be seen in non-HLH patients. ⁵ As a result, associated laboratory data are considered to be essential to making this diagnosis. ⁷ Similarly, differentiating hemophagocytic histiocytosis from other types of histiocytic proliferations in NSG mice may present a diagnostic challenge.

The pathogenesis of the granulomatous/histiocytic hyperinflammatory lesions remains unexplored in both articles, ^2,6 but the authors offered instructive and pathophysiologically plausible arguments. Blümich et al ² proposed that the inflammatory lesions may represent features of both graft-versus-host disease (GVHD) and HLH. They argued that inflammatory infiltrates targeting the skin in a few of the animals were consistent with features of GVHD while the multisystemic granulomatous infiltration (with rare hemosiderin-laden macrophages) was suggestive of HLH. The early onset and morphology (multisystemic granulomatous lesion without epithelial/tissue damage) described by Blümich et al ² were indeed atypical for chronic GVHD, which generally has a delayed onset and presents as a fibro-inflammatory lesion that often targets the skin, lung, and liver. ^8,14 Janke et al ⁶ persuasively argued that the predominance of myeloid-derived cellular proliferation (mast cells, eosinophils, and macrophages including multinucleated giant cells), hemophagocytosis, and absence of tissue (epithelial) damage disfavored GVHD and was more in line with HLH/MAS, as previously described in NSG-SGM3 mice. ¹⁹ Janke et al hypothesized that the NSG strain background of the NSG-SGM3 mice is already prone to the proliferation of histiocytes and a mild degree of hemophagocytosis, and that the addition of the myeloid cytokine transgenes triggers the HLH/MAS-like syndrome. They ruled out the usual potential infectious instigators for HLH including Epstein-Barr virus (EBV) and cytomegalovirus (CMV) in their study cohorts. ⁶ These viruses are among those that have been associated with acquired/secondary HLH/MAS in human patients, and a humanized mouse model of EBV-associated HLH has been described. ¹² Other methodological approaches, such as immunophenotyping of the T cell repertoire and cytokine quantification, would have provided further insight into the pathogenesis of these inflammatory lesions in these humanized mice, as recently reported. ¹⁹ The pathogenesis and the clinicopathological outcome in this study could have been impacted by multiple variables affecting the engrafted cells or tissues (source, storage, and ex vivo manipulation) and the mouse (strain, genetics, irradiation and other manipulations). ⁹ However, the limited number of study cohorts may not have allowed parsing these and other factors in the study by Janke et al. ⁶

The second striking and novel finding in NSG-SGM3 mice reported by Janke et al ⁶ was mast cell hyperplasia in the pancreas, which the authors attributed to the human SCF that can potentially cross-activate the murine SCF receptors. Mast cell hyperplasia, however, exhibited interesting peculiarities between naïve NSG-SGM mice and those engrafted with human cells with regard to the origin and distribution of mast cells. Mast cells in the naïve NSG-SGM mice were of murine origin and periductular in distribution, whereas in the manipulated cohorts they were of human origin and infiltrative throughout the pancreas. The basis and relevance of this preferential homing of mast cells to the pancreas and their disparate distribution between the naïve and manipulated NSG-SGM mice may signify a pathobiological significance yet to be discovered. In this regard, the putative role of mast cell distribution and activation in human pancreatitis is an intriguing possible connection. ^4,18 Hence, one cannot help wondering if the peculiar proliferation and interstitial homing of human mast cells in the pancreas of humanized NSG-SGM3 mice would be a harbinger of pancreatitis had the mice survived longer. For now, the authors ⁶ perceptively alerted us to a diagnostic conundrum that may arise from potential misinterpretation of these human mast cells as other cell types because they do not resemble the typical mouse connective tissue mast cells.

The third feature of NSG-SGM3 mice engrafted with human cells was their predisposition to develop eosinophil hyperplasia, attributed to the combined action of IL-3 and GM-CSF from the transgene and IL-5 from implanted human cells or from the mouse itself. ⁶ Janke et al proposed that all of the lesions in their NSG-SGM3 mouse cohorts were attributed to the potential of engrafted human stem cells to differentiate into mast cells, eosinophils, and macrophages. They provided an interesting and plausible explanation as to the potential origin of precursor cells from the respective humanization approaches employed in their study, namely, hCD34+ HSCs, non-transduced hCD34+ HSCs in leukemia models, and potential stem/progenitor cells within the leukemia patient-derived xenograft (PDX). ⁶

The findings described in these 2 article ^2,6 will be relevant to pathologists and scientists alike because these changes potentially complicate or confound histopathological interpretation and experimental results obtained from these strains and their ever-increasingly refined versions. One anticipates the most obvious impact may be on preclinical cancer and infectious disease studies for therapeutic safety and efficacy endpoints. For instance, what would be the effect of the immunoinflammatory phenotype on tumor growth and response to therapy, notably immune-based therapies? There is evidence showing a difference in immunotherapeutic response against tumor xenografts in NSG mice versus their humanized cohorts. As a case in point, therapeutic antibody that targeted programmed cell death protein 1 (PD-1) was shown to cause significant growth inhibition of tumors in humanized NSG but not in nonhumanized cohorts. ^11,16 What may be the role of the inherent hyperinflammatory phenotype in the humanized NSG in modulating tumor response/therapeutic outcome? What are the potential on- and off-tumor effects including target modulation/expression by the tumor and nontumor tissues, immune cell trafficking, hypercytokinemia, and the other potentially numerous variables that may muddle the predictive potential and translatability of therapeutic outcomes from these humanized platforms? Presumably these complications are likely amplified in preclinical evaluation of cellular immunotherapies, such as chimeric antigen receptor-modified T (CAR-T) cell therapy, which in effect constitutes another layer of humanization. It may be equally problematic, if not more, if these models are to be used for preclinical safety studies; hence their use should be based on scientific need and a strong understanding of their limitations. For example, NSG mice may not be appropriate for the safety study of CAR-T therapy because of potential off-tumor on/off-target effects that may make the results of questionable relevance to humans and because GVHD may drive T cell expansion independent of target engagement. Equally, one could expect that this hyperinflammatory phenotype may also compromise preclinical studies of pathogens and infectious agents in humanized mice. Another layer of complication is the fact that viral infection, as described above, is one of the most frequent triggers for HLH. For example, SARS-CoV-2 has recently been shown to cause HLH in a subset of patients with severe COVID-19. ¹⁰ Hence, veterinary pathologists and investigators need to be aware of these potential confounders as these humanized mice are increasingly employed for modeling viral infections including SARS-CoV-2.

Notwithstanding some of the shortcomings, mainly ascribed to the retrospective and hence descriptive nature of the studies, the articles by Blümich et al ² and Janke et al ⁶ provided information that will be invaluable for pathologists and bioscientists working on humanized mice models. Their findings have practical relevance given the ever-increasing expansion and use of humanized mice and the continued efforts to enhance and refine their “humanization.” Both articles present evidence, supported by cogent argument and striking photomicrographs, that humanization has a price and more humanization comes with a heftier price. Awareness of these unintended immunological consequences of humanization would inform the cost-benefit and/or cost-effectiveness tradeoffs that come with using humanized mouse models in preclinical safety and efficacy studies.