Using brain organoids to understand Zika virus-induced microcephaly

The rapidly advancing field of stem cell biology continually provides new insights into basic biology and human disorders, as well as innovative approaches to develop new patient treatment strategies. In the past 10 years, human pluripotent stem cells (hPSCs) have emerged as an invaluable tool in modeling human disorders, especially those with complex genetic origins that are challenging to model in animals (Takahashi et al., 2007; Wen et al., 2016). Significant progress has been made in the targeted differentiation of hPSCs into various cell types with high purity in order to study human cell biology and develop cell replacement therapies. A new frontier of stem cell research is the generation of three-dimensional (3D) tissue structures to model organogenesis and developmental disorders. Human brain organoids are stem cell-derived 3D tissues that self-assemble into organized structures that resemble the developing human brain (Jo et al., 2016; Kadoshima et al., 2013; Lancaster et al., 2013; Mariani et al., 2015; Paşca et al., 2015; Qian et al., 2016). The ability to engineer brain organoids from genetically tractable patient stem cells holds tremendous potential to transform our understanding of human neural development, which is otherwise inaccessible to manipulation and detailed analyses. In this Spotlight article, we highlight the use of brain organoids to study Zika virus (ZIKV) infection, identify advantages and limitations of brain organoids as a model system, and discuss perspectives for future studies.

ZIKV, a mosquito-borne flavivirus, has reportedly spread in over 70 countries and territories globally (Box 1) (CDC, 2016; Heukelbach et al., 2016). While ZIKV infection in adults usually results in mild symptoms, much attention has been drawn towards the co-occurrence of ZIKV outbreaks and an increased incidence of newborns with microcephaly, a condition in which infants are born with an abnormally small head. Since the declaration of ZIKV as a global health emergency by the World Health Organization (WHO), clinical examinations of microcephalic fetal tissues have shown the presence of ZIKV in damaged fetal brains (Heymann et al., 2016; Mlakar et al., 2016; Ventura et al., 2016). More recently, the United States Centers for Disease Control and Prevention (CDC) evaluated existing evidence and concluded that ZIKV causes microcephaly and other severe fetal brain defects (CDC, 2016). Live infected human fetal tissues are not accessible and postmortem tissues are variable in their quality and genetic backgrounds, and clinical studies alone cannot provide sufficient insights for understanding how ZIKV causes this damage. Therefore, researchers have adopted brain organoid models to study the cellular tropism and pathogenesis mechanisms of ZIKV in controlled settings (Ming et al., 2016).

Brain organoids include at least two categories, namely cerebral organoids and brain region-specific organoids (Kelava and Lancaster, 2016). The cerebral organoid system utilizes the intrinsic signaling of hPSCs to allow spontaneous differentiation into neural tissues that resemble features of different brain regions (Lancaster et al., 2013). The cerebral organoid offers an opportunity to model whole brain development and interactions among different brain regions. Brain region-specific organoids, by contrast, use patterning factors to induce the differentiation of hPSCs into specific lineages, such as cerebral cortex (Kadoshima et al., 2013; Mariani et al., 2015; Paşca et al., 2015; Qian et al., 2016), midbrain (Jo et al., 2016; Qian et al., 2016) and hypothalamus (Merkle et al., 2015; Qian et al., 2016; Sakaguchi et al., 2015; Wang et al., 2015). Compared with cerebral organoids, brain region-specific organoids model individual brain regions of interest and generally result in more uniform and reproducible tissue, providing a platform for quantitative characterization.

Despite the different methodologies, a common characteristic that distinguishes brain organoids from other in vitro models, such as monolayer and neurosphere cultures, is that, in addition to producing the relevant cell types, brain organoids are capable of recapitulating much of the architecture of the fetal brain (Fig. 1). For example, cortical organoids contain well-organized neural progenitor cell (NPC) layers and neuronal layers, resembling the laminated structure found in the cerebral cortex (Kadoshima et al., 2013; Paşca et al., 2015). This provides a unique opportunity for direct investigation of structural phenotypes, such as the relationship between ZIKV infection and cortical layer thickness. Embryonic primate and human brains exhibit a well-defined outer subventricular zone (oSVZ) layer containing outer radial glial cells (oRGCs), the major progenitors believed to be responsible for primate and human cortical expansion (Lui et al., 2011; Nowakowski et al., 2016b). One forebrain organoid model recapitulates the progenitor zone organization with an oSVZ layer formed by specialized NPCs displaying the distinct molecular signature of human oRGCs (Qian et al., 2016). oRGCs are few in number and do not form a discrete layer in rodents, highlighting the potential of brain organoids to recapitulate human-specific developmental features that cannot be studied in animal models. Forebrain organoids also exhibit functional cortical neuron subtypes found in all six cortical layers, as well as diverse interneuron subtypes (Qian et al., 2016). In addition, astrocytes are present in brain organoids (Paşca et al., 2015; Qian et al., 2016).

Compared with monolayer and neurosphere cultures, which usually contain relatively pure populations of one or a few cell types, brain organoids better recapitulate the composition, diversity and organization of cell types found in the developing human brain (Fig. 1). By exposing the whole organoid to ZIKV, it is possible to analyze and compare the efficiency with which ZIKV infects different cell types, layers and regions within the same organoids, as well as to investigate potential non-cell-autonomous effects. In addition, compared with 2D differentiation protocols, brain organoids of different stages can more accurately mimic the developing trajectory of fetal brain within the first and second trimesters of pregnancy. Because these spatiotemporal transitions take place as a continuous process, researchers can examine brain organoids at any time point of interest to gain insight into both short- and long-term effects of viral infection over the course of fetal brain development. Therefore, in the context of a global health emergency, such as the ZIKV outbreak, brain organoid systems can provide direct insight into human neural development and expedite therapeutic progress.

The first study to model ZIKV infection during human brain development used monolayer cultures of forebrain-specific NPCs, which provided the initial hint that ZIKV more efficiently infects NPCs over hPSCs or immature neurons (Tang et al., 2016). A number of studies have since modeled ZIKV infection using human NPC monolayer and neurosphere cultures, acute fetal brain slices, and brain organoids (Barrows et al., 2016; Cugola et al., 2016; Dang et al., 2016; Garcez et al., 2016; Hanners et al., 2016; Liang et al., 2016; Onorati et al., 2016; Qian et al., 2016; Retallack et al., 2016; Simonin et al., 2016; Wells et al., 2016) (Fig. 1). An in-depth analysis of forebrain organoids transiently exposed to ZIKV at different developmental stages showed that ZIKV exhibits tropism towards NPCs, including ventricular RGCs and oRGCs, over intermediate neural progenitors or immature neurons (Fig. 2) (Qian et al., 2016). This finding was later confirmed in studies of fetal human brain slices in vitro (Onorati et al., 2016; Retallack et al., 2016) and embryonic mouse brains in vivo (Li et al., 2016; Nguyen et al., 2016). Infected NPCs become viral factories to produce more infectious viral particles, leading to an increasing number of infected cells over time (Qian et al., 2016). In addition to the reduced overall size of forebrain organoids, quantitative analysis revealed that ZIKV infection caused depletion of NPCs by at least two means: suppression of NPC proliferation and increased cell death of both infected NPCs and uninfected neurons. This resulted in a decrease in volume of both NPC and neuronal layers, resembling microcephaly (Fig. 2). Notably, ZIKV infection-induced NPC death was much more dramatic in forebrain organoids than in monolayer cultures (Qian et al., 2016; Tang et al., 2016).

Infection of human neurospheres and cerebral organoids with ZIKV and dengue virus 2 (DENV2) showed that ZIKV, but not DENV2, attenuates NPC growth, suggesting that the detrimental consequence of ZIKV infection is not a general feature of the flavivirus family (Garcez et al., 2016). Indeed, a recent transcriptome analysis of infected monolayer NPCs showed that ZIKV has a more selective and substantial impact on the expression of genes involved in DNA replication and repair, in contrast to the robust global transcriptome changes induced by DENV (Zhang et al., 2016). Transcriptome analysis of ZIKV infection in human cerebral organoids and mouse neurospheres also identified upregulation of toll-like receptor 3 (TLR3), an innate immune receptor (Dang et al., 2016). Analysis of gene expression changes highlighted TLR3-mediated downregulation of neurogenesis and upregulation of pro-apoptotic pathways. Functionally, a TLR3 inhibitor partially rescued the reduced size of ZIKV-infected neurospheres and organoids, and it will be interesting to further determine whether cell viability and organoid structures are also rescued.

Initial studies in the field used the only available strain at the time, the original African strain of ZIKV (ZIKV^M), which shares 87-90% RNA sequence identity with the Asian strain (ZIKV^C) and recent clinical isolates from Brazil (ZIKV^B) and Puerto Rico (ZIKV^P) (Faria et al., 2016). One study combined mouse models and cerebral organoids to compare the effect of infection by different ZIKV strains. Compared with ZIKV^M, ZIKV^B appeared to exhibit stronger deleterious effects in cerebral organoids and in embryonic mice, including more severe effects on NPC depletion and neuronal layer disruption (Cugola et al., 2016). Transcriptome analysis of monolayer NPCs infected with different ZIKV strains showed similar overall expression profiles, yet ZIKV^C, but not ZIKV^M, induced upregulation of viral response genes and TP53 (Zhang et al., 2016). Functionally, TP53 inhibitors can block the apoptosis induced by both ZIKV^C and ZIKV^M in human NPCs, with higher potency against ZIKV^C-induced apoptosis. It should be noted that any strain-specific phenotypes in experimental models need to be considered in the context of the passage histories of the strains. For example, ZIKV^C has been passaged extensively in suckling mouse brain and then in vitro, whereas ZIKV^B and ZIKV^P have only been passaged a limited number of times in the laboratory.

Within a short period of time, brain organoids have been applied in multiple laboratories to investigate ZIKV infection (Ming et al., 2016). Consistent results from these studies have consolidated the causal link between fetal ZIKV infection and birth defects. Despite these advances, it is important to acknowledge challenges when applying current organoid technologies to ZIKV research. The first challenge is the limited accessibility of organoid technologies. Methodologies for generating monolayer NPCs and neurospheres are well-established and commercially available, but culturing organoids requires specialized expertise and it would take extensive training and resources for a laboratory specializing in virology to adopt organoid technology. Although commercialization of organoid technology is unlikely to fulfill the urgent need for addressing the current ZIKV emergency, researchers world-wide have overcome this challenge by establishing extensive collaborations outside of their own discipline, as the majority of studies published so far on ZIKV organoid modeling are collaborative efforts involving multiple laboratories that would not normally overlap in their topics of research.

The second challenge is the heterogeneity that accompanies the structural complexity and cell type diversity offered by brain organoids over simpler models, such as neurospheres (Kelava and Lancaster, 2016). In cerebral organoids, spontaneous differentiation of PSCs determines composition and properties, leading to considerable sample-to-sample, batch-to-batch and cell line-dependent variations. These variations present obstacles for quantitative analysis and reproducibility of results. Moreover, owing to the long-term nature of organoid culturing, small perturbations in culture conditions could potentially change the outcome significantly. While efforts have been made to generate brain region-specific organoids under defined conditions, it is important to establish common standards of organoid quality by carefully characterizing multiple aspects, including lineage specificity, proportion of different cell types and, in the case of cortical organoids, the relative size of each layer.

A third challenge in applying organoid technology to ZIKV research is the low-throughput nature of culturing and analyzing organoids. This represents a major obstacle to conducting drug screening and other high-throughput applications. Although technologies such as miniaturized bioreactors offer promise for scaling up organoid cultures with reduced cost and increased throughput (Qian et al., 2016), current methods for structural and cellular characterization have relied on labor-intensive procedures, including cryosectioning, immunostaining and image collection. For large-scale screening involving hundreds of conditions, new methods are needed to process these organoids in batches and to automate readouts to obtain quantitative data.

A final and perhaps most difficult challenge in brain organoid research is their simplistic nature compared with the actual developing human fetal brain. All current reported studies evaluate direct exposure of ZIKV to brain organoids in the absence of immune and vascular systems, which might skew the actual effects of ZIKV infection in humans. In the future, new and more sophisticated models, which might involve the addition of endothelial cells or microglial cells to organoids, or that add blood-derived components to the medium, might provide more physiologically relevant features for modeling ZIKV infection during fetal human brain development.

Most published studies have so far focused on characterizing the pathological impact of ZIKV infection. Brain organoids can be easily manipulated, both genetically and chemically, offering a valuable platform for the next phase of ZIKV-related research. There are at least two major directions: (1) to delve into the underlying molecular and cellular mechanisms responsible for ZIKV-induced damage; and (2) to use organoids as a screening and validation tool for therapeutic development.

Genetic manipulation of brain organoids can be achieved via stable genome editing of hPSCs, or transiently via electroporation or retrovirus/lentivirus/adeno-associated virus infection of organoids directly. Therefore, it is possible to examine how an individual or combination of ZIKV-encoded proteins, or ZIKV-derived noncoding RNAs, affects brain organoid development (Liang et al., 2016). Insights into downstream molecular events will enhance our understanding of how ZIKV interacts with the host machinery to suppress NPC proliferation and induce cell death, as well as help to identify genes and pathways necessary for ZIKV replication (Ming et al., 2016). These studies will not only provide potential therapeutic targets against ZIKV, but might also enhance our general knowledge of human brain development.

Although current brain organoid systems have yet to meet the demand for full-scale screening of hundreds to thousands of conditions, they provide a relevant platform for testing specific hypotheses and for validation of therapeutic candidates. For example, the observation that the AXL receptor tyrosine kinase protein is highly expressed in human RGCs and oRGCs has led to the hypothesis that it could function as a ZIKV entry receptor in NPCs (Nowakowski et al., 2016a). To test this hypothesis, AXL was deleted via genome editing in hPSCs and the resulting cerebral organoids showed no difference in ZIKV infection and pathophysiology, suggesting that AXL is dispensable for ZIKV infection in human NPCs (Wells et al., 2016). In another example, a recent study screened over 6000 drug candidates and approved drugs using monolayer NPC cultures and identified a number of hits that were then applied to forebrain organoids to test drug efficacy and to evaluate potential toxicity (Xu et al., 2016). The same principle applies to validating antibodies against ZIKV for vaccine development. Whereas most screens are conducted by measuring one or two specific parameters, characterization of brain organoids can provide comprehensive insights into the effectiveness of therapeutic candidates to rescue the effects of ZIKV at molecular, cellular and structural levels.

Studies using brain organoids derived from hPSCs have contributed substantially to a growing body of knowledge on the effects of ZIKV infection on fetal brain development (Ming et al., 2016). Brain organoids have proven to be an invaluable model system for studying human cortical development and diseases in general. With future improvements, the brain organoid system presents a unique and comprehensive platform to investigate human-specific developmental features as a continuous process in a highly relevant context, and holds great promise for therapeutic development.

We thank Dr. Kimberly M. Christian for comments.