An epithelial organoid model with Langerhans cells for assessing virus-host interactions

Abstract

Persistent infection with oncogenic human papillomavirus (HPV) may lead to cancer in mucosal and skin tissue. Consequently, HPV must have developed strategies to escape host immune surveillance. Nevertheless, most HPV infections are cleared by the infected host. Our laboratory investigates Langerhans cells (LCs), acting at the interface between innate and adaptive immunity. We hypothesize that this first line of defence is vital for potential HPV elimination. As an alternative to animal models, we use smaller-scale epithelial organoids grown from human primary keratinocytes derived from various anatomical sites. This approach is amenable to large sample sizes—an essential aspect for scientific rigour and statistical power. To evaluate LCs phenotypically and molecularly during the viral life cycle and onset of carcinogenesis, we have included an engineered myeloid cell line with the ability to acquire an LC phenotype. This model is accurately tailored for the crucial time-window of early virus elimination in a complex organism and will shed more light on our long-standing research question of how naturally occurring HPV variants influence disease development. It may also be applied to other microorganism–host interaction research or enquiries of epithelium immunobiology. Finally, our continuously updated pathogen–host analysis tool enables state-of-the-art bioinformatics analyses of next-generation sequencing data.

This article is part of the theme issue ‘Silent cancer agents: multi-disciplinary modelling of human DNA oncoviruses’.

1. Introduction

Since the late 1970's, there has been an evolution of growing human keratinocytes in a three-dimensional (3D) rather than two-dimensional manner for ‘the reconstitution of living skin’ [1] in a cell culture dish with nomenclatures such as ‘organotypic culture’ [2], ‘keratinocyte raft cultures’ [3], ‘organotypic raft cultures’ [4] or most recently skin ‘organoids’ [5] (figure 1). Initially, the main driver for in vitro skin cultivation was grafting. To grow keratinocytes as stratified epithelium was first reported by Rheinwald & Green [6], followed by Bell's ‘full thickness skin equivalent’ [1,7,8]. In the early 1990's, primary keratinocytes infected with high-risk (HR) human papillomavirus (HPV) were found to have premalignant characteristics in organotypic culture [9] and various models capable of reproducing the HPV infectious cycle were established by several independent groups [10–12]. Globally, HPV—a DNA virus, is the most common sexually transmitted infectious agent with approximately 10% prevalence in healthy women [13] and more than 300 types known to date (the majority of the 500+ types of human and non-human PVs identified [14], based on the papillomavirus episteme, PaVE: https://pave.niaid.nih.gov). A subset of 12 HR types, with HPV16 being the most common, cause cancer (e.g. oropharyngeal and gynaecological) in humans [13,15]. The malignant potential of HR HPVs is largely owing to the E6 and E7 oncogenes encoding their corresponding proteins which interfere with cellular integrity ([16] and references therein). HPV16 variants implicated in a higher risk for cervical cancer [17] showed a higher degree of dysplasia in 3D raft cultures compared to lower risk variants [18–20]. The next generation of rafts included immune components to study the role (or lack thereof) of innate and adaptive immunity in combatting HPV. For instance, models including lymphocyte infiltration [21] and the microenvironment of Langerhans cells (LCs) [22]—epithelium-specific dendritic cells (DCs), were developed. More recently, a bone marrow-derived cell line called MUTZ-3 [23] has been used to produce DCs and LCs [24]. Such a cell line allows far better reproducibility in a research context where many biological replicates are necessary for robust research results with high statistical power.

Figure 1.

Historical overview of epithelial models. Significant developments in the evolution of three-dimensional epithelial models (focusing on the incorporation of Langerhans cells and HPV16 variant research) in the last four decades are depicted along with potential future directions. Owing to space constraints we are unfortunately unable to showcase all the excellent model studies that exist.

HR HPV such as type 16 is the main risk factor for cervical cancer provided it can persist in the host [25]. Variant designations are based on their geographical region of origin ([26] and references therein). The European prototype (EP) was the first HPV16 genome published more than 30 years ago [27] and at present, four (formerly five) lineages are known: A (European sub-lineages A1–A3 and Asian sub-lineage A4), B (African-1 sub-lineages B1–B4), C (African-2 sub-lineages C1–C4), and D (North American sub-lineages D1 and D4, and Asian-American (AA) sub-lineages D2 and D3) [26,28]. Recently, we have put our efforts on the common EP and AA variants which differ in only three amino acid changes at residues 14 (Q > H), 78 (H > Y) and 83 (L > V) in the major transforming protein E6 [18–20,29,30]. Epidemiological studies revealed that the AAE6 variant is a higher risk factor for dysplasia as well as an earlier onset of invasive tumours than EPE6 [31,32]. AAE6 has a greater transforming, migratory and invasive potential than EPE6 when retrovirally transduced into primary human keratinocytes during recent long-term in vitro immortalization studies [18,29]. Furthermore, AAE6 is more prone to integrate into the host cell genome [20] and demonstrated an altered metabolic phenotype reminiscent of the Warburg effect [30]. These results suggest that coding changes in E6 have strong mechanistic and functional consequences for infection and thus contribute to marked differences in cancer risk.

To decipher the fundamental biology of HPVs and their tumourigenic features in a model system, the organotypic 3D infection model (raft culture, or organoid, the latter will be used henceforth) has the advantage of allowing reproducible and simultaneous epithelial differentiation and therefore the occurrence of an active viral life cycle. Our approach is a joint venture between the biological, clinical and computer sciences, with like implications for clinical and basic research. In addition to how E6 variants are implicated in the development of HPV-related diseases, this model has been developed to adapt to new enquiries regarding how the early cell-based innate immune system fights a common virus such as HPV. Here, we discuss an organoid epithelial model drawing on previous research and experience by us [18–20] and others [1–12,21–24]. Its strength lies in the unique combination of components essential for our application: primary, HPV-permissive host cells from various anatomical sites, the recapitulation of the HPV infectious cycle using full-length HPV16 genomes, naturally existing HPV16 variants throughout its genome, controllable copy numbers of HPV-positive keratinocytes, tissue-residing LCs, phenotypical and immunological characterization as well as refined bioinformatics tools for next-generation sequencing (NGS).

2. The ‘silent killer’: how human papillomavirus evades host immune recognition

HPV, the ‘silent killer’, is a master in evading the host immune response, causing illness and in some cases death of the infected host. The Indigenous, supernatural monster Windigo comes to mind serving as a metaphor when our group explained the danger of HPV to First Nations women in Northwest Ontario, Canada [33]. Here, we outline an eclectic, experimental approach appropriate to study HPV infection. We will employ healthy mucosal or skin tissue including epidermis and dermis either from the uterine cervix (non-keratinizing, cervical keratinocytes), the oropharyngeal area (non-keratinizing, gingival keratinocytes) or the skin (keratinizing epithelial keratinocytes) with matching fibroblasts. Fortunately, these cells can now be bought, avoiding the barrier of lengthy procurement processes [34]. Epidermis on its own or combined with dermis models are also commercially available. However, our context precludes the use of such models since donor background cannot be controlled and to ‘infect’ with HPV, we need to grow 3D cultures to allow the viral life cycle to take place for the production of infection-competent virions. Our laboratory has the necessary experience with organoids, as we can rely on a decade or so of in-house experience in 3D culturing [18–20,29,34,35].

Under normal physiological conditions, infiltrating T-cells from the underlying blood vessels migrating into the dermis and even the epidermis has not been observed. However, immature and mature LCs belong to the normal mucosa and skin landscape and hence only their integration in our current model has been considered. Rodrigues-Neves & Gibbs [36] pointed out that a 3D keratinocyte model with both immune components is still lacking in the scientific literature. This is clearly a draw-back when researching various allergens where inflammation is the biggest obstacle to tackle. However, in the context of persistent HPV in immune-comprised organs such as the cervix, the sheer lack of these components may provide major viral immune evasion strategies to escape host immune surveillance. Indeed, HPV is thought to keep a low inflammatory profile—meaning that immune cells are not attracted to the site. Without causing epithelial inflammation, HPV has developed many strategies to go undetected, both during innate and adaptive phases [37].

(a). The human papillomavirus clearance hypothesis: the role of Langerhans cells

LCs are a distinct DC subset comprising about 2–3% of epidermal cells [38] that are at the interface between innate and adaptive immunity. They reside in the supra-basal part of the epithelium and their adhesion is mediated by E-cadherin [39]. Keratinocytes are important for LC activation and maturation as they modulate immunity through their production of chemokines and pro-inflammatory cytokines. We hypothesize that in hosts clearing their HPV infection (the majority) [40], keratinocytes stimulate LC activation, e.g. via TNF-α and TGF-β secretion which promotes LC migration from the epidermis to dermis and lymph nodes. Melief [41] depicts such a scenario where ‘danger signals’ and ‘activated dendritic cells' lead to the activation of cytotoxic T-cells—the first step towards tumour elimination. While we generally agree with this notion, further details were added to illustrate our model and adapt it to the viral infectious cycle and early carcinogenesis (figure 2).

Figure 2.

Epithelial microenvironment and immune landscape. HPVs infect epithelial tissues lining the upper digestive or genital tract most likely by infecting the stem cells of the proliferating basal layer. If the viral genome persists in the basal layer, then it can be amplified in the differentiating layers. Thereafter, new virions are released without causing an inflammation. Langerhans cells (LCs) reside in the epithelia as immunosurveillance cells. Many parameters (markers) to investigate LC patterns of activation, differentiation and migration from epidermis to dermis (and potentially to draining lymph nodes) are known. They will be closely investigated in the depicted model in the context of the two HPV variants under study. Viral life cycle will be confirmed, e.g. with viral capsid proteins L1/L2. The epithelial phenotype will be assessed via appropriate keratin (K) markers in the various anatomical sites, e.g. K5/10 in the skin, and keratinocyte immune markers for an environment commensal for LC epidermis to dermis migration, e.g. MHC I/II, E-cadherin, TNF-α and TGF-β. We will further test markers of epidermal LCs, e.g. MHC II, CD11c, CD207 (langerin), EpCAM and markers of LC migration ability, e.g. MHC II, CD28, CD40 and CXCR4 and CCR7 for a two-step migration. The above markers will all be tested in situ by immunofluorescence. Monitoring early HPV infection by the host needs danger signals (cytokines and chemokines) to attract LCs travelling to sentinel lymph nodes. Such LC attractants, e.g. CCL2/5/20, and CXCL12 (fibroblast-derived) and CCL27/28, IL-18 and type I IFNs (KC-derived) will be characterized via supernatant. In collaboration with the Alizon group in Montpellier, we use mathematical models to discover novelties/unknowns, test new hypotheses and research questions in silico [42].

LCs are patrolling antigen-presenting cells (APCs) acting as immunological sentinels in mucosal and skin tissues. They probably engulf and phagocytose viral particles and/or dead HPV+ keratinocytes before a lesion develops and may be the first encounter of an infected host to fight an HPV infection. In turn, keratinocytes secrete pro-inflammatory cytokines and dermal fibroblasts secrete chemokines/chemo-attractants, helping LCs to mature and migrate to lymph nodes for cross-presenting HPV peptides to CD4+ and CD8+ T-cells. LCs, however, seem to be less frequent in the transformation zone [43] where most cervical cancers develop, and E6 negatively interferes not only with E-cadherin but also with the expression of LC chemo-attractants ([44] and references therein). It is this phase of the immune system that we think is one of the key interfaces between pathogen and host, and which the virus must escape to persist in its host. Interestingly, the oncoprotein E6 of HPV16 downregulates E-cadherin ([45,46] and references therein), and an E6 variant with only one amino acid change of AA at residue 83, showed this to be increased compared to EP [45]. This, and the potential role of LCs in HPV clearance, prompted us to investigate LC immune functions in the AAE6 variant context.

(b). Implication of HPV16 variants in modulating the host immune system

In the past decennium and outlined in the introduction, our laboratory has provided ample evidence that E6 variants within the HPV16 genome strongly promote functional changes in the mammalian host. More recently, we discovered that the highly oncogenic AAE6 variant may also be involved in immune escape like the L83V variant (see §2a). In our given context, other than promoting the differentiation of LCs, and in contrast to keratinocytes, TGF-β stimulates stromal fibroblasts to proliferate and to synthesize matrix proteins [47]. An interesting finding from the Jackson et al. [20] RNA-seq data is worth exploring further: in addition to one of the significantly downregulated clusters of genes being involved in the ‘negative regulation by host of viral transcription’, we also found that the TGF-β innate immune pathway signature seems to be downregulated in the AAE6 but not in the EPE6 organoids. This is intriguing since TGF-β is the key cytokine for LC maturation. TGF-β was expressed approximately one-third lower in AAE6 versus HPV-negative and EPE6-containing organoids. Although this is not significant when considering the differential expression results which use a conservative filtering technique (given the high-throughput nature), the similarly listed ‘TGF-β induced’ (TGFBI) is significantly lower. TGFBI, an extracellular matrix (ECM) protein, promotes inflammation, integrin-mediated monocyte adhesion, migration and chemotaxis [48]. TGFBI downregulation by AAE6 may be a means to further tone down inflammation in the HPV environment. Moreover, innate immune receptor signalling [49] may also be affected by the AAE6 variant. In a recent protein–protein interaction screen, we found that AAE6 but not EPE6 binds to various E3 ligases (other than E6AP), indicating that innate immune-regulated transcription factors (e.g. NF-kβ and IRF7) as well as anti-microbial peptides (e.g. defensins and cathelicidins) and pro-inflammatory cytokines (e.g. IL-1β) may be downregulated (Mehran Masoom 2018, unpublished observations). Lastly, our finding that AAE6 interferes with host cell metabolism by shifting to a Warburg effect [30] has also been looked at from the perspective of tumour-associated macrophages (TAMs) in the stroma [50]. Interestingly, it was reported that tumour-derived lactic acid caused the M2-like polarization in TAMs. Altogether, we conclude that in an immunological context, the AAE6 variant may indirectly modify LC biology as well as the tumour microenvironment in the stroma.

3. An eclectic methodological approach for an epithelial organoid

We will perform systematic organotypic epithelial culturing [19,20] in a stepwise manner through three primary experimental phases: preparation, cultivation and characterization. The first phase provides a foundation focused on a priori experimental design for reproducibly answering biological questions in a ‘life-like’ model of human epithelium. The second phase continues with the process of growing (figure 3) and harvesting these laboratory-grown tissues. Finally, the third phase concludes with thoroughly characterizing the tissues via biological analysis and interpretation. While additional components and complexities (e.g. beyond LCs) could be included in an organoid model to make it ever more ‘life-like’, we strive to present a simplistic yet useful approach with a focus on studying HPV biology (and specifically, the molecular underpinnings of increased tumourigenic risk owing to viral variants, such as via LC interactions). The overall objective is to establish an immune-competent 3D in vitro organoid to study two commonly found HPV16 E6 variants undergoing their active HPV life cycle. We will determine an initial host–pathogen interaction, i.e. the suppression by HPV at the innate immunity level—for which the life cycle needs to be activated. Our previous model, even without LCs, fulfilled closely the needs of in vivo but prevented us from controlling the number of HPV+ cells after transfection [19,20]. The new model, on the other hand, can be adapted to such needs as we are able to use various ratios of HPV+ and – cells.

Figure 3.

Flow diagram of the epithelial model. To culture immunocompetent organoids: on day 1, in a 24-well plate, fibroblasts are embedded in a collagen matrix as a dermal equivalent that supports keratinocyte growth. On day 2, keratinocytes (containing HPV episomes) and Langerhans cells are seeded on top of the dermal equivalent at a ratio of 1 : 1. On day 3, the rafts are lifted and transferred to a six-well plate with a membrane insert to create an air–liquid interface exposing only the dermal equivalent to media. Subsequently, the rafts are fed for two weeks with a media change every second day. This environment encourages differentiation and stratification. With our experimental model, we strive to mimic what happens in vivo. With this model, the viral life cycle can be propagated, and this is only possible when cells are grown in three dimensions as organoid cultures. The fully grown epithelium depicts mucosal, non-keratinizing epidermis and dermis.

(a). Preparation of the revised model

Here, we will use the entire HPV16 AA variant genome's single-nucleotide polymorphisms (SNPs) (i.e. altogether approx. 150 not just the three SNPs within the E6 gene as done previously) [19,20]. Using the Cre-Lox recombination system [51,52] with subsequent selection potentially all transfected cells will carry the full HPV16 genome, which allows us to control the number of HPV+ cells within the 3D tissue. An alternative approach will also be used to increase DNA transfection efficiency threefold, i.e. from approximately 10% using chemical transfection to approximately 30% with electroporation [53] prior to selection. We have found a way to create smaller 3D cultures enabling us to have more biological replicates. This will increase the effect and sample size for more robust and reliable statistics. For LCs, we will use the MUTZ-3 cell line [23] as this allows experimental reproducibility and avoids donor variability when using DCs derived from peripheral blood monocytes [54]. We performed experiments to assess the suitability of this line using flow cytometry to identify a ‘differentiated’, double positive population (langerin/CD1a) (figure 4a,b). Another group has used this approach successively, albeit in a context of skin allergens and irritants [24].

Figure 4.

Characterization of the epithelial model. Langerhans cells (LCs) can be derived from MUTZ-3 cells via a cytokine-mediated differentiation over approximately 10 days (a). Morphologically they appear more irregular and clustered together than their progenitors and differentiation success can be verified via flow cytometry for cell-surface markers CD1a and CD207 (langerin) as well as immunofluorescence for CD207 (b) and adhesion molecule E-cadherin. In addition to LCs, a major focus of our research is to study HPV16 genome variants, using full-genomes which contain all naturally occurring polymorphisms, and to introduce these into our epithelial model. In this example case, near-diploid immortalized keratinocytes (NIKS, [55]) are co-transfected with LoxP-flanked HPV genomes and a Cre-recombinase expression plasmid to yield selected populations of keratinocytes containing viral DNA (c) and able to undergo a viral life cycle when grown as epithelial organoids (evidenced by capsid protein production, L2). As well, epithelium can be further characterized for proliferation markers (such as Ki-67) and cell-cycle dysregulation markers owing to E7 expression (p16^INK4A) (d). European prototype (EP) genomes were designed to contain a LoxP site in either the SphI or PmlI restriction site, both in the non-coding upper regulatory region of the viral genome.

The first step of our epithelial organoid approach will be to prepare the experimental model by carefully designing experiments with the research question at the forefront. This includes first establishing the hypotheses to be tested, defining the experimental variables (dependent and independent), and performing sample size estimations through power calculations based on expected effect size and variability (which can be informed via previous work or preliminary ‘pathfinder’ experiments). Biostatistics are one facet of our eclectic approach, and while under-represented in past literature, may be applied to help resolve reproducibility concerns in cancer biology. When estimating sample sizes to appropriately test hypotheses, the concept of ‘biological independence’ is important to consider. While truly independent replicates would be altogether uniquely derived biological samples (such as different donor individuals), it can be helpful to think of biological independence as a spectrum ranging from these uniquely derived specimens on one end, to increasingly more feasible options, albeit with a proportional loss in true biological independence. For epithelial cultures, we suggest that the ‘level of interrogation’ should be the main consideration when determining the level of biological independence required. For example, with our research question focused on viral variants and their differential tumourigenic risk to the host, the level of interrogation would be at the interface between virus and host, such as the introduction of viral genomes into keratinocytes via transfection (chemical or physical). So long as the research question is focused on the difference between the viral variants, a simple yet effective design would be to replicate the experiment with independent transfections using the same donor pool of cells (controlling this background if possible, to some extent, given inherent variability in passaging cells over time). If the research question was broader, by including donor/host variability, then the level of interrogation would correspondingly be at the host-level, requiring unique donors such as via patient-derived samples [34]. Once the experimental design has been established, the next stage of the approach is to perform material calculations and establish the organoid cultivation time-course.

This includes growing the required host cells (in a humidified incubator at 37°C and 5% CO₂) either with the intention of incorporating into organotypic epithelia (such as keratinocytes, fibroblasts and Langerhans cells differentiated from the MUTZ-3 myeloid leukaemia cells), or in support of those cells (such as 5637 bladder carcinoma cells and J2/3T3 mouse embryonic fibroblasts). While these cells vary in their media requirements, adherence (monolayer versus suspension cultures for the immune cells), and culture technique, an important consideration is the tissue origin and donor-background of the cells used, and whether they are primary or immortalized, and whether they will be matched for an experiment (e.g. gingival fibroblasts with gingival keratinocytes to model the oral epithelia). We have tried near-diploid immortalized keratinocytes (NIKS, [55]), primary human foreskin/epidermal keratinocytes (PHFKs or HEKs) and primary human oral/gingival keratinocytes (HGKs). Also, we have generated epithelial organoids from patient-derived cervical biopsies (via suspected lesions at colposcopy, [34]). Primary cervical cells (both keratinocytes and fibroblasts from the uterine cervix) are now commercially available. The decision on which cells to use may be relevant due to tissue region susceptibility differences [56], tissue microenvironments, and variations in signalling affecting differentiation and possibly the viral life cycle and immune environment. Components of the dermal equivalent (typically collagen and fibroblasts), making up the ECM may also be relevant factors.

Finally, beyond the host components, the sourcing and preparation of the viral genomes is at the crux of our experimental design and research question. Viral genomes have previously been from isolates, where a small number of modifications (e.g. in the E6 gene) can be introduced via mutagenesis, but now it is also possible to synthesize whole genomes, including all desired SNPs, or intentional modifications/deletions, based on reference genomes. Using gene synthesis and cloning, it is also possible to introduce LoxP sites which are useful for Cre-LoxP-mediated transfections as a method of introducing viral genomes along with selection genes into host keratinocytes (figure 4c). Using this technique, we performed a proof of principle study where we achieved an active viral life cycle using synthesized HPV16 EP whole genomes with LoxP sites in the SphI restriction site (previously used by others [51,52]) as well the alternatively chosen PmlI site (with a previously unknown effect on the viral life cycle), further upstream in the upper regulatory region and outside of potential transcription factor binding sites (figure 4d). Synthesized HPV16 EP whole genomes both yielded an active life cycle, with the PmlI LoxP site performing as good if not better than the default SphI location.

(b). Organoid cultivation and characterization

While many past studies have focused on the cultivation of epithelial organoids (figure 1), including detailed methodology papers that serve as a good foundation and provide more thorough descriptions of the process, as well as video tutorials [4], it will now be important to expand the perspective to different experimental design considerations (such as multi-component rafts) and a variety of endpoints and time-courses. Recently, novel methods for studying the HPV life cycle have been described elsewhere [57].

Different size organoids can be grown, and we have tried small (96-well sized), medium (48-well sized) and large (24-well sized) rafts, where smaller rafts allow for increased sample size for a variety of downstream applications. While being cost-effective (with the exception of plate inserts and per sample materials), there is increased difficulty manually handling smaller rafts and the extraction yields of tissues are lower. Hence, we settled for medium-sized rafts as an optimal balance. As well, different amounts of fibroblasts can be seeded into the dermal equivalent, where we have found that an increased number helps with epithelial differentiation. Varying the number of infected cells within the epithelium could be a way to model varying stages of disease, from low-grade to high-grade lesions (where higher-grade lesions have a greater number of HPV+ cells [58]).

The duration of culturing (typically 1–21 days) can be assessed using time-series rafts, as we have done previously to aid in mathematical modelling [42]. While 14 days is typically used as the peak of the viral life cycle, shorter or longer durations (to a limited extent) may be relevant for assessing changes over time as well as the interaction between differentiation and viral replication, genome amplification, and transcription as they relate to persistence and integration. Prior to harvesting (typically less than 24 h before), the thymidine analogue BrdU can be added to culture media to assess proliferation (where suprabasal proliferation is indicative of keratinocytes infected with HPV). When culturing and harvesting, biosafety precautions are required (e.g. biosafety level II in Canada), as active viral particles may be produced. Harvesting can be done to preserve structure, via fixation (e.g. formalin), or through tissue dissection (e.g. manual separation of the epidermis from dermis) followed by homogenization of the relevant compartment and molecular extractions (e.g. DNA, RNA, protein), or processing for single-cell suspensions. For formalin-fixed and paraffin-embedded tissues, typical histological assessment is performed using haematoxylin & eosin staining, whereas in situ techniques such as immunohistochemistry and immunofluorescence can be used for qualitative and semi-quantitative characterization of host markers (e.g. Ki-67 proliferation marker, BrdU-incorporation proliferation marker, p16^INK4A cell cycle dysregulation and surrogate E7 marker, cytokeratin 5 and 10 differentiation pattern markers, and viral markers (L1 and L2 capsid proteins)). Data and statistical analysis can be performed using open-source software. Extracted and purified molecules, such as nucleic acids (DNA or RNA), can be used for common assays such as viral copy number or viral/host gene expression, or used for more high-throughput techniques such as microarray or NGS.

(c). Immune-competent component

The feasibility of incorporating MUTZ-3-derived LCs into 3D epithelial cultures has been reported previously by the Gibbs group [36]. While LCs normally make up approximately 2–3% of the epithelium, this group used a ratio of 1 : 1 or even 2 : 1 in 3D cultures compared to the number of keratinocytes. This high number was deemed necessary because only a subset of cytokine-treated MUTZ-3 cells differentiate to double-positive langerin/CD1a LCs: e.g. 30–70% (S.W. Spiekstra, Gibbs group 2018, personal communication) or even lower at 10–20% in our hands. Moreover, a proportion of differentiated LCs die or do not attach. Nevertheless, in our preliminary results from monolayer attachability experiments with half, equal or double ratios of cytokine-differentiated LCs versus keratinocytes, we yielded similar proportions of attached LCs (average of approx. 13%). The efficacy may be less in a 3D scenario owing to LCs maturing and migrating out of the epithelium too soon. Instead of compensating this loss with an increased ratio of LCs, our populations will be further enriched for an increased number of differentiated LCs using anti-Langerin conjugated microbeads. The starting population before differentiation will also have to be monitored carefully so that CD14+ and CD34+ proportions are close to equal [59]. Likewise, timely LC maturation, which should be triggered by cytokine production of surrounding keratinocytes and chemokines from dermal fibroblasts will have to be considered. LC maturation may not happen through these ‘natural’ means and instead require a cytokine boost, e.g. of TNF-α and IL-1β and prostaglandin E2, or an allergen or irritant [60].

So far, the outlined LC incorporation approach only allows a qualitative assessment of LC characteristics in the tissue. Hence, a quantitative assay is also needed for our research question to detect any measurable effect between the two variants under study. Therefore, we will use a modified transwell migration assay adding artificial ECM on top of the upper transwell membrane before adding various ratios of LCs and keratinocytes and delay LC maturation for at least two days to measure any effect owing to HPV. Fibroblasts will be seeded into the underlying well. With this system, cell chemotaxis of LCs can be measured like in or ex vivo. Because the transfected HPV+ cells are not yet expected to be tumourigenic, only a few will spontaneously pass through the ECM but LCs are expected to do so if they receive the appropriate signals to ‘mature’ from keratinocytes and fibroblasts.

Finally, it is tempting to use two different methods in parallel: to include LCs in one and omit them altogether in the other. Indeed, even without LCs, we can still test keratinocytes and fibroblasts for markers that render a milieu suitable for LC migration from epidermis to dermis (summarized in [62]). Consequently, both approaches will be attempted using immunofluorescence and bead-based multiplex assays to detect immunological markers (or lack thereof) in keratinocytes and culture supernatant.

(d). Next-generation sequencing and the pathogen–host analysis tool

NGS of extracted nucleic acids (DNA and RNA) allows for a comprehensive analysis of the molecular background of our organoids and the ability to determine differences in HPV16 variant interactions with host tissue [20]. Important considerations include library preparation (whether or not a sequence capture or enrichment step will be used to enrich the low abundance viral sequences relative to the host), sequencing (platform used, short versus long reads, read depth anticipated), as well as bioinformatics analysis pipelines to be used (custom-scripts, high-performance computing cluster access, or desktop tools). Ultimately, next-generation or high-throughput sequencing analysis can enable hypothesis-testing as well as hypothesis-generation via exploration, and possibly spur future research questions and additional cycles of organoid culturing.

To aid researchers with these data, we developed a platform to analyse pathogen–host relationships in NGS data by using industry standard methods while reducing barrier to entry (https://github.com/chgibb/PHAT, [62]). For the PHAT ‘toolbar’, or graphical user interface (GUI), obtained sequence data is added (Input), quality-controlled (QC) and aligned with the appropriate reference sequence; this platform gives the user access to the alignment summary of obtained NGS data (DNA and/or RNA-seq) compared to a pathogen or host reference sequence. Number and percentage of reads that have aligned to the reference are provided, and the user has access to an interactive (scrolling and zooming) linear visualization of viral reads and coverage across a reference genome, which also highlights SNPs that are within reads, compared to the reference, that can be helpful for pathogen genotyping. The Output button gives access to customizable data tables that can be output and saved in preferred spreadsheet format (such as .csv or Excel) for use in other software or for publication/reports. This allows the user to select exactly which samples and columns of data they would like to output, such as quality control information, alignment statistics, SNPs/genotyping, etc. The Genome Builder is a tool within PHAT that can be used to create circular visualizations of pathogen reference genomes (rather than just linear maps for the round HPV plasmids). Viral genes, SNPs within the genes, and circular read coverage (which is important for assessing episomal/integrated forms in the case of HPV) can be added and annotated. We recently added the ability to work with pre-aligned data from large datasets, where users can take output from high-performance computing clusters and work with it on their desktop or laptop computer, and we are currently adding enhanced viral integration detection features (given one of our most pertinent research questions is the pattern of viral integration into host DNA between HPV16 variants). Overall, PHAT is under continued development and users are automatically notified of any updates, which can then be downloaded and installed seamlessly.

4. Limitations of existing research models

While we advocate for a simple yet useful model of human epithelium, we appreciate that achieving this goal has limitations. George E. P. Box's wisdom on model utility can be extended to biological models: ‘Since all models are wrong the scientist cannot obtain a ‘correct’ one by excessive elaboration’ [63, p. 792]. Hence, avoiding excessive elaboration may be important for keeping experiments manageable and focused, but an overly simplistic model may generalize or altogether lack essential components or characteristics that are found naturally to provide insight for the biological phenomenon studied. Box continues ‘… following William of Occam, he [one] should seek an economical description of natural phenomena. Just as the ability to devise simple but evocative models is the signature of the great scientist so [,] over-elaboration and over-parameterization is often the mark of mediocrity’ [63, p. 792]. This limitation can be overcome by adapting the model to include only the essential components to answer the actual research question. In this focused approach, we knowingly exclude parts in our organoid model, such as lymphoid cells, that eventually may become involved in the host immune process but that currently are outside our study frame. With continued advances in high-throughput assays as well as computing power and analytics, it may be possible to simulate human epithelia in silico using mathematical modelling [42]. Such a technique could be a powerful tool in combination with biological models to provide dataset training and verification, for answering a variety of questions pertaining to epithelial responses (e.g. owing to viral infection) and testing hypotheses by modulating key parameters. Finally, advances in single-cell sequencing and in situ analysis may uncover new information about tissue heterogeneity and molecular signatures during a viral life cycle.

5. Conclusion

This commentary describes our concept accurately tailored to the crucial time-window of early virus infection, e.g. HPV elimination (or not) in a complex organism such as humans. We describe a unique, robust and reproducible alternative to animal models to study HPV immune biology during the full viral life cycle. While most other investigations centred around the skin, we and a select few others [60,61] also consider anatomical areas lined by mucosal tissue mostly affected by HPV: the uterine cervix and the oral cavity. Consequently, immune molecules in the skin are well characterized [36] including a global approach [64], while this is still largely lacking for both mucosal sites. Most importantly, HPV variants have not been addressed at all. Our approach will shed new light on host immune evasion by HPV in the context of HPV variants focusing on HPV and its interactions with epithelial LCs. In particular, we investigate the molecular signature of LCs surrounding keratinocytes in the epidermis and underlying fibroblasts in the dermis in the context of a high-risk HPV, i.e. type 16 and what role HPV variants, differently implicated in cervical disease have ([19,20] and references therein). Here, for the sake of creating a reproducible and complex model to study two common, naturally occurring HPV16 variants, we do not wish to include genetic variables (in addition to those found in the viral genomes under study). However, the outlined model is amenable to be expanded to an epidemiologic study with multiple host genome backgrounds involving individual or all three anatomical sites with an appropriate sample size. To bear in mind: preparing organoids the way we describe here requires an interdisciplinary mind set, meticulous preparedness coupled with a good pair of laboratory hands, a large portion of patience while maintaining these cultures, and a great deal of understanding and appreciation for bioinformatics. The lucky, successful candidate will be hugely rewarded with new discoveries.

Data accessibility

The data and materials supporting this article have been described in text or are available upon request.

Authors' contributions

R.J. and I.Z. have contributed equally to conception and design, acquisition of data and literature, and overall interpretation. They were both instrumental in drafting and critically revising the article. S.E. created the final figures, assisted with acquisition and interpretation of data and formatted the manuscript for submission. All authors have contributed to revisions and approve the final version to be published.

Competing interests

We have no competing interests other than a co-authorship with one of the guest editors, Dr Alizon [42].

Funding

This work was supported by Natural Sciences and Engineering Research Council of Canada (NSERC) grants to I.Z. (no. 355858-2008, no. 435891-2013, no. RGPIN-2015-03855), an NSERC Alexander Graham Bell Canada Graduate Scholarship-Doctoral (CGS-D) to R.J. (no. 454402-2014), and a Northern Ontario Heritage Fund Corporation (NOHFC)-sponsored internship for S.E. The funding bodies had no role in study design, data collection, data analysis and interpretation, or preparation of the manuscript.