Thanks for the invitation to discuss our paper at the #NephJC discussion group. As a clinician who started my lab based PhD only 2 years ago with no prior laboratory experience, I can absolutely relate to anyone who might feel a little overwhelmed by scientific articles.
On that note, thanks also to Divya who has done a wonderful job of summarising the article and obviously doesn’t feel at all overwhelmed!! Rather than provide another summary of the findings, I thought I’d try and expand on the basis for the study, some of the intricacies of stem cell culture and future directions of the field in the hope this will make other similar articles more accessible for the group.
Stem cell disease modelling in lay terms.
For over a decade now scientists have been able to take somatic cells (for example a skin fibroblast or white blood cell) and turn them into pluripotent stem cells. These are called induced pluripotent stem cells or iPSC’s. iPSC’s possess key desirable properties for scientific purposes:
- They resolve the ethically contentious need to harvest cells from of a human embryo.
- They are self-renewing with a high rate of proliferation and tolerate liquid nitrogen storage and thawing quite well, which means that once you’ve made them, you essentially have an endless supply of them.
- The can be generated from any patient.
- You can turn a stem cell into any cell type of the body, so long as you have good recipe (or protocol). This is called directed differentiation.
Since the discovery of iPSC (DOI:10.1016/j.cell.2007.11.019), groups around the world have generated protocols for the directed differentiation of iPSC to many different cell types of interest including neurons, hepatocytes, intestine, cardiomyocytes, retinal cells and many others. The directed differentiation of stem cells to a kidney tissue was made difficult by virtue of the fact that kidney tissue is complex, with over 25 known epithelial and interstitial cell types, each with its own highly specialised functions arranged in in a highly organised structure essential for normal function. However our group found that dissociating and re-aggregating the culture early in the process allowed the cells to self-organise into nephrons adapting a technique previously used to form chimeric kidney organoids from two dissociated mouse embryonic kidneys (doi:10.1038/nprot.2016.098).
All of these in vitroprotocols are founded on a sound understanding of the in vivo embryonic development of the target cell or tissue type. In short, we use commercially available growth factors to manipulate signalling pathways to mimic growth factor gradients within specific parts of the human embryo aiming to replicate the path of the human embryonic stem cell to kidney. Whilst the hype of this technology has revolved around the regeneration of tissues for the purpose of autologous transplantation (and we are starting to see this approach in clinical practice - eg. corneal or bone marrow transplants), it is felt we are at best a few decades away from this realisation for regenerating kidney. A far more realistic short term application for iPSC derived kidney tissue is in disease modelling.
Disease modelling aims to replicate a model of a human disease in vitro for the purposes of learning the molecular basis of disease and novel therapeutic discovery. Traditionally, this involves using animal models or non-renal human cellular models. However there are inescapable differences in signalling pathways and cell function between species and even between cell types within the same organism. Regenerating living human kidney tissue from easily accessible patient fibroblasts represents an exciting development for the study of inherited kidney diseases and allows the study of a patient’s kidney disease using their own kidney cells, without then need to biopsy their kidney.
“[this technique] allows the study of a patient’s kidney disease using their own kidney cells, without then need to biopsy their kidney”
What are the important things to consider when reading papers about stem cell science?
As for all new technologies, we are likely to see an exponential rise in scientific research using stem cells. The stem cell field has developed a number of preferences/standards when assessing the quality of stem cell base research. Here are a few of the key factors to consider:
The Starting Material
How were the iPSC reprogrammed? How were they gene edited? How were they maintained?
There are many commercially available human derived iPSC lines available. For example, all of the work within the Takasato et al publication (establishing the kidney protocol we used) was done using a cell line called CRL1502, clone C32. This was reprogrammedto pluripotency by the supplier from the fibroblasts ofa foreskin sample. The benefit of these commercially available cell lines is that they have a well validated genome and well validated pluripotency at the time of purchase. There are even commercially available iPSC cells lines available which have been derived from patients with common genetic conditions (eg. ADPKD).
Alternatively and iPSC can be made in house relatively easily from any patient somatic cell. There are multiple methods of reprogramming a somatic cell to pluripotency. The forced, transient expression of 4 key pluripotency genes is sufficient to induce pluripotency. This was originally done by Sendai virus or lentivirus transfection which integrated these genes into the genome of the host cell, but newer systems involve transfection of an episomewhich allows the transient expression of these genes without disturbing the host genome (called ‘footprint free’). Less disruption of the cell genome is obviously desirable when studying a genetic disease and minimises the chance of a new finding being related to such a genomic disturbance rather than the gene of interest.
It is generally accepted that microarray karyotype is sufficient to rule out major genomic imbalances in newly developed iPSC cell lines and all cells must be proven free of Mycoplasma infection which is a recognised hazard in cell culture and can cause significant disturbances in cell behaviour and expression.
Stem cells self-replicate quickly. They can fill a culture plate within a few days and tend to get very unhealthy (or dead) if they are too tightly packed for too long. So when we’re working with stem cells we have to intermittently break up the cell layer and replate a fraction onto a new plate – this is called passaging the cells. The more a stem cell line is passaged, the more likely it is to develop a spontaneous genetic errors and/or lose pluripotency. For this reason, a scientist must keep a record of the number of times a cell line is passaged. There is varying opinion in the field as to when a cell line gets too old and this will vary between clones. A cell line beyond passage 25 is getting questionable.
The simultaneous reprogramming and gene editing performed by Sara Howden in this paper is desirable because it saves time (each process individually takes about 3 months and costs between AUD $10-15K) and is more likely to yield clones which are clonal (ie. not chimeric) for the genotype of interest. This simultaneous approach is currently only possible using fibroblasts because fibroblasts are large and can be plated individuallywith reasonable viability. When an individual fibroblast is successfully reprogrammed it changes morphology from a single large cell to a small colony of very small cells (a clone). These clones can be easily recognised as they arise, transferred to a new plate, expanded and then screened by Sanger sequencing for the gene edited genotype. Stem cells, on the other hand, don’t like being alone, they like to hang out in groups. When gene editing is attempted on a stem cell line (as opposed to a fibroblast), it is more likely that a colony will arise from a pair or trio of cells with varying gene edited status – producing a chimeric clone where some cells are edited and others aren’t. (doi:10.1038/nprot.2018.007)
Finally, stem cell culture methods have come a long way over the last 5 years. Where previously iPSC cells had to be cultured on a layer of irradiated mouse embryonic fibroblasts in media containing uncharacterised foetal bovine serum (producing all manner of variability and contamination) now they are cultured in ‘feeder free’ conditions and fully characterised media without serum. This is preferred as it improves standardisation and reproducibility.
The End Product
How certain are we that the cultured kidney is kidney?
This is really important as the applicability of the findings hinge on the experiment yielding and isolating the right cell type from the organoid before they move on to further analysis.
Whilst it is tempting to think of cell type as a categorical variable, it is actually a continuous variable. Just like some stem cells can be more pluripotent than others, a cultured proximal tubule cell (for example) will have a variable degree of proximal tubular identity during its differentiation protocol. Most current protocols will start with a period of between 4 and 14 days of growth factor treatment and then leave the cell to do its own thing in a base media without growth factor support. In this context, a proximal tubular cell will most likely reach a peak quality of proximal tubular identity before dedifferentiating away from the target cell identity. Any stem cell paper needs to demonstrate that efforts have been made to identify the point at which the differentiated cells are most like the target cell of interest. Therefore I maintain some scepticism about protocols that differentiate to kidney beyond about 4 weeks of in vitro culture.
It is reasonable to assume that all differentiation protocols will produce some cells within the organoid that are not the cell or tissue type of interest. We are quite up front about this with our kidney organoids which often grow their kidney tissue around the edge and unidentified, usually stromal, material in the middle. There are various methods of minimising this off-target tissue, but I feel strongly that efforts should be made to isolate the cell type of interest before further experiments are performed. This is why we sorted our organoids for the epithelial cell marker EPCAM. Failure to do this adequately could result in artefactual differential gene expression (essentially false positive findings resulting from a different spectrum of off-target cell types between the organoids).
Finally, cell identity in these cultures is sufficiently precarious that removing cells from the 3D micro-environment in which they developed is also likely to lead to a degree of dedifferentiation. When I have attempted to reculture epithelial cells sorted from my organoids to a 2 dimensional culture format, the result has been a heterogeneous mess of cells with various morphologies and no identifiable cell identity. Potential best standards for confirmation of cell type identity include transcriptional comparison to human foetal tissues either by bulk RNA sequencing or single cell RNA sequencing, which has become very popular over the last few years.
Variability Vs Reproducibility
It is a basic principle of science that any finding must be replicated at least once in order to be publishable. One might expect that a differentiation experiment will yield the same result every time it is executed. However variability in cell number, cell media or growth factor batch, time, temperature and manual handling are just some of the long list of potential variables between individual experiments. An understanding of how a protocol endpoint varies from one differentiation to the next is critical to the accurate interpretation of results.
We were fortunate that Prof Little had the foresight to generate a variability dataset for our protocol ahead of the IFT140 experiments which demonstrated and characterised the greatest variability in gene expression occurs between organoids cultured at different times and the narrowest variability between organoids cultured at the same time. (https://doi.org/10.1101/238428) One major strength of our report is the bioinformatics approach to control for this variability. This served to separate transcriptional variation from the genetic difference between the two clones themselves. I believe this sets a precedent for the field which will be an expectation for future publications.
Isogenic Controls
Historically in disease modelling, genetically mutated cells or animals have been compared to unaffected or unrelated genetically ‘normal’ controls. One only needs to examine the whole exome variant list of one patient to recognise the enormous number of genomic variants that exist between variants. An isogenic control iPSC clone refers to a clone which is assumed to be genetically identical at every locus except for the gene edited locus of interest. This is rapidly becoming (arguably already is) the minimum standard in the field.
Multiple Clones
Ideally, a scientific finding will be reproduced in multiple iPSC clones representing each genotype. This is to prove that the finding has not arising from an idiosyncrasy arising during the reprogramming or gene editing process which is specific to that clone. These are called biological replicates.
In our paper we were forced to use technical replicates (a triplicate of experiments performed on one pair of iPSC clones) as we only yielded one successfully gene corrected stem cell clones from almost 50 screened colonies. This is an acknowledged limitation of the study.
These are just some of the many principles of stem cell science that will hopefully help you to critique similar papers in the future.
What is transcriptional profiling?
Genomic sequencing is for DNA what transcriptional profiling is for RNA. The terms ‘RNA sequencing’ and ‘transcriptional profiling’ are often used interchangeably. I would suggest, however, that RNA sequencing is a scientific technique which provides a dataset for transcriptional profiling (which also encompasses bioinformatics analysis).
Pure RNA can be isolated from tissue using a commercially available extraction kit. Using a separate kit, that RNA can be reverse transcribed into DNA (called copy DNA or cDNA). This means, if the tissue was transcribing gene A five times as much as gene B, there will now be five times more gene AcDNA in the sample compared to that for gene B. These samples are now run through a high throughput DNA sequencer (the same used to sequence a genome with a slightly modified protocol), and we can work out how much each gene is being expressed based on how many copies of the cDNA exist in the sample. It is effectively running a quantitative PCR for all genes in the genome simultaneously.
Using a variety of bioinformatics approaches, comparison of the normalised gene counts can be compared between two samples to determine an increase or decrease in gene expression. One can generate lists of the most differentially expressed genes which can be input into online curated analysis tools (for example ToppGene: https://toppgene.cchmc.org/). Such tools seek to identify cellular structures, pathways and processes associated with those genes.
Is there any relevance of this work to clinical practice?
For now, the lessons learned from this work remain of academic significance only and will not feed back to the patient from whom the original stem cells were derived. However, as a proof of concept, this study establishes that when you make iPSC from a patient’s cells and differentiate those iPSC to kidney organoids, the organoids are capable of replicating the patient’s disease. We have also demonstrated that computer based analysis can be used to analyse outputs from the organoids.
It is my hope, that further automation of stem cell based kidney culture platforms and analyses will identify the disease causing pathways leading to novel treatment targets for inherited kidney diseases. Indeed, Prof Beno Freedman’s laboratory in Seattle has further developed this approach, describing robot-assisted culture, staining and analysis or CRISPR edited ADPKD iPSC-derived organoids.
As discussed in the paper, there is also a role for gene correction of candidate novel variants in patients with genetically undiagnosed renal disease with a view to novel variant and gene discovery/validation. The challenge here is to find the assay for the disease within the organoid. We were blessed with the ability to examine the ciliary morphology in the IFT140 case but it may not always be this straightforward.
From a clinical point of view, this understanding of the pace of discovery and innovation in the field has allowed me to approach patients with inherited kidney diseases with more optimism.
So what next?
There are many areas in which organoids could be further developed to improve their disease modelling power. There is no blood flow within the organoid, and we know the vascular endothelium contributes the glomerular basement membrane. This makes glomerular and GBM diseases more difficult to study using organoids. As a result of the absence of a blood supply, there is no flow within the organoids tubules. Signalling within the primary cilium is thought to be altered by mechanic stress produced by tubular flow. How this might alter the transcriptional profile of the epithelial cells remains to be seen.
Kidney organoids are transcriptionally equivalent to trimester one or two human fetal kidneys and lack many of the transporters seen in the mature human kidney. It is not known whether iPSC-derived kidney organoids from patients with adult onset genetic disease will be as faithful as we have demonstrated in this paper for an infant-onset phenotype.
Many of these limitations could be overcome by the transplantation of human iPSC derived kidney organoids into animals where the animal’s blood supply has been demonstrated to invade the organoid’s glomeruli and even produce tubular flow (https://doi.org/10.1016/j.stemcr.2018.01.041). A balance between in vitro organoid techniques, in vivo organoid techniques and traditional animal models are likely to be complimentary disease modelling platforms as the field moves forward.
I look forward to the discussion. Thanks again for involving me.
Dr Tom Forbes MBBS FRACP (@kidney_tom)
Paediatric Nephrologist, Royal Children’s Hospital, Melbourne, Australia
PhD Candidate, Kidney Disease, Development and Regeneration Group, Murdoch Children’s Research Institute, Melbourne, Australia