Refers to Ackermann A. M. et al. GABA and artesunate do not induce pancreatic α-to-β cell transdifferentiation in vivo. Cell Metab. https://doi.org/10.1016/j.cmet.2018.07.002 (2018).

Loss of functional β-cell mass is the key event in the pathogenesis of type 1 diabetes mellitus (T1DM) and also contributes to type 2 diabetes mellitus (T2DM)1. Similar to neurons, human β-cells have very low proliferative capacity2, and the goal of being able to regenerate the lost β-cell mass in patients with diabetes mellitus remains a distant prospect.

In 2013, Nouha Ben-Othman and co-workers showed that genetically induced downregulation of Arx, a key transcription factor for glucagon-producing α-cell specification and phenotype, in α-cells converts them into insulin-producing β-like cells in vivo3. Departing from these findings, the team performed a screen to identify compounds that caused a similar outcome and found that the neurotransmitter GABA inhibits Arx expression in α-cells4. They then combined lineage tracing, transmission electron microscopy, immunohistochemistry and functional tests (including GABAA receptor blockade) to demonstrate that long-term (at least 4 weeks) administration of GABA in vivo to mice ‘convinced’ α-cells to convert into β-like cells. Of note, GABA treatment induced two rounds of β-like cell neogenesis following two cycles of streptozotocin-induced β-cell death, twice reverting severe streptozotocin-induced diabetes mellitus.

In parallel experiments, performed at two independent laboratories, it was shown that GABA converted human and rat α-cells into β-cells (the rat experiments were performed blindly at Decio L. Eizirik’s laboratory); these experiments were not as conclusive as the mouse experiments owing to the lack of lineage tracing4. In the same issue of Cell, another independent study found that antimalarial drugs from the artemisinin family (particularly artemether) also induced the conversion of α-cells into β-like cells by enhancing GABA signalling in vivo5. These effects, first observed in a mouse cell line, were then confirmed in vivo in mice (using lineage tracing), rats and zebrafish. Artemether also improved insulin secretion in human islets.

These exciting observations could open the way for clinical trials with GABA or artemisinins to restore the lost β-cell population in patients with T1DM. The immune system would probably attack these newly formed β-cells, but GABA could be a promising adjuvant therapy for new (we are still in the process of understanding why the immune system attacks β-cells6) therapies to re-educate or mildly suppress the immune system7.

In short, based on these data, the translation of a series of elegant basic research findings into the clinic felt as though it was just around the corner. However, later in 2017, another article indicated that artemether did not convert mouse α-cells into β-cells over a 72 h follow-up and, of particular concern, induced de-differentiation and dysfunction of mouse β-cell and, at least for some experiments, human β-cells8. Furthermore, a subsequent mouse study indicated that neither GABA nor the artemisinin artesunate induced mouse α-cell to β-cell transdifferentiation in vivo9. In this study, Ackermann and colleagues used in vivo genetic α-cell lineage tracing with a tamoxifen-inducible Glucagon–CreERT2 system, which labels >90% of α-cells and enables accurate quantification of mature α-cell to β-cell transdifferentiation. The investigators treated mice for 3 months with intraperitoneal GABA injections9 at the same concentrations used by Ben-Othman and colleagues4. At the end of this period, the authors reported no changes in blood levels of glucose or insulin as determined by intraperitoneal glucose tolerance test and found no indications of α-cell to β-cell transdifferentiation above basal levels9.

In the four papers described above there are some obvious differences in experimental models used, which might contribute to explanation of the opposite findings. For example, all the experiments conducted by Ackermann and colleagues9 were performed in a single normoglycaemic mouse model, while the study by Ben-Othman and co-workers4 included experiments in which GABA was used to revert diabetes mellitus in mice, and the findings were validated in two other species, namely human and rat. In addition, the tamoxifen-induced lineage-tracing model used by Ackermann and colleagues9 aims to follow transdifferentiation of adult α-cells into β-cells, while the constitutive Glucagon–Cre mice used by Ben-Othman et al.4 can also detect β-cell neogenesis via a putative and transient cell state where glucagon is expressed. Furthermore, different mouse strains were used in the two laboratories; however, this should not be an issue as humans are not inbred and a novel treatment that is robust enough to deserve to be tested in patients with diabetes mellitus should induce similar beneficial effects in different mouse strains. Moreover, the GABA used in the different experiments was not originally made for in vivo use and major differences in biological effects have been observed between batches. Finally, diet and housing conditions are different between the laboratories, which could affect the murine microbiome (the microbiome is always a prime suspect) and consequently drug metabolism and systemic mouse metabolism.

To solve this controversy and clarify whether GABA and related compounds will move into clinical trials in patients with diabetes mellitus, we suggest a collaborative effort between two (although more would be ideal) laboratories that have access to validated mouse models for lineage tracing of α-cell to β-cell conversion as well as human islets (Fig. 1). These laboratories should use the same GABA preparation for all experiments and preferably use GABA that is produced under good manufacturing practice conditions (such is the case with GABA for future human use) that have been validated by Patrick Collombat’s laboratory (as Collombat described the phenomenon) or by another laboratory involved in the experiments. The laboratories should use exactly the same protocol to treat normoglycaemic or streptozotocin-induced diabetic mice with GABA or saline for 3 months, and evaluate body weight and glucose tolerance (with measures of plasma insulin) and immunofluorescence analysis of the pancreata to determine the extent of α-cell to β-cell conversion. This last and crucial step should be done on coded samples in an independent expert laboratory.

Fig. 1: Investigating the effect of GABA on β-like cell transdifferentiation from α-cells.
figure 1

Presently, the results on whether GABA induces transdifferentiation of α-cells into β-cells are conflicting. This lack of clear conclusions prevents investigators from conducting clinical trials of GABA — a drug that might be effective for regenerating lost β-cell mass — in patients with diabetes mellitus. We propose a multicentre study to determine whether GABA induces α-cell to β-cell transdifferentiation in mouse and human cells.

Full size image

In addition to these steps, 500 human islets, ideally from the same donors (that is, islet isolation would be performed at a single place and distributed in parallel to the laboratories), should be transplanted under the kidney capsule of immunodeficient mice and subsequently treated with GABA, as described previously4. Importantly, the investigators should perform at least 4–5 independent experiments with islets from different donors. At the end of the treatment phase, an independent laboratory should count the percentage of glucagon-positive and insulin-positive cells in coded samples. If lineage tracing of human islet cells becomes available (for example, using lentiviral vectors) in the meantime, this approach should be incorporated into the protocol.

These proposed experiments will require a sincere desire to unveil the truth, as well as marked efforts and funding. In our view, and considering the potential clinical translation, they are worth doing.