The incidence of type 1 diabetes is increasing worldwide and pancreatic islet replacement has emerged as a therapy, especially in the case of recurrent severe hypoglycemic events [1–4]. Transplantation of donor islets, obtained by mechanical and enzymatic dissociation of the pancreas, provides sustained improvement of glycemic control with an efficient prevention of severe hypoglycemia in the large majority of recipients, thus improving patients’ quality of life [5]. Moreover, islet transplantation has been demonstrated to prevent the progression of chronic diabetes complications [6–8].

Islet graft function can be assessed as clinical outcome by several methods, such as the β-score [9] or the Igls criteria [10]. The islet grafts’ potency can be assessed in vivo by transplanting a set fraction of the islet preparation into immunodeficient rodents, however, the read-out is retrospective [11, 12]. We are still lacking predictive criteria for evaluating islet quality immediately prior to transplantation [12], an issue that has been called for early on [13]. This issue will also be of considerable importance for potential future therapies using stem cell-derived surrogate islets [12].

Therefore, we investigated whether donor islet quality could be ranked according to their electrical activity. Considerable knowledge in the electrophysiology of human islets has been acquired during the last decade and allows to define meaningful electrophysiological parameters to evaluate their function and establish a ranking score [14]. Indeed, changes in ion fluxes are the first integrative signals of islet activity. Slow potentials (SPs), as recorded here by dynamic multi-electrode arrays (MEAs), reflect the physiological important coupling between islet β-cells, are tightly linked to insulin secretion and exhibit the same biphasic profile, a hallmark of islet activation [15, 16]. Moreover, these SPs are regulated by relevant hormones, such as adrenalin or incretins, and are deteriorated during aging and glucotoxic condition [15, 17]. Finally, these recorded electrical activities can regulate glucose homeostasis in silico in the FDA-approved simulator of glucose homeostasis in type 1 diabetes patients [18, 19]. In view of these characteristics of the recorded electrical islet signatures, we hypothesized that a detailed and dynamic electrophysiological analysis may reflect donor islet quality. Moreover, the use of extracellular electrophysiology applied here only requires routine expertise in cell culture, which is available in most clinical laboratories. The analysis of recorded data can either be automated and performed online or offline with commercial software or after electronic data exchange with expert groups.

Islet signals were recorded using electrodes that measure extracellular islet field potentials due to ion channel activities. In our approach approximately 100 IEQ donor islets suffice to analyze their quality (Figure 1A) after seeding on a commercial microelectrode array (MEA); thus, only a minute aliquot of islets was necessary from a single donor as compared to the 10,000 IEQ per kilogram body weight required for transplantation. The recorded signals were amplified, digitized, and processed in real-time using dedicated hardware that applies a series of filters and detection algorithms to extract SP frequencies, which are representative of islet activity (Figure 1B) [15, 22]. The analysis can be performed directly online in real-time using custom electronics as given in Figure 1B, which provides automated filtering of recordings, detection of electrical signals (events) and feature extractions [23, 24].

As extracellular electrophysiology is non-invasive without rundown, even over several days, repetitive dynamic measures are possible over physiologically meaningful time spans. This allows to test a series of physiological relevant parameters in a dynamical fashion, as opposed to simple glucose-induced increase of islet activity [15, 18, 22]. To rank the performance of donor islets according to SP frequencies, we established a number of criteria and ensuing testing scenario. The score rankings were established prior to actual recordings ranging from CHIP-scores 1 to 6, reflecting the least physiological performance, i.e., glucose insensitivity, to the most physiological performance, i.e., glucose concentration dependency and biphasic activity pattern [15, 17] (see Figure 1C; Supplementary Table S1). The criteria applied here are based on well-known islet physiology [14, 26]: at low glucose, islet β-cells show no or only minor spontaneous activity; the neurohormone epinephrine inhibits β-cell activity, an important feature during physical activity or stress; increasing levels of glucose over its physiological range considerably enhances electrical activity and in the best case, this electrical activity is biphasic; the stimulation by glucose is further augmented postprandially by the incretin hormones, such as GLP-1, at physiological levels of 50 pmol/L and less [14, 15, 22]; and finally, sulfonyl-urea drugs such as glibenclamide stimulate islet β-cells independent from glucose via pharmacological closure of K_ATP channels. The absence of glucose, GLP-1 or glibenclamide induced activity was ranked as least performant with a score of 1 (Figure 1C, red numbers) and those islets were not further investigated. If at least glibenclamide or GLP-1 effects were observed, a score of 2; if responses to GLP-1 and to the stress hormone adrenalin were observed, a score of 3; if glucose concentration dependency was evident only at high glucose concentrations (15 mM vs. 1 mM), a score of 4 was attributed; if glucose concentration dependency was observed over the physiological range of 5.5 mM–11 mM glucose, a score of 5 was given; finally, if biphasic glucose-induced activation was observed, the (highest) score of 6 was assigned. Islets were consequently tested for basal and non-β cell activity using adrenaline known to inhibit β- and stimulate α-cells (Figure 1C; ➊,➋), responsiveness to elevated glucose (Figure 1C; ➌), glucose concentration dependency (Figure 1C; ➍), biphasic activity at physiological glucose concentrations (Figure 1C; ➎), and the effects of drugs such as the sulfonylurea glibenclamide or hormones (GLP-1, adrenaline) on glucose-induced activity (Figure 1C; ➏,➐,➑).

Two MEA recordings are given in Figure 1D (black traces) as examples of lowest (CHIP-score 1; maximum frequency 0.095 ± 0.014 Hz) and highest CHIP-score (CHIP-score 6; maximum frequency 0.409 ± 0.034 Hz) and their corresponding β−scores are provided in Figure 1D. Recordings from electrodes not covered with islets were provided (Figure 1D, blue traces) and show the high signal-to-noise ratio of the MEA approach. Recordings with the lowest CHIP-score (1; upper panel) showed neither clear glucose-dependency nor any effect of GLP-1 or even glibenclamide and were not submitted to further tests. Recordings with the highest CHIP-score of 6 (lower panel) passed successfully steps 1 (as shown), as well as 3 and 4 (traces not shown) and exhibited a clear biphasic increase in SP frequency in step 5 (as shown). The peak of the first phase occurred after 5 minutes, in-line with reported electrical behavior of human islets [14, 15, 17] and exhibited even typical 5–10 min oscillations in the second phase.

The Islet CHIP study included 8 recipients from the STABILOT clinical islet transplantation study (4 men and 4 women; mean diabetes duration, 34 ± 11 years; Table 1; Supplementary Table S1). The evolution of primary graft function evaluated using the β-score prior to the second infusion is shown in Figure 2A. Groups of CHIP-score vs. β-score are apparent: patients who improved rapidly after the 1st infusion and further before the 2nd infusion (Figure 2A, green), patients who improved only partially after the 1st infusion (Figure 2A, blue), and patients who changed little after the 1st infusion and progressed little afterwards (Figure 2A, orange). There was a correlation between the β-score established prior to the second islet infusion and the CHIP-score (Figure 2B; ρ = 0.51, p = 0.1), as well as between HbA_1c levels and CHIP-score; ρ = −0.556, p = 0.08) but statistical significance was not attained in either case. Islet purity, GSIS or total amounts of insulin secreted did not correlate with the β-score, HbA_1c level, or CHIP-score as reported previously. CHIP-score did not correlate with main donor criteria such as age, sex, BMI, cause of death, warm or cold ischemia time. These latter parameters as well as volume or amount of IEQ infused were also not correlated with β-scores.

The most relevant score in terms of success is given by the β-score and final clinical outcome will evidently depend on numerous parameters including recipients’ characteristics. It is thus reasonable to expect that donor organ quality may only be one of the many factors influencing the clinical outcome [27]. Obviously, determining the quality of the main therapeutic agent remains a major issue as in any clinical intervention. Function of human islet transplanted in nude mice correlates well with clinical outcomes. Although this highlights the importance of quality of islets to be transplanted, the read-out of this assay is only retrospective [11]. New approaches to quality control are required and they may also provide means for better comparison of results between transplantation cohorts in view of the diverging criteria applied in donor selections [28].

Previously, a number of islet parameters were tested for their predictive value in animal transplantation studies [29]. Mitochondrial markers have been reported as significant indicators in the case of allo- or auto-transplants [30, 31]. However, islets were generally of lower purity, and the clinical endpoints used were either insulin dependency or independence, which is quite different from the currently used scaled β-score. Studies on the size of islets used for transplantation have not shown any correlation with clinical outcome [32]. Investigation of donor long noncoding RNA repertoire revealed that MALAT1 expression predicts the quality of human islets prior to their isolation [33] but a potential correlation of this signature with clinical outcomes has not been published.

Similar to our study, insulin secretion of donor islets prior to transplantation has been reported to correlate poorly, if at all, with outcome in animal transplantation studies [29, 34]. In contrast to GSIS, the CHIP-score described here relies on combining a proven technology [15, 17, 18, 22, 23] with parameters such as a range of glucose concentrations, as compared to a single concentration of high glucose, relevant hormones and direct assessment of K_ATP-channel function, central to islet activity. Moreover, the potential clinical relevance of the electrical signals recorded here by MEAs is underscored by the observation that they can be used in an FDA approved simulator of human metabolism of patients afflicted by type 1 diabetes to control in silico glucose homeostasis via insulin delivery [18, 19]. For those reasons the static evaluation of insulin secretion by GSIS may not provide sufficient details on islet function, including physiologically relevant parameters, such as biphasic activity and β-cell coupling [14]. Additional parameters as used here in the electrophysiological characterization (hormones and detailed glucose concentration dependency), may per se also be determined in classical dynamic secretion assays. However, this would considerably increase the workload and costs as compared to an automated electrophysiological analysis. Moreover, dynamic measurements of insulin release do not inform per se about the important physiological parameter of islet β-cells, which is coupling. Finally, the MEA technology required is fully compatible with the expertise of a clinical laboratory.

As expected from a small sample sized pilot study, the limitation of our study is the absence of statistical significance, despite a good correlation, that may also reflect the influence of multiple confounding clinical factors.

In conclusion, to the best of our knowledge, this pilot study is the first to correlate donor islet functional quality and clinical outcomes prior to human allotransplantation. Biomimetic potency tests have been strongly advocated for islet transplantation, and recent progress in islets-on-chip may provide solutions [12, 35]. Our observations indicate a potential usefulness of our islets-on-chip system presented here in evaluating islets before grafting and might consequently improve clinical outcomes. The approach used here may also be developed as microfluidic device thus further reducing the number of islets required [15, 23, 36, 37]. In the long run, such a system might also qualify for evaluation of stem cell-derived pseudo-islet organs prior to their implantation [12].