Autofluorophores Assessed by Hyperspectral Microscopy Indicate Perturbation and Transplant Viability in Pancreatic Islets

doi:10.21203/rs.3.rs-2058969/v1

Download PDF

Research Article

Autofluorophores Assessed by Hyperspectral Microscopy Indicate Perturbation and Transplant Viability in Pancreatic Islets

https://doi.org/10.21203/rs.3.rs-2058969/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background

Islets prepared for transplantation into type 1 diabetes patients are exposed to compromising factors that contribute to early graft failure necessitating repeated islet infusions for clinical insulin independence. Lack of reliable pre-transplant measures to determine islet viability severely limits the success of islet transplantation. We imaged cell autofluorophores using hyperspectral microscopy to characterise loss of viability in islets and define a non-invasive methodology for predicting transplant outcomes in mice.

Methods

Islet viability was compromised by exposure to oxidative stress (menadione), hypoxia (dimethyloxalylglycine), cytokine injury (TNF-α, IFNγ, and IL-1β), or warm ischemia (30- and 60-minute delayed pancreas collection). The impact of islet encapsulation in a conformal coating of hydrogen-bonded poly(N-vinylpyrrolidone)/ tannic acid (PVPON/TA) multilayer film was investigated. LED illumination produced excitation at 358 to 476 ± 5 nm in 18 steps, emission was detected using filters at 414, 451, 575, 594, and 675 ± 20 nm. Syngeneic (C57BL/6Ausb) mice with diabetes (alloxan tetrahydrate) were used to test viability on transplantation. Discriminative analysis and unsupervised principal component analysis were used to differentiate groups. Unmixing of spectral signals to identify component fluorophores was carried out using the unsupervised algorithm Robust Dependent Component Analysis (RoDECA).

Findings:

The autofluorophores NAD(P)H, flavins, collagen-I and cytochrome-C were successfully unmixed. Redox ratio (NAD(P)H/flavins) was significantly increased in islets exposed to ROS, hypoxia, cytokine injury and warm ischemia, typically driven by elevated NAD(P)H. Receiver operating characteristic assessment showed that our models were able to detect; oxidative stress (ROS) (AUC = 1.00) hypoxia (AUC = 0.69), cytokine exposure (AUC = 0.94), or warm ischemia (AUC = 0.94) compared to islets harvested from pristine anesthetised heart beating mouse donors. Significantly, we defined an unsupervised autofluorescent score for ischemic islets that accurately predicted restoration of glucose control in diabetic recipients. Similar results were obtained for islet single cell suspensions, suggesting translational utility in the context of emerging beta cell replacement strategies.

Conclusions

Hyperspectral microscopy of autofluorescence has the potential to give a non-invasive indication of islet viability, prior to transplantation. This would inform clinical decision making and enable patients to be spared transplantation attempts with no potential to reduce their dependence on exogenous insulin.

Cell Survival and Cell Death

Type 1 diabetes

islet

transplantation

hyperspectral

multispectral

autofluorescence

Transplantation of pancreatic islets is effective for patients with T1D and hyperglycaemic unawareness (1), however its efficiency is low, with most patients requiring high islet numbers, over multiple islet infusions (transplants) to become insulin independent (2, 3). Though transplantation failure is multifactorial, it has been linked to islet viability (3) highlighting the need for methods to characterize islet quality prior to transplantation to inform clinical decisions (4). However, common methodologies for assessing islet viability, including the consideration of islet morphology by a manual observer, live/dead assays (chiefly via membrane impermeable fluorescent dyes), ATP/ADP ratio and glucose stimulated insulin secretion (GSIS) are not strongly predictive of insulin independence after transplant (4–9). Hypothesised causes of this include the measurement of living versus dead cells pre-transplantation not capturing apoptotic and pre apoptotic cells, as well as quiescent β-cells recovering activity once reintroduced to physiological conditions (4). Oxygen consumption showed promising accuracy but has not achieved clinical implementation (10).

Inflammation and oxidative damage (reactive oxygen species (ROS)-induced) triggered through islet isolation can impact β-cell metabolism, potentially affecting islet viability and later graft function (11–13). The reduced form of nicotinamide adenine dinucleotide (NADH) and flavin adenine dinucleotide (FAD), the principal electron donors and acceptors of oxidative phosphorylation (14), are key indicators metabolic state. The relative concentrations of these coenzymes – determined by a measure termed the optical redox ratio – has been linked to apoptosis (15), neoplasia (16), and stem cell differentiation (17). NADH and FAD are autofluorophores, with NADH having excitation maxima at 290 and 351 nm and emission maxima at 440 and 460 nm, and FAD having an excitation maxima at 450nm and its emission maxima at 535 nm (18). As such, the relative abundance of these factors can be detected by their defined spectral signatures allowing a real time, and sensitive readout of the cellular redox state which in turn reflects cell viability and function (15–17).

Here we have applied hyperspectral imaging technology to determine islet autofluorescence following isolation under optimal (heart beating anesthetised mouse donor) conditions. A broad light spectrum was analysed to capture data from an extensive cross-section of autofluorescent molecules within the cell – not just NAD(P)H and flavins which are the typical targets of autofluorescent microscopy – giving a deep signature of biological status. Systems that have been characterized by this methodology include cell cycle stage (19), arthritic cartilage (20), kidney function (21), levels of ROS (22), and age (23). The spectral signature of pristine pancreatic islets was then compared to islets exposed to viability-compromising insults – warm ischemia, hypoxia, oxidative stress, cytokine injury – in order to investigate whether a spectral signature could be defined to detect damage. As a secondary objective we unmixed the spectra in order to study changes in specific component autofluorophores. Furthermore, we investigated whether a signature could be developed to predict post-transplant graft function in diabetic recipients (mouse). A non-invasive, label free assay to assess functional viability of islet preparations would provide an evidence-based platform for determining specific islet preparations' likelihood of post-transplant success. This would improve patient experiences by avoiding the burden of unsuccessful procedures and could contribute to increasing the rate of insulin independence in T1D patients whilst reducing the need for multiple islet infusions.

Islet collection and culture

Pancreatic islets from 2–3 C57BL/6Ausb mice (Australian BioResources) per experiment (mixed sex) were harvested according to our protocol (24), exposed to a compromising intervention, then imaged on a hyperspectral microscope. Ethics approval was given by the Garvan Institute of Medical Research Animal Research Authority (20_18). All culture was carried out at 37^oC, 5% CO₂ in islet culture media (24). Interventions included ROS clearance inhibitor menadione (30µM), hypoxia inductor dimethyloxalylglycine (DMOG), pro-inflammatory cytokines (200U/µL TNF-α, 200U/µL IFNγ, and 25U/µL IL-1β), or warm ischaemia (30 and 60 minutes delayed pancreas collection). Menadione treatment was 2 hours of exposure (in culture media) followed by 24 hours culture. DMOG treatment was 0.5mM/L for 16 hours prior to immediate imaging. Cytokine exposure was 4 or 24 hours in islet culture media to induce moderate or major inflammation, respectively. Control islets were maintained in culture media for identical time-courses. Similarly, islets for the warm ischaemia experiment were maintained in culture for 24 hours before imaging. For the ROS, hypoxia and inflammatory models, islets from different mice were pooled prior to being sorted into the treatment groups, ensuring that differences were not the consequence of animal or isolation factors. This was not possible for warm ischemia as the intervention was applied at the animal level. Three mice were used for each treatment group to accommodate this. Disaggregation of islets to individual cells was performed using 0.5 mM EDTA.

Islet encapsulation was investigated to determine whether the hyperspectral technology could still be applied in the context of this strategy for escaping immune detection. A conformal coating of hydrogen-bonded poly(N-vinylpyrrolidone)/ tannic acid (PVPON/TA) multilayer film was applied according to the protocol in (25). First islets were pelleted in 15 mL Eppendorf centrifuge tubes and washed twice with islet culture media. PVPON was allowed to adsorb onto islet surfaces from a 1 mg mL^− 1 solution (RPMI 1640) for 8 min on a circular roller followed by the deposition of TA layer from a freshly dissolved 0.3 mg mL^− 1 solution (pH = 7.4) for 8 min. After each deposited layer, islets were collected by centrifugation for 1 min at 1000 rpm and rinsed with RPMI 1640. Islets were encapsulated with 4 bilayers of (PVPON/TA) with tannic acid on the outer layer. All solutions were filter-sterilised (0.22 µm pore size) with polystyrene non-pyrogenic membrane systems (Corning).

Islet transplantation

Diabetes was induced in 8–10-week-old C57BL/6Ausb mice by intravenous injection of alloxan tetrahydrate (Sigma-Aldrich) (20 mg/ml), 110 mg/kg body weight in injectable grade water. Only mice with blood glucose ≥ 20 mmol/l over two consecutive readings were eligible as transplant recipients. Islets were isolated from the pancreas of donor mice and transplanted into syngeneic recipients (26). Islets were prepared from the pooled pancreata of three donor mice per transplant (ensuring reliably sufficient numbers) and 100 hand counted islets were transplanted into individual recipient mice. Donors and recipients were female. Islets were either control (isolated immediately) or exposed to 60-minutes of warm ischemia (27). The kidney was accessed by left flank incision and brought into the wound by gentle blunt dissection. A small nick was made in the kidney capsule at the inferior renal pole and islets were deposited toward the superior pole. Blood glucose levels were monitored daily for seven days, then every second day up to thirty days. Two replicate experiments were performed.

Hyperspectral fluorescence microscopy

Hyperspectral (wide-field) fluorescence microscopy used an Olympus IX83 microscope with a NuVu electron multiplying charge coupling device camera (EMCCD, hnu1024). LED illumination produced excitation at 358, 371, 377, 381, 385, 391, 397, 400, 403, 406, 412, 418, 430, 437, 451, 457, 469, and 476 ± 5 nm while emission was detected using filters at 414, 451, 575, 594, and 675 ± 20 nm. The objective was 40× oil objective lens (UAPON340, Olympus). Imaging was carried out using a warm-stage at 37^oC with cells in Hank's Balanced salt solution (HBSS; Gibco), which is non-fluorescent.

Image preparation and analysis

Image preparation was carried out (28, 29) to remove image artefacts (i.e. Poisson’s noise, dead and saturated pixels, illumination curvature, background fluorescence). Regions of interest were segmented from the channel images to produce single-region images, and a variety of colour intensity features were extracted. These features included mean channel intensity and their associated statistical measures such as channel intensity ratio (30). Further, features related to the histogram of the cell images such as pixels’ standard deviation and skewness were also considered, which characterized the colour distribution of an image (31).

Modelling and unmixing

To assess optimal separation in each case under consideration, data points representing multidimensional feature vectors for all islets or cells were projected onto an optimal (for this case) two-dimensional (2-D) space created by discriminative analysis (32). This space maximizes between-group distance while minimizing within-group variance, and it is spanned by two canonical variables equal to a selected linear combinations of cellular features (33). Finally, a classifier was employed to predict the pre-defined region labels (34). Receiver operating characteristic (ROC) area under the curve (AUC) analysis was used to quantify the accuracy of models (values of 1.0 indicate perfect accuracy, values of 0.5 indicate no better accuracy than chance). In Fig. 8a conventional, common approach of principal component analysis (PCA) was applied to the same multidimensional feature vectors for unsupervised assessment.

For unmixing spectral signals to identify component fluorophores, linear mixed modelling was used where the signal of each pixel was held to be a linear combination of a small number of endmember component spectra with weights attributable to the concentration of the fluorophores responsible. Relative concentrations were expressed as abundance fractions (35–37), and the unsupervised algorithm Robust Dependent Component Analysis (RoDECA) was used to identify dominant endogenous fluorophores and their abundance. We compared extracted spectral characteristics to known characteristics of fluorophores for validation. RoDECA has been shown to be able to discriminate individual fluorophores in the presence of overlapping spectra and image noise (20, 28, 38).

Statistical analysis

Statistical analysis was carried out in Matlab (R2017b). As the data did not meet the assumptions of parametric testing (normal distribution, equal variance), group comparisons were made using the Mann-Whitney U test which ranks data to compare between group differences. Data are presented as median values and 95% confidence intervals. Groups were accepted as being significantly different at an alpha value of 0.05.

Data and Resource Availability

Data will be made available on reasonable request.

Islets were exposed to ROS damage, hypoxia, pro-inflammatory cytokines and warm ischemia. In all cases the spectral signals for the fluorophores NAD(P)H, flavins, collagen-I and cytochrome-C were successfully unmixed using RoDECA. These four fluorophores have had their spectral characteristics identified and assigned in a fluorophore reference bank of purified fluorophores maintained at physiological concentrations and pH (20, 28, 38). Discriminant analysis was used to investigate whether a hyperspectral signal sensitive to the presence of compromised viability could be constructed. This was repeated for single cell suspensions from disaggregated islets as this strategy would have greater clinical utility. Unmixing was also undertaken for single cells where findings paralleled the results for whole islets (Supplementary Material 1).

ROS damage

ROS damage by menadione created a visible change in the spectral morphology of many islets (Fig. 1). This effect was not apparent in all islets, however the model developed from the hyperspectral data achieved perfect separation (Fig. 2A and Fig. 2B; AUC = 1.00). Similar accuracy was achieved for single cells (AUC = 0.99, Supp Fig. 1A) despite the loss of the whole islets' distinctive morphology. Unmixing in whole islets was able to identify the autofluorescent coenzymes NAD(P)H and flavins, the structural protein collagen-I, and the mitochondrial protein cytochrome C. Additionally, redox ratio was calculated as the ratio of unmixed abundances of NAD(P)H and Flavins. The only statistically significant difference was for redox ratio, which was elevated in islets exposed to ROS damage (Supp Fig. 2).

Hypoxia

The induction of hypoxia in islets through DMOG exposure represented a more subtle form of damage, with no apparent impact on islet morphology and weaker discrimination (Fig. 2C and D; AUC = 0.69). A minor improvement was seen for single cells, which could be separated with AUC = 0.72 (Supp Fig. 1B). Unmixing showed a significant elevation in redox ratio (Fig. 3D), but here the increase for NAD(P)H was also significant (Fig. 3A).

Inflammatory cytokines

The inflammation model was more complex with three comparisons. Good accuracy was achieved with islets exposed to moderate pro-inflammatory signalling (4 hours) being discriminable from pristine islets with AUC = 0.79 (Fig. 4A, B), islets exposed to major pro-inflammatory signalling (24 hours) being discriminable from pristine islets with AUC = 0.94 (Fig. 4C and Fig. 4D), and the two groups of exposed islets being discriminable from one another with AUC = 0.95 (Fig. 4E and Fig. 4F). Single cells from islets exposed to major and moderate pro-inflammatory signalling could be discriminated from cells from pristine islets with AUC 0.94 and 0.98 (Supplementary Material 1C), respectively. Unmixing analyses show significant increases in NAD(P)H for both 4 hours and 24 hours of exposure (Fig. 5A). Flavins were not significantly affected (Fig. 5B), however RR was increased for 24 hours cytokine exposure compared to 24 hours control (Fig. 5D). For both exposure times, less collagen-I was detected in cytokine treated islets (Fig. 5C), however cytochrome C was not affected. Comparisons were also made between 4-hour control and 24 hours control as well as 4 hours cytokine and 24 hours cytokine exposure. The only effect was for NAD(P)H where the difference between the two periods was p = 0.05.

Warm ischemia

Islets were exposed to warm ischemia to directly model the damage which human islets for transplantation may experience prior to pancreatic retrieval (Fig. 6). Good accuracy was achieved for the discrimination of pristine islets from those with moderate exposure (Fig. 6A and Fig. 6B, AUC = 0.82), but better accuracy was achieved for the discrimination of islets with major exposure (Fig. 6C and D, AUC = 0.92). Discrimination of islet with moderate compared to major exposure was 0.80 (Fig. 6E and Fig. 6F). Three-way discrimination which sorted islets into each of the three groups (control, moderate, major) achieved AUC = 0.76. This suggests hyperspectral imaging of autofluorescence can give a scaled indication of islet preparation viability. Results for single cells were also good, although, surprisingly, accuracy was similar for moderate and extreme exposure, with moderate exposure being very slightly ahead (AUC = 0.86 and 0.84, Supp Fig. 1D).

Unmixing showed a significant increase in NAD(P)H for 60 minutes of warm ischemia compared to both control and 30 minutes (Fig. 7A), Flavins were significantly reduced by both 30 and 60 minutes warm ischemia compared to the controls, but not relative to each other (Fig. 7B). This was reflected by warm ischemia significantly increasing RR after both 30 and 60 minutes (Fig. 7D). Effects for collagen-I were highly significant (Fig. 7C), but without a clear pattern, with 30 minutes increasing collagen I relative to control and 60 minutes, and 60 minutes being significantly lower than control. In the assessment of single cells, wherein the extracellular matrix had been further disrupted through exposure to EDTA for disaggregation, no significant differences were found between groups (supplementary material 1). Cytochrome-C was significantly elevated by 60 minutes of warm ischemia compared to both controls and 30 minutes (Fig. 7E).

Encapsulation

The encapsulation of islets in a multilayer conformal film was investigated to determine whether the hyperspectral imaging of islet autofluorescence could still detect viability compromising insults in this context. First, we investigated whether the presence of the conformal coating had an impact on the hyperspectral properties of the islets by comparing the profiles of encapsulated islets to those which had been taken through the encapsulation process without application of the conformal coating (Unencapsulated islets). Perfect accuracy was achieved for their discrimination (AUC = 1.0), which was not surprising given the visible impact of the conformal coating on the spectral properties of the islets (Supp. Figure 3). The comparison of Unencapsulated islets to control pristine islets also achieved perfect discrimination (AUC = 1.0), suggesting that the encapsulation process itself had a major effect on islet characteristics. This raised the question as to whether other sources of potentially compromising insult could still be detected in encapsulated islets. To assess this, we reperformed the previously described cytokine experiment. If encapsulation material had too large of an impact on the hyperspectral characteristics of islets it would overshadow our ability to use hyperspectral microscopy to detect damage. This was not the case, however, as 4 hours and 24 hours of exposure to pro-inflammatory cytokines could be discriminated with accuracy comparable to previous values (AUC 0.73 and 0.89). Other implications of these findings are that the impact of the encapsulation process is not so major that the comparatively minor impact of 4 hours exposure to cytokines cannot still be detected, and that, not surprisingly, the encapsulation envelop does not exclude our proinflammatory cytokines.

Prediction of transplant success

Glucose control after transplantation was assessed over 30 days (Fig. 8E). All transplanted control islets maintained excellent glucose control from day 1 to the end of follow-up. Three of the five islet preparations subjected to warm ischemia (Fig. 8E) did not restore glucose control, while two did. In both cases the ischemic islets which restored glucose control had partial function in the initial period (< 5 days) but maintained normo-glycemia by the completion of monitoring (Fig. 8E). The inclusion of islets exposed to warm ischemia in both groups (functional/non-functional) is important as it means that this analysis is not synonymous with the discrimination of islets exposed to warm ischemia and those not.

Unsupervised PCA was applied to multidimensional feature vectors for all islets to investigate their emergent sorting (Fig. 8). Groups could be seen with a bias for clustering islets from functional and non-functional preparations (Fig. 8A) more so than Control and Ischemic preparations (Fig. 8B). Principal component 2 emerged from this assessment (Fig. 8C) as a potential informative "viability score". Functional and non-functional preparations could be separated in all cases by their being above or below a threshold value for this "viability score" (Fig. 8C). An ROC curve of individual islets had lower accuracy (AUC = 0.81); however, these preparations should be expected to be heterogenous, with overall functional preparations containing a minority of compromised islets and non-functional preparations containing some of viable islets. These findings are particularly meaningful as they are the result of unsupervised analysis – that is, the model was not structured to separate preparations that restored glucose control from those which did not, instead it was "blind" to group membership and the distinction emerged when it sorted like with like. Data for single cells which are more vulnerable to motion artifacts, had too high variance for unsupervised assessment to be undertaken.

There are several assays for post-transplant islet function, however they have all been found to be insufficiently reliable predictors for robustly informing clinical decision making (4–9). Optimally, technologies that interrogate an islet preparation's functional capacity will be non-invasive (4) in order to maximally preserve islet numbers in clinical contexts and enable direct follow-up experiments in research contexts. Hyperspectral microscopy of autofluorescence enabled the differentiation of individual islets that were damaged as compared to those in a pristine resting state. Additionally, it was able to discern the individual components of damaging stimuli, and so could discern islets exposed to ROS damage, pro-inflammatory cytokine signalling, or warm ischemia. Strong accuracy (AUC > 0.9) was achieved for detecting islets exposed to ROS, pro-inflammatory cytokine signalling or warm ischemia, while further refinements may improve the detection of hypoxia damage. Further, an unsupervised algorithm prospectively identified which islet preparations restored glucose control in diabetic mice. This nominally outperforms any prior technology (4–9), and was achieved under challenging (discriminating islets capable of restoring glucose control from those which were not, even when both preparations were exposed to ischemia) and blind conditions, however further replication is needed to demonstrate reliability. These data show hyperspectral microscopy has strong potential to be translated to assess islet viability and inform clinical decision making with novel information on likelihood of transplant success. As well as reducing patient burden, the withdrawal of immunosuppression following failed islet transplants results in heightened sensitisation to human leukocyte antigens (39, 40), increasing the importance of avoiding islet preparations with a low likelihood of successful transplantation.

The 3D nature of islets could interfere with their emission spectrum via absorption. Single cell imaging avoids this possibility, but at the expense of structural information. However, when islets were disaggregated the accuracy of the hyperspectral models was not compromised. Due to their size only one to three islets can be imaged at a time, which would reduce the efficiency of assessment. To investigate a strategy to overcome this limitation we homogenised suspensions of single cells to enable collection of 50–80 datapoints per field-of-view, presumptively representative of all islets disaggregated. This allowed the rapid collection of a characteristic dataset with a lower image preparation burden, decreasing time taken for assessment. These findings also demonstrate the potential of this technology to be applied for emerging beta-cell replacement strategies, such as those based on stem cell culture and differentiation, beyond deceased donor islet transplantation.

Redox ratio was significantly increased in islets with compromising exposures across all conditions, with the exception of moderate exposure to pro-inflammatory cytokines where statistical significance was not reached (p = 0.11). An associated significant increase in NAD(P)H was also frequently observed, sometimes accompanied by a reduction in flavins. These effects are well supported by the literature. ROS generation from menadione exposure in pancreatic β-cells is driven by elevated NADH (41), and NADPH maintains systems which defend against cellular ROS damage (41). Hypoxia necessitates greater reliance on anerobic glycolysis, which is marked by a shift from flavins towards NAD(P)H (42). Activation of immune cells by inflammatory signalling has been linked to glycolysis and increased redox ratio (43), however our observation of the redox ratio in pancreatic islets being increased by pro-inflammatory signalling appears to be novel. Increasing levels of NADH in blood have also been observed with post-mortem interval (44). Furthermore, both regional and global myocardial ischemia induced in isolated rat hearts resulted in a rapid, substantial increases in the intensity of NADH autofluorescence (45).

The potential of elevated redox ratio to act as a consistent biomarker of islet viability is further supported by the observation of increased NAD(P)H relative to flavins in cells undergoing apoptosis (46). Furthermore, cell metabolism shifting away from oxidative phosphorylation and towards anerobic glycolysis, indicated by the elevated redox ratio (42), represents a critical junction in islet physiology, as the decreased production of ATP from glycolysis relative to oxidative phosphorylation cripples glucose-triggered insulin secretion, rendering islets non-responsive to fluctuating glucose in the external milieu (11, 47).

We also isolated the spectral signal of cytochrome-C, although the only significant findings were made for warm ischemia where it was elevated in islets with major exposure relative to pristine islets and moderate exposure. Cytochrome-C is generally located between the inner and outer mitochondrial layers, where it is an essential component of the electron transport chain. The link between islet viability and Cytochrome-C may be via its connection to apoptosis. Cytochrome-C being released from the mitochondria into the cytosol is a primary drivers of apoptosis as it activates a caspase cascade that commits the cell to the death process. This change in localisation is generally not expected to alter the relative strength of its total spectral signal, however the upregulation of cytochrome-C production has also been observed to accompany apoptosis (48).

Islets collected for transplantation are exposed to a number of viability compromising insults. We have shown that hyperspectral microscopy has the sensitivity to discern these insults by the spectral signature they impart on an islet. Furthermore, we demonstrated that hyperspectral imaging of islets is able to produce a “viability score” which reflects the ability of otherwise compromised islets to yield good glucose control after transplantation into diabetic mice. Such a score is critically required for clinical islet transplantation where islets come from deceased donors and are known to have been subjected to multiple insults, including warm ischemia, hypoxia, oxidative stress and cytokine injury. Potential limitations of this technology primarily relate to translation from mouse to humans, including both biological variation between species and the heterogeneity introduced by the clinical environment (e.g. variable insulin secretion function, patient and donor specific factors). The increased presence of acinar tissue seen in human preparations compared to mouse should not present a challenge as this tissue is morphologically distinct from islets under normal brightfield microscopy (49). Hyperspectral images took approximately six minutes to collect, however this would be significantly reduced in translation through the use of a task specific instrument and protocol which could have uninformative channels removed and a wider field of view. The absence of phototoxicity from the excitation illumination wavelengths affecting islet function and viability should also be validated in future research. However, as we have shown that this technology does not compromise in vitro cultured embryos – a notoriously fragile system – such an effect is unlikely (50).

This study offers a clear pathway for grading islet preparations and predicting their future performance. The strength of our findings are reinforced by the use of fully unsupervised assessment. An additional strength is that this approach could be applied to alternate sources of beta-cells, including quality assessment of xenogenic islets (51), and potentially stem cell derived islets, where distinguishing differentiated functional beta-cells from undifferentiated cells would be advantageous (52). For application to adult cadaveric islets, as currently used in clinical practice, this technology has the potential to give a non-invasive indication of islet viability, prior to transplantation, which would inform clinical decision making and enable patients to be spared transplantation attempts with no potential to reduce their dependence on exogenous insulin.

Ethics approval and consent to participate

Ethics approval was given by the Garvan Institute of Medical Research Animal Research Authority (20_18)

Consent for publication

Not applicable

Availability of data and materials

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests

The authors declare that they have no competing interests.

Funding

This research was supported by the Juvenile Diabetes Research Foundation through the JDRF award 1-INO-2019-788-S-B to S.T.G and E.G. and partially supported by a Discovery Project grant from the Australian Research Council (DP170101863) and the Centre of Excellence scheme (CE140100003) as well as the UNSW SHARP scheme to E.G. The research was also supported by National Health Medical Research Council (NHMRC) grants (GNT1130222; GNT1146493; GNT1189235) and the Twin Towns Services Community Foundation to S.T.G. S.T.G. is a NHMRC Senior Research Fellow (GNT1140691).

Author’s contributions

JMC SNW, STG and EMG conceived of and contributed to the design of the project, JMC, AGA and SNW participated in data collection, AH and SBM undertook the analysis, and all authors contributed to interpretation. The manuscript was drafted by JMC with all authors participating in its critical revision. All authors give approval for publication and accept accountability for the work. JMC is Guarantor.

Results from this manuscript have been presented at 'Label-free Biomedical Imaging and Sensing (LBIS) 2021' as "Pancreatic islet characterisation by the hyperspectral Assessment of autofluorescence: a non-invasive, label-free measure of viability".

AUC; Area under curve

DMOG; Dimethyloxalylglycine, N-(Methoxyoxoacetyl)-glycine methyl ester

ECM; Extra cellular matrix

FAD; Flavin adenine dinucleotide

FLIM; Fluorescence lifetime imaging microscopy

HBSS; Hank's balanced salt solution

LMM; Linear mixed model

NADH; Nicotinamide adenine dinucleotide hydride

NAD(P)H; Nicotinamide adenine dinucleotide (with or without) phosphate hydrogen

PVPON/TA; poly(N-vinylpyrrolidone)/ tannic acid

RoDECA; Robust Dependent Component Analysis

ROS; Reactive oxygen species

ROC; Receiver operating characteristic

T1D; Type 1 diabetes

O'Connell PJ, Holmes-Walker DJ, Goodman D, Hawthorne WJ, Loudovaris T, Gunton JE, et al. Multicenter Australian trial of islet transplantation: improving accessibility and outcomes. American journal of transplantation: official journal of the American Society of Transplantation and the American Society of Transplant Surgeons. 2013;13(7):1850–8.
Anazawa T, Okajima H, Masui T, Uemoto S. Current state and future evolution of pancreatic islet transplantation. Ann Gastroenterol Surg. 2019;3(1):34–42.
Barton FB, Rickels MR, Alejandro R, Hering BJ, Wease S, Naziruddin B, et al. Improvement in outcomes of clinical islet transplantation: 1999–2010. Diabetes Care. 2012;35(7):1436–45.
Papas KK, Suszynski TM, Colton CK. Islet assessment for transplantation. Curr Opin Organ Transplant. 2009;14(6):674–82.
Ricordi C, Lakey JR, Hering BJ. Challenges toward standardization of islet isolation technology. Transplant Proc. 2001;33(1–2):1709.
Bertuzzi F, Ricordi C. Prediction of clinical outcome in islet allotransplantation. Diabetes Care. 2007;30(2):410–7.
Street CN, Lakey JR, Shapiro AM, Imes S, Rajotte RV, Ryan EA, et al. Islet graft assessment in the Edmonton Protocol: implications for predicting long-term clinical outcome. Diabetes. 2004;53(12):3107–14.
Salgado M, Gonzalez N, Medrano L, Rawson J, Omori K, Qi M, et al. Semi-Automated Assessment of Human Islet Viability Predicts Transplantation Outcomes in a Diabetic Mouse Model. Cell Transplant. 2020;29:963689720919444.
Goto M, Holgersson J, Kumagai-Braesch M, Korsgren O. The ADP/ATP ratio: A novel predictive assay for quality assessment of isolated pancreatic islets. American journal of transplantation: official journal of the American Society of Transplantation and the American Society of Transplant Surgeons. 2006;6(10):2483–7.
Kitzmann JP, O'Gorman D, Kin T, Gruessner AC, Senior P, Imes S, et al. Islet oxygen consumption rate dose predicts insulin independence for first clinical islet allotransplants. Transplant Proc. 2014;46(6):1985-8.
Cantley J, Walters SN, Jung MH, Weinberg A, Cowley MJ, Whitworth TP, et al. A preexistent hypoxic gene signature predicts impaired islet graft function and glucose homeostasis. Cell Transplant. 2013;22(11):2147–59.
Cowley MJ, Weinberg A, Zammit NW, Walters SN, Hawthorne WJ, Loudovaris T, et al. Human islets express a marked proinflammatory molecular signature prior to transplantation. Cell Transplant. 2012;21(9):2063–78.
Abdelli S, Ansite J, Roduit R, Borsello T, Matsumoto I, Sawada T, et al. Intracellular stress signaling pathways activated during human islet preparation and following acute cytokine exposure. Diabetes. 2004;53(11):2815–23.
Bartolome F, Abramov AY. Measurement of mitochondrial NADH and FAD autofluorescence in live cells. Methods Mol Biol. 2015;1264:263–70.
Heaster TM, Walsh AJ, Zhao Y, Hiebert SW, Skala MC. Autofluorescence imaging identifies tumor cell-cycle status on a single-cell level. J Biophotonics. 2018;11(1).
Ostrander JH, McMahon CM, Lem S, Millon SR, Brown JQ, Seewaldt VL, et al. Optical redox ratio differentiates breast cancer cell lines based on estrogen receptor status. Cancer Res. 2010;70(11):4759–66.
Meleshina AV, Dudenkova VV, Bystrova AS, Kuznetsova DS, Shirmanova MV, Zagaynova EV. Two-photon FLIM of NAD(P)H and FAD in mesenchymal stem cells undergoing either osteogenic or chondrogenic differentiation. Stem Cell Res Ther. 2017;8(1):15.
Ramanujam N. Fluorescence spectroscopy of neoplastic and non-neoplastic tissues. Neoplasia. 2000;2(1–2):89–117.
Campbell JM, Habibalahi A, Mahbub S, Gosnell M, Anwer AG, Paton S, et al. Non-destructive, label free identification of cell cycle phase in cancer cells by multispectral microscopy of autofluorescence. BMC Cancer. 2019;19(1):1242.
Mahbub SB, Guller A, Campbell JM, Anwer AG, Gosnell ME, Vesey G, et al. Non-Invasive Monitoring of Functional State of Articular Cartilage Tissue with Label-Free Unsupervised Hyperspectral Imaging. Sci Rep. 2019;9(1):4398.
Mahbub SB, Nguyen LT, Habibalahi A, Campbell JM, Anwer AG, Qadri UM, et al. Non-invasive assessment of exfoliated kidney cells extracted from urine using multispectral autofluorescence features. Sci Rep. 2021;11(1):10655.
Habibalahi A, Moghari MD, Campbell JM, Anwer AG, Mahbub SB, Gosnell M, et al. Non-invasive real-time imaging of reactive oxygen species (ROS) using multispectral auto-fluorescence imaging technique: a novel tool for redox biology. 2020:101561.
Campbell JM, Mahbub S, Habibalahi A, Paton S, Gronthos S, Goldys E. Ageing human bone marrow mesenchymal stem cells have depleted NAD(P)H and distinct multispectral autofluorescence. Geroscience. 2020.
Grey ST, Longo C, Shukri T, Patel VI, Csizmadia E, Daniel S, et al. Genetic engineering of a suboptimal islet graft with A20 preserves beta cell mass and function. J Immunol. 2003;170(12):6250–6.
Pham-Hua D, Padgett LE, Xue B, Anderson B, Zeiger M, Barra JM, et al. Islet encapsulation with polyphenol coatings decreases pro-inflammatory chemokine synthesis and T cell trafficking. Biomaterials. 2017;128:19–32.
Walters S, Webster KE, Sutherland A, Gardam S, Groom J, Liuwantara D, et al. Increased CD4 + Foxp3 + T cells in BAFF-transgenic mice suppress T cell effector responses. J Immunol. 2009;182(2):793–801.
Hesse UJ, Schmitz-Rode M, Voss S. Islet counting by staining with dithizone in various concentrations. Transplant Proc. 1992;24(6):2989.
Mahbub SB, Plöschner M, Gosnell ME, Anwer AG, Goldys EM. Statistically strong label-free quantitative identification of native fluorophores in a biological sample. Scientific reports. 2017;7(1):15792%@ 2045–2322.
Habibalahi A, Bala C, Allende A, Anwer AG, Goldys EMJTOS. Novel automated non invasive detection of ocular surface squamous neoplasia using multispectral autofluorescence imaging. 2019.
Gosnell ME, Anwer AG, Mahbub SB, Perinchery SM, Inglis DW, Adhikary PP, et al. Quantitative non-invasive cell characterisation and discrimination based on multispectral autofluorescence features. Scientific reports. 2016;6:23453%@ 2045–2322.
El Aziz MA, Selim IM, Xiong S. Automatic detection of galaxy type from datasets of galaxies image based on image retrieval approach. Scientific Reports. 2017;7(1):4463%@ 2045–322.
Jombart T, Devillard S, Balloux F. Discriminant analysis of principal components: a new method for the analysis of genetically structured populations. BMC genetics. 2010;11(1):94%@ 1471–2156.
Naganathan GK, Grimes LM, Subbiah J, Calkins CR, Samal A, Meyer GE. Visible/near-infrared hyperspectral imaging for beef tenderness prediction. Computers and electronics in agriculture. 2008;64(2):225 – 33%@ 0168–1699.
Vapnik V. The nature of statistical learning theory: Springer science & business media; 2013.
Keshava N, Mustard JF. Spectral unmixing. Signal Processing Magazine, IEEE. 2002;19(1):44–57.
Keshava N, Kerekes JP, Manolakis DG, Shaw GA, editors. Algorithm taxonomy for hyperspectral unmixing2000.
Keshava N. A survey of spectral unmixing algorithms. Lincoln Laboratory Journal. 2003;14(1):55–78.
Mahbub SB. Unsupervised Hyperspectral Unmixing Analysis for Label-free Quantitative Identification of Native Fluorophores in a Biological Sample by a Robust Dependent Component Analysis (RoDECA). Macquarie University; 2017.
Campbell PM, Senior PA, Salam A, Labranche K, Bigam DL, Kneteman NM, et al. High risk of sensitization after failed islet transplantation. American journal of transplantation: official journal of the American Society of Transplantation and the American Society of Transplant Surgeons. 2007;7(10):2311–7.
Cardani R, Pileggi A, Ricordi C, Gomez C, Baidal DA, Ponte GG, et al. Allosensitization of islet allograft recipients. Transplantation. 2007;84(11):1413–27.
Blacker TS, Duchen MR. Investigating mitochondrial redox state using NADH and NADPH autofluorescence. Free Radic Biol Med. 2016;100:53–65.
Kolenc OI, Quinn KP. Evaluating Cell Metabolism Through Autofluorescence Imaging of NAD(P)H and FAD. Antioxid Redox Signal. 2019;30(6):875–89.
Griffiths HR, Gao D, Pararasa C. Redox regulation in metabolic programming and inflammation. Redox Biol. 2017;12:50–7.
Donaldson AE, Lamont IL. Biochemistry changes that occur after death: potential markers for determining post-mortem interval. PLoS One. 2013;8(11):e82011.
Papayan G, Petrishchev N, Galagudza M. Autofluorescence spectroscopy for NADH and flavoproteins redox state monitoring in the isolated rat heart subjected to ischemia-reperfusion. Photodiagnosis and photodynamic therapy. 2014;11(3):400–8.
Levitt JM, Baldwin A, Papadakis A, Puri S, Xylas J, Munger K, et al. Intrinsic fluorescence and redox changes associated with apoptosis of primary human epithelial cells. J Biomed Opt. 2006;11(6):064012.
Cantley J, Grey ST, Maxwell PH, Withers DJ. The hypoxia response pathway and β-cell function. Diabetes Obes Metab. 2010;12 Suppl 2:159–67.
Chandra D, Liu JW, Tang DG. Early mitochondrial activation and cytochrome c up-regulation during apoptosis. J Biol Chem. 2002;277(52):50842–54.
Ricordi C, Lacy PE, Finke EH, Olack BJ, Scharp DW. Automated method for isolation of human pancreatic islets. Diabetes. 1988;37(4):413–20.
Tan TCY, Mahbub SB, Campugan CA, Campbell JM, Habibalahi A, Chow DJX, et al. Non-invasive, label-free optical analysis to detect aneuploidy within the inner cell mass of the preimplantation embryo. bioRxiv. 2020:2020.11.01.364133.
Cardona K, Korbutt GS, Milas Z, Lyon J, Cano J, Jiang W, et al. Long-term survival of neonatal porcine islets in nonhuman primates by targeting costimulation pathways. Nat Med. 2006;12(3):304–6.
Odorico J, Markmann J, Melton D, Greenstein J, Hwa A, Nostro C, et al. Report of the Key Opinion Leaders Meeting on Stem Cell-derived Beta Cells. Transplantation. 2018;102(8):1223–9.

Download PDF

Version 1

posted

You are reading this latest preprint version

Autofluorophores Assessed by Hyperspectral Microscopy Indicate Perturbation and Transplant Viability in Pancreatic Islets

Status:

Version 1

Abstract

Background

Methods

Findings:

Conclusions

Figures

Introduction

Methods

Islet collection and culture

Islet transplantation

Hyperspectral fluorescence microscopy

Image preparation and analysis

Modelling and unmixing

Statistical analysis

Data and Resource Availability

Results

ROS damage

Hypoxia

Inflammatory cytokines

Warm ischemia

Encapsulation

Prediction of transplant success

Discussion

Conclusions

Declarations

Abbreviations

References

Status:

Version 1