Semiautomatic assessment of endothelial density and morphology in organ-cultured corneas — potential predictors for transplantation suitability and clinical outcome?

Corneal diseases are a major cause of visual impairment and blindness worldwide. According to the estimations of the World Health Organization (WHO), corneal opacities are the 3^rd most common cause of blindness worldwide, accounting for 7% of the world blind population [1-3].

Over a century after being first established as a surgical procedure, allogeneic cornea transplantation still plays an important role as a remedy of last resort for a great variety of corneal conditions. It is the most common type of solid tissue transplantation worldwide: about 150,000 keratoplasties are performed worldwide every year, predominantly in North America and Europe .

A high transplant-quality — mainly defined by biomechanical integrity and transparency — is essential for the long-term success of the treatment of corneal diseases by keratoplasty. One of the key factors is the functionality of the corneal endothelium, as it regulates nutrition and osmotic equilibrium of the cornea, thus warranting corneal transparency. The correlation between endothelium cell density (ECD) and functionality is well known and has been studied in detail [4-6].

Several factors have been studied for a possible influence on the ECD in donor corneas such as donor diseases, cause of death, corneal retrieval techniques and proficiency of the staff performing the microsurgery procedures [7, 8]. However, only donor age turned out to have a significant effect on the ECD of corneal transplants: statistically, ECD decreases throughout life by 0.5 to 0.6% per year [9, 10].

In functional corneal endothelium, mainly hexagonal, but also pentagonal cells are prevalent, as their geometry allows a mathematically economic, homogenous cover of the Descemet membrane following the corneal curvature [11]. In case of apoptotic or necrotic degradation of endothelial cells (due to increasing age or harmful environmental influences), neighboring endothelial cells increase their size to fill in any gaps in the endothelium layer, as they are unable to undergo mitosis in adult individuals [12, 13]. In consequence, these neighboring cells lose their hexagonal or pentagonal shape and become polygonal. Thus, the proportion of hexagonal and pentagonal cells in relation to polygonal cells decreases with the declining endothelium cell density resulting from cell degradation. While this correlation is well understood, it is obvious that morphology is not exclusively dependent on cell density: independent cellular- and molecular-biological effects emerging during normal (cell-) aging reducing the endothelial functionality also may affect endothelial morphology — and thus on the suitability of the corresponding corneas intended for transplantation.

New semi-automated methods of microscopic assessment of donor cornea endothelium not only allow the standardized quantification of mean cell density, but also facilitate an easier assessment of the morphology data of endothelial cells. Nonetheless, to our knowledge, no surveys have studied whether the endothelium cell morphology itself may have a cell-density-independent effect on the stability of human donor corneas in culture or on the clinical outcomes after transplantation.

The objective of this study is to explore whether endothelial morphology data of donor corneas (obtained after donation and during organ culture) could serve as an independent predictive parameter for the functional stability during in vitro cultivation and the clinical outcome after transplantation; this is done in comparison with data on the endothelial density in donor corneas, the actual main discriminative parameter for transplantation approval of donor corneas.

This retrospective case study was conducted at the Hamburg Eye Bank (HEB).

For the purpose of this study, we analyzed 1031 donor corneas that were processed in the Hamburg Eye Bank between 2012 and 2016. For this study, endothelial cell counts and endothelial cell morphology of the initial and the final examinations of donor corneas (see below), cultivation data (as the retrieval method, the time between donor death and cultivation of the cornea, the total duration of the cultivation, days in dextran-free and dextran-containing media) and donor data (donor age, donor death cause, basal diseases) were assessed for all corneas.

All corneal tissue was retrieved from deceased donors by the department of forensic medicine of the University Hospital Hamburg. Consent calls with relatives of the donor were performed in accord with the current law. Cornea recovery was performed by the medical forensic staff or by the trained technicians of the HEB. Tissue processing, preparation, and cultivation were performed by trained technicians of the HEB.

One of the main problems in corneal transplant-surgery is the scarcity of suitable donor corneas [15]. Selecting donor corneas for transplantation is still mainly based on the absence of optical opacities and especially on sufficiently high endothelial cell densities (ECD), as a proper function of the endothelial cell layer is crucial for nutrition, osmoregulation, transparency and structural integrity of the cornea.

A regular hexagonal pattern provides the most stable covering of a plane and a deviation from this pattern may reflect a less stable monolayer: irregular shapes cause greater surface tension on the monolayer and cause a suboptimal geometric and thermodynamic configuration thus reducing the ability of corneal endothelium to maintain corneal homeostasis [16, 17].

The principal aim of this study was to examine whether endothelial cell morphology (ECM) could be used as a supplemental parameter to estimate the prospective quality of donor corneas after organ culture — and thus their suitability for transplantation.

A unique feature of our study is the additional analysis of clinical data obtained after the transplantation of the donor corneas. To our knowledge, this has not been done so far.

It is obvious that such analyses face several challenges:

First, the base collectives (donor characteristics, donated transplant-corneal ECD/ECM attributes and clinical data of transplanted patients) used for the comparisons should be widely homogeneous. Although there was a prevalence of male donors in our study group (1.68 to 1; Table 1), we found no difference either in ECD or in ECM between both sexes; also, death causes did not display any independent correlation with these variables (data not shown, Supplementary Figs. 1–2). These findings correspond well to previous studies [8; 19]. Donated transplant-corneal ECD/ECM attributes in organ culture (Table 3) also display homogeneous distribution patterns. However, the collective of cornea-transplanted patients is (inevitably) more heterogeneous (Table 2); this point may complicate the interpretation of our results (see below).

Second, we could demonstrate that a parameter of the analyzed collective — donor age — is associated with the compared variables ECD and ECM: The higher the donor age, the lower the corresponding ECD- and ECM values (see the respective correlation in Table 4). This is a well-known fact already demonstrated in previous studies [8; 20–23]. The average age of donors in this study was 66.1 years old, and only 5% of all corneas were retrieved from donors younger than 40 years of age. Thus, our conclusions on the results of our analysis might only apply to corneas from donors of elevated ages.

Third, we compared two parameters — ECD and ECM (and their stabilities in culture) — which can be assumed to influence each other; we indeed demonstrated that these parameters are correlated positively: the higher the corneal ECDs, the better the corresponding corneal ECMs (the same applies for their stabilities in culture; Table 4). Comparative analyses on the predictive capacities of these parameters have to consider their correlation.

Now, comparative ROC analyses represent a mathematically elegant method to compare the predictive powers of ECD and ECM (or their stabilities in culture) despite their evident correlation: as their individual areas under the curve (AUCs) always include the mutual correlation, differences of the AUCs necessarily are independent from it. A higher AUC of one of these variables (f.e. ECD or ECM) thus could indicate a higher predictive power which is independent from the other variable.

The main findings of our study are as follows:

Firstly, initial ECD or ECD development in culture displays only a limited predictive power (AUC of 0.655 or less; Fig. 2) for the approval (or rejection) of a donor cornea for transplantation.

Secondly, initial ECD or ECD development in culture does not have any predictive power (AUC values nearby 0.5; Fig. 3) for the clinical development after transplantation.

Thirdly, in no case — neither in organ culture before approval for transplantation nor in clinical development after transplantation — did initial ECM or ECM development in culture displays a predictive power superior to that of ECD or ECD development in culture (Figs. 2 and 3).

In summary, we could demonstrate that the initial ECM (measured after retrieval from the donor) as well as the ECM development during organ-culture do not provide any additional information to estimate the suitability of donor corneas for transplantation or their visual performances after transplantation — at least when semi-automated measurements were employed.

At the same time, the relatively low predictive power of initial ECD of donor corneas for transplantation-suitability is somewhat surprising. Even more, initial ECD lacks any prediction ability for the visual performance after transplantation. A possible explanation (at least for the latter observation) is that the usual cut-off level of corneal ECDs at their final assessment (at the end of organ culture) for transplantation approval (or against) in our institution lies at minimum 2000 cells/mm² (for deep lamellar transplantation even at 2500 cells/mm²). Now, corneal endothelium usually still is functional at ECDs of 600 to 800 cells/mm² [6]; thus, the elevated final ECD cut-off levels of 2000-/2500 cells/mm² for transplant-cornea approvals could be too high to enable a detection of a putative predictive power of initial ECD for visual performance after transplantation in our study setting (i.e. within 2 years of observation). Noteworthy, in side analyses on sub-collectives from donors older than 80 years (with generally lower ECD distributions), the predictive capacity of initial ECD for transplantation approval proved to be somewhat higher (AUC = 0.796, not shown, Supplementary Fig. 3). In consequence, our observations of course do not categorically exclude an influence of initial endothelial density on the final approval for transplantation or the clinical outcomes of transplanted corneas. The poor predictive capacity of this variable in our analyses however highlights the need for further studies on the issue, whether current cut-off levels for transplantation approval could be downscaled; especially long-term observational studies on cornea-transplanted patients (longer than 2 years post transplantation) would be interesting.

In this study, high numbers of donors, transplant-corneas and (to a lesser extent) transplanted patients were analyzed thoroughly. However, its results and conclusions still need to be tested carefully in future surveys, as our study has some limitations.

The first obstacle is a consequence of the nature of semi-automated morphology analyses; it can only be as accurate, as the software performing it. In its current version, the EAS system used in this work mainly quantifies the number of 6- and 5-edged cells compared to pleomorphic cells. Some phenomena as cell agglomeration or degenerative aspects were not considered. A new version of the software will improve accuracy in the future.

The second limitation is that we measured the clinical outcome of corneal transplantation by comparing postoperative visual acuity and objective refraction. It is well known that there are a multitude of factors influencing these parameters — such as concurrent (known and unknown) ocular pathologies, subjective collaboration of the patient, size and position of the transplant, etc. While we presumed that — given a sufficient number of outcomes analyzed — these effects could cancel each other out, we admit that only a robust predictive corneal-quality marker could overcome all of these effects. This in mind — and under the premise that the main interfering factor might be the diagnostic background of the eye — we repeated the post-clinical comparative ROC analyses on two diagnostic groups (results not shown, see Supplementary Table). The results of these sub-analyses support the main observations and indicate that the poor/lacking predictive capacities of initial ECD/ECM (or their stability in organ culture) on the clinical outcomes of transplanted patients might be largely independent from their diagnosis. However, due to low (analyzable) case numbers, the reliability of these sub-analyses should be viewed critically — more comprehensive analyses are needed to test these observations.

Third, the vast majority of analyzed corneas was transplanted by penetrating keratoplasty (n = 178). In contrast, only 24 posterior-lamellar keratoplasties (DMEKs) were performed (see Table 2). Of course, this numerical imbalance does not allow reliable statements on putative differences of predictive powers of endothelial variables between these transplantation methods.

Finally, due to the retrospective setup of this study, we did not dispose on comprehensive data on corneal ECDs and ECMs after their transplantation. This data certainly is of considerable value and should be assessed in further studies.

In conclusion, we state that the initial endothelial cell density of donor corneas assessed before organ culture cultivation remains the only — and just limited — predictor for their final approval for transplantation, while the endothelial cell morphology seems to have no quantifiable importance. In contrast, the clinical outcome of visual acuity up to 2 years after transplantation could be less dependent from endothelial cell density (and morphology) than expected. These observations could suggest considerations on less restrictive cut-off levels for the endothelial cell density for transplantation approval; however, further studies are needed.