High-Resolution Episcopic Microscopy (HREM): Looking Back on 13 Years of Successful Generation of Digital Volume Data of Organic Material for 3D Visualisation and 3D Display

: High-resolution episcopic microscopy (HREM) is an imaging technique that permits the simple and rapid generation of three-dimensional (3D) digital volume data of histologically embedded and physically sectioned specimens. The data can be immediately used for high-detail 3D analysis of a broad variety of organic materials with all modern methods of 3D visualisation and display. Since its ﬁrst description in 2006, HREM has been adopted as a method for exploring organic specimens in many ﬁelds of science, and it has recruited a slowly but steadily growing user community. This review aims to brieﬂy introduce the basic principles of HREM data generation and to provide an overview of scientiﬁc publications that have been published in the last 13 years involving HREM imaging. The studies to which we refer describe technical details and specimen-speciﬁc protocols, and provide examples of the successful use of HREM in biological, biomedical and medical research. Finally, the limitations, potentials and anticipated further improvements are brieﬂy outlined.


Introduction
Imaging plays a central role in all areas of modern science and is one of the most expanding fields of biomedical research. Therefore, a plethora of highly sophisticated imaging methods spanning from ground-penetrating radar to Brillouin microscopy were developed in the last century [1][2][3].
In particular, the life sciences had a massive benefit from the availability of novel imaging modalities designed for 3D visualisation of living humans and animals, harvested embryos and tissue samples. Techniques such as single-photon emission computed tomography (SPECT) and positron emission tomography (PET), micro-computed topography (µCT) and optical coherence tomography (OCT) are already used as routine tools for the diagnosis of pathologies in the clinical routine [4][5][6][7][8][9]. Others, such as micromagnetic resonance tomography (µMRI), atomic force microscopy (AFM) and optical projection tomography (OPT), are still restricted to the scientific analysis and characterisation of biological samples [10][11][12][13][14][15][16]. Their focus rests on the interpretations of the three-dimensional (3D) arrangement of organs, tissues, cells and molecules in healthy and diseased organisms, which is the basis for understanding the genetic, epigenetic and functional aspects of developmental processes, the genesis of pathologies and the effect of strategies to treat diseases.
Each imaging method has unique advantages for a small field of applications. But due to technical constraints, most are restricted to highly specific research fields and a narrow selection of specimen types. A method that is highly potent, especially in structural 3D visualisation of embryos of biomedical models and biopsy material, is high resolution episcopic microscopy (HREM).
HREM [17] is an episcopic imaging method which generates a series of inherently aligned digital images of histologically processed and embedded specimens. The images are virtually stacked to be analysed by scrolling through the original or virtual resections and to lend themselves to volume rendering or segmentation and surface rendering. Thus, HREM is in line with other episcopic imaging techniques, such as serial block-face scanning electron microscopy (SBFSEM) [18], serial reconstruction technique [19][20][21], Epi-3D [22], surface imaging microscopy [23,24], imaging cryomicrotome [25], serial block-face imaging [26] and episcopic fluorescent image capturing (EFIC) [27], although it is optimised for a low microscopic level of resolution.
In a nutshell, materials are harvested and processed as for traditional histology. During dehydration, they are stained with eosin red mixtures, and resin (JB4) dyed with eosin or eosin/acridine orange is used for infiltration and as an embedding material. The resulting resin blocks are sectioned on a microtome, while digital images of subsequently exposed block surfaces are captured in fluorescence-mode using yellow fluorescent protein (YFP) (excitation 500/20 nm, emission 535/30 nm) or green fluorescent protein (GFP) (excitation 470/40, emission 525/50) filter sets for visualising eosin-contrasted structures. Cells and tissues specifically stained with LacZ or NBT/BCIP (nitro-blue tetrazolium and 5-bromo-4-chloro-3'-indolyphosphate) appear heavily contrasted when also using a Texas Red filter system as shown in Figure 1. Several protocol variations optimised for various materials are already published and extensively discussed [28][29][30][31][32].

Applications of HREM
HREM has been employed in various areas of the life sciences and for visualising a number of different organic materials. Its chief domains of application are phenotyping of embryos of biomedical model organisms and structural analysis of human biopsy material. Besides those, there HREM data typically have a voxel size of 2 × 2 × 2 µm 3 and are created from specimens with a volume of 6 × 6 × 6 mm 3 , although larger and smaller samples and higher or lower resolutions are possible. Data contrasts fit for identifying cells and nuclei, nerve fibres, capillaries and collagen bundles in the context of overall morphology and tissue architecture. Data generation varies according to specimen size, but it takes approximately 4-5 h for data comprised of 2000 single section images. Since numerous publications comprehensively describe the method [33][34][35][36] and ready-to-go HREM apparatuses are already commercially available (OHREM, Indigo Scientific Ltd., Baldock, England), this paper will not provide protocols or technical descriptions, but will focus on discussing published scientific results and potential applications.

Applications of HREM
HREM has been employed in various areas of the life sciences and for visualising a number of different organic materials. Its chief domains of application are phenotyping of embryos of biomedical model organisms and structural analysis of human biopsy material. Besides those, there are a number of unconventional applications.

HREM in Embryo Research
HREM was used for analysing whole embryos or embryonic organs and tissues of various species, including humans. However, the highest impact in this field was gained from phenotyping mouse embryos and from the selection of HREM as the tool for phenotyping mouse mutants harvested at embryonic day (E) 14.5 in the Deciphering the Mechanisms of Developmental Disorders (DMDD) program.

Mouse Embryos
Moderate breeding costs, short generation times and the existence of sophisticated molecular tools to manipulate the mouse genome facilitate efficient engineering of mutant mouse lines by random or targeted disruption of specific genes. Due to the conservation of basic developmental mechanisms and gene function, such lines are then used for studying normal mammalian gene function and the genetic components of diseases.
Identifying morphological defects of mouse embryos of genetically engineered mouse lines is an important step to characterise the function of the disrupted gene during embryo development and to define its role in the genesis of hereditary diseases and malformations. HREM was developed to permit the production of 3D volume data of whole E14.5 mouse embryos in a resolution and quality to fit for a holistic qualitative and quantitative analysis of embryo morphology, organ, blood vessel and nerve topology and tissue architecture as shown in Figure 2. In the scope of DMDD [37], which was linked with the International Mouse Phenotyping Consortium (IMPC), HREM was employed to build a virtual resource of fully annotated phenotype data of over 200 normal and over 500 mutant mouse embryos harvested at E14.5 stemming from 87 individual single knock-out or knock-down lines producing lethal or subvital homozygous offspring.    In the scope of this program, a standardised and ergonomic protocol for the comprehensive scoring of the morphologic phenotype of E14.5 embryos [38], a novel system for defining developmental substages of E14.5 mouse embryos [39], and reference data for the correct interpretation of phenotype abnormalities and for distinguishing variations from abnormalities [40] were created. Approximately 200 novel mouse phenotype (MP) terms were added to the MP ontology list as a result of HREM allowing for the detection of details that could not be detected with alternative imaging techniques. Numerous publications in highly ranked journals were produced based on HREM data generated in DMDD [38][39][40][41][42][43][44][45][46][47][48].
HREM has proved an excellent fit for pure descriptive but also metric analysis, which will have a great impact on researching the influence of biomechanical forces on prenatal tissue and organ formation and remodelling. Simple descriptive studies allowed, for example, the visualisation of pharyngeal arch artery development, proved the transitory existence of a 5th pharyngeal arch artery in the mouse and defined the dimensions of large arteries in mouse embryos bred on various genetic backgrounds [69][70][71][72][73][74]. Precise visualisation of septation and cardiac cushion, the outflow tract and valve development were used as a basis for learning the concepts underlying associated malformations [75][76][77][78][79][80][81][82][83][84][85][86]. Besides its importance for research, digital 3D models are of course eagerly anticipated teaching aids, as they can exemplify complex developmental remodelling processes.
Simple metric data were produced for defining the dimensions of the great intrathoracic arteries of embryos bred on various genetic backgrounds. However, based on HREM data, more sophisticated models could be calculated, which could mathematically define the description of the heart looping process [52] or use 3D fractal analysis for defining the complexity of trabeculation during ventricular development of the heart [87-89].

Chick Embryos
Chick embryos of very early to late developmental stages were already examined, as shown in Figure 2. Embryos of the blastoderm stage [90,91] were used for researching the effect that the storage of eggs has on embryonic development and viability. Embryos of late developmental stages have been successfully used for analysing the effect of hemodynamic alterations on the topology and remodelling of the cardiac outflow [72], and for producing metric definition of the dimensions of the great intrathoracic arteries to serve as reference data for such studies [73]. In addition, HREM analysis of an embryo with cephalothoracopagus malformation was successfully conducted, which permitted the formulation of a novel concept explaining the genesis of this abnormality [92].

Embryos/Fetuses of Other Species
HREM is not restricted to research mouse and chick embryos. It also has been shown to work on a broad variety of embryos of biomedical models. This includes quail and Xenopus embryos, where the proof of principle was published [17,28], but also includes the zebrafish, for which HREM was shown to fit for visualisation and mathematical characterisation of complex developmental processes, such as dentation and cardiac jogging [17,93].
Yet, HREM is not restricted to small biomedical models. 3D visualisation of voluminous fetal material is also possible. In particular, joint development was visualised in horse fetuses [94]. Furthermore, the potential of HREM to analyse the heart morphology of human fetal hearts in the first trimester has been shown [95,96]. Also, other (data not yet published) fetal materials fit as well for HREM imaging, but due to restricted availability of human embryos for morphological studies, collecting such materials is rather problematic and limited to highly specific scientific projects.

HREM for Visualising Adult Material
Though originally developed for imaging embryonic samples, protocols are already available for preparing biopsy material from adult individuals of various species for HREM imaging, as shown in Figure 3. Examples are materials stemming from rodents, zebrafish, turquoise killfish, ferrets and the fruit fly, for which there are preliminary data, but no publications yet exist.

HREM for Other Organic Materials
The materials primarily subjected to HREM imaging are sourced from animals. However, very recently, HREM expanded its application into the realm of plant sciences. More concretely, the morphology of wild-type and genetically altered tomato plants were visualised in order to study abnormal leaf axil patterning [105]. In pilot studies, the capacity of HREM for testing paper quality [28,106] and the fibre arrangement and the formation of keratinocytes seeded on skin replacement material [107] were evaluated. These experiments demonstrated that HREM is not restricted to biomedical research, but has yet-unknown capacities to aid research in many other fields of modern science.

Conclusions and Further Perspectives
HREM was developed in 2006, and for about a decade, operated on self-assembled apparatuses. Published data does exist for the pig and the mouse, in which vascularisation in wound healing was researched [97,98], and for humans. In humans, HREM was successfully employed to develop a novel concept for dermal vascularisation in thick [99,100] and thin skin [101] and to analyse the topology of arteries and nerves in the auricle [102]. Besides this, it recently permitted the characterisation of plaques in the coronary arteries [103].
HREM could also show the structure of dermal matrix skin substitutes with and without seeded keratinocytes, and the effect of dermal matrix skin substitutes in combination with skin graft transplants on vascularisation in a porcine wound model. Recently, HREM was an integral part of a multimodal imaging pipeline for comprehensively characterising the vascularisation of murine tumour models [104].

HREM for Other Organic Materials
The materials primarily subjected to HREM imaging are sourced from animals. However, very recently, HREM expanded its application into the realm of plant sciences. More concretely, the morphology of wild-type and genetically altered tomato plants were visualised in order to study abnormal leaf axil patterning [105]. In pilot studies, the capacity of HREM for testing paper quality [28,106] and the fibre arrangement and the formation of keratinocytes seeded on skin replacement material [107] were evaluated. These experiments demonstrated that HREM is not restricted to biomedical research, but has yet-unknown capacities to aid research in many other fields of modern science.

Conclusions and Further Perspectives
HREM was developed in 2006, and for about a decade, operated on self-assembled apparatuses. In 2015, a fully operable, all-inclusive HREM machine became commercially available under the trade name of 3D Optical HREM imaging (OHREM). This increased the number of projects using the HREM technique, which was illustrated by a quadruplication of publications based on HREM data over the last ten years, and it boosted the development of new protocols and technical advancements. However, HREM is still to be considered as a technique in its beginnings with great potential for refinement and improvement. As pilot data has shown, it has a yet-unexplored high potential for visualising a broad variety of organic materials in unmatched resolution and data quality.

Stitching
One of the strengths of HREM is its ability to image volumes of up to 6 × 6 × 12 mm 3 in a numeric resolution of 3 × 3 × 3 µm 3 . Increasing the numeric resolution is possible, but requires focusing on a smaller volume to be scanned [34,108]. First prototypes of HREM machines, which scan several images of the same block-surface and incorporate stitching algorithms for combining them, overcome these limitations and are already in use. It is to be expected that this promising approach will soon advance to become the HREM routine.

Pipelines
Even by using block scanning and stitching, the use of HREM is limited to exploring the microanatomy of specimens of a relatively small size. For large specimens, HREM imaging has to be combined with other imaging modalities to gain HREM detail in the context of overall specimen information. Since this is quite simple with techniques such µMRI, the first imaging pipeline that includes HREM dates back to the very early phase of HREM imaging [51]. Large batches of embryos were scanned with high-throughput µMRI in moderate resolution. Interesting specimens or specimen parts were then identified, selected and subjected to HREM imaging for providing tissue detail. This approach was modified and expanded in the last years [109], and the first publications providing protocols and demonstrating the benefits of high complex multimodal, multiscale imaging pipelines integrating HREM with imaging modalities such as µMRI, US, µCT, OCT, PAT and histopathology are already in preparation [104].

Specific Stainings
HREM data volumes comprise thousands of digital images, each resembling a greyscale image of a hematoxylin-eosine-stained histological section. Thus, HREM offers 3D information of morphological details of organic materials in a straightforward and simple way. Attempts to expand HREM to permit 3D visualisation of specifically labelled structures and molecular signals as well are as old as the HREM technique itself. Even the first publication included an example for visualising LacZ-stained tissues in mouse embryos and NBT/BCIP signals after whole-mount in situ hybridisation in zebrafish embryos [17]. However, despite these efforts, visualising specific stained tissues is still experimental and restricted to small specimens and specimens composed of loose and easy-to-penetrate tissues [103,110,111]. Protocols for late embryos and dense tissues are eagerly anticipated and will open new fields of applications for HREM. We are confident that with the growing community of HREM developers and users, solutions for this problem will be presented within the next few years.