Our new method for fluorescent marker photooxidation yielded high quality and detailed images of neuronal morphology in nearly all processed tissue samples ( and ). Numerous cell bodies, dendritic arbors, dendritic appendages and axonal projections were labeled in all DiI-stained regions ( and ). Dendritic terminal clusters (, Panels A–C) and axon terminal boutons (, Panel D) could be clearly identified. Very similar results were obtained in control photooxidation reactions performed using the Eclipse 80i microscope, which confirms that the quality of photooxidation-based DAB staining performed with the new apparatus is on equal footing with current standard methodology ( and Figure S3
). Likewise, the total illumination time necessary for high quality staining of neuronal cells was similar for both methods: 2 to 3 hours (). We were also capable of reliably visualizing the same set of cells under fluorescent (pre-photooxidation) and conventional (post-photooxidation) light microscopy without any obvious loss in staining quality (Figure S3
Representative photomicrographs of human retinal neurons and axons stained by the new photooxidation method.
In samples of developing rat neocortex, we retrieved images of DiI labeled Cajal-Retzius cells (). The large unipolar horizontal-projecting dendrite of the Cajal-Retzius cell was clearly identified, as well as the dendritic side branches and spine-like appendages that are typical of this cell type (, Panels A–B). The detailed visualization of these dendritic appendages highlights the acuity of the new method.
We were also able to successfully photooxidize DAB in fixed post-mortem human retinas. Most labeled neurons were identified as putative horizontal cells (, Panels A–C). Photomicrographs of DAB-labeled cells revealed detailed images of neuronal cell bodies, dendritic arbors and axonal projections (). Dendritic tufts, a typical characteristic of horizontal cells, were also clearly visualized (, Panels A–C).
The produced results are well within the quality range for morphological reconstructions or neurite mapping studies 
. Moreover, we confirmed that the quality of the images obtained with the new illumination method is at the same level as that obtained with a mercury lamp microscope, as both methods wielded highly detailed stainings of the entire cell and its processes (Figure S3
and ). However, a few requirements must be followed to obtain a high quality labeling: a) the DAB solution should be filtered before use; preferentially with filters with a 40 µm pore diameter or less; b) the DAB solution should be periodically exchanged during the photooxidation process; we obtained optimal results by exchanging the solution every 15 minutes; c) materials used in the experiment must be regularly cleaned and always kept free of chemical or biological contaminants; d) the main area of interest for photooxidation should be kept in the center of the LED's light beam during the reaction.
Methodological advantages of the new method
In addition to the already known advantages of photooxidation, including improved contrast and absence of photo-bleaching, our method makes the use of fluorescent microscopes unnecessary and allows for the simultaneous photooxidation of multiple labeled sites across large tissue samples, as it does not depend on the focal area of a microscope objective lens. These advantages make the method extremely cost and time efficient when compared with conventional photooxidation protocols, especially for large tissue samples, such as human retinas.
Wide-field illumination for fluorescent photooxidation
One of the main advantages of the photooxidizer apparatus is its relatively wide light beam in comparison to microscope-based illumination methods. We measured the effective illumination field of both a conventional fluorescent microscope and the photooxidizer by quantifying the area of tissue that was successfully photooxidized within a single reaction session (2–3 hours). The criterion for a successful photooxidation was the same in both experiments, i.e., the imaging of high resolution DAB staining in cell bodies and neurites. We found that the illumination field in both cases can be approximated to a perfect circle (). The region of tissue that was successfully photooxidized with the fluorescence microscope was fitted within a circle with approximately 1.32 mm of radius and an area of 5.5 mm2, while the region of tissue that was successfully processed with the photooxidizer was fitted within a circle with a radius of around 3.66 mm and area of 42 mm2 (). Therefore, the new photooxidizer apparatus is capable of processing an area of tissue over 7.5 times larger than what would be possible with currently available methods, all within similar time frames and quality standards. This methodological advantage may be of high value for the detailed investigation of relatively large cellular and histological structures, such as neuronal projection fibers and individual axonal collaterals.
Cost-efficiency of the new method
By eliminating the need for fluorescence microscopy, and consequently of mercury arc light bulbs, for the photooxidation reaction, the presented method provides a highly cost-efficient alternative to current protocols. This can be demonstrated by comparing the cost per hour of each methodological approach (summary in ). Currently, high efficiency mercury arc lamps for fluorescence microscopy have a marketed average lifetime of approximately 2,500 hours and cost around US$500. This implies an estimated average cost of US$0.20/hour of lamp usage. Our apparatus, on the other hand, has a total cost of less than US$100, including the high power LED and all of its electronic and custom built components. Given that the high power LED we used has a 100,000 hour life expectancy, photooxidation with our method costs less than US$0.001/hour of usage (). In other words, we estimate the new method to be at least 200 times cheaper than current photooxidation techniques.
This cost difference becomes even more significant when one factors in the advantages of wide-field illumination (). Considering the measured illumination field areas in the conventional fluorescence microscope (5.5 mm2) and the photooxidizer (42 mm2), the life expectancy of high-end mercury arc lamps (2,500 hours) and high power LEDs (100,000 hour) and the fact that a successful photooxidation reaction takes a maximum of 3 hours with both methods, we estimate that the maximum possible area of tissue photooxidized with a mercury lamp during its entire lifetime is approximately 4,583 mm2, while the maximum possible area of tissue photooxidized with our system during the entire lifetime of a high power LED is 1,400,000 mm2. By dividing the cost of the mercury lamp and the photooxidizer apparatus by these area estimates, we predict a lifetime cost per area of processed tissue of around US$0.11/mm2 for conventional microscope-based techniques, while the cost per area of processed tissue with our new method would be around US$0.00007/mm2 (). Thus, by combining the benefits of high power LEDs and wide field illumination, our new method can potentially surpass the cost efficiency of currently available techniques by three orders of magnitude. We believe this massive decrease in cost might represent a major advantage for high throughput structural research in cell biology and neuroscience, such as high resolution mapping of brain structures.
Note that all estimates presented here considered the total cost of the apparatus; a comparison strictly between the use of mercury lamps and high power LEDs would have yielded an ever greater cost difference. It should also be noted that while many companies have recently presented microscope models that use high-power LEDs as an illumination source, this technology is far from being the standard in cell biology research. Furthermore, even with a LED-based light source, photooxidation efficiency with these microscope models would still be restricted by the relatively small illumination field of microscope optics. Thus, our new method offers immediate advantages that, to our knowledge, are currently unavailable to the general scientific community.
Customizability and potential applications
Here we proved the concept of our new method by performing fluorescent photooxidation using intracellular DiI. However, the structure of the photooxidizer allows for a large degree of customization. By changing the high power LED for another with a different emission spectrum one can use this method to perform photooxidation reactions with any form of fluorescent marker, including commercial dyes and genetically encoded fluorescent proteins. Considering that LEDs have relatively broad emission spectra (), all that is needed is for the investigator to use a LED whose spectrum contains the appropriate excitation wavelength for the marker of choice. For example, an appropriate high power LED for photooxidation using the yellow fluorophore DiO should have an emission spectrum that encompasses the 484 nm wavelength range. For Lucifer Yellow, the ideal emission spectrum would need to include a 428 nm wavelength. The same principle can be applied for genetically encoded markers: a good LED for photooxidation using EGFP would have an emission spectrum that encompasses the 488 nm wavelength range 
, while for the novel MiniSOG marker it would need to include the 448 nm wavelength range 
. High power LEDs are cheap (approximately US$10) and easily replaceable within our apparatus schematics, both factors that highlight the customizability potential of the new approach.
Photooxidation techniques have been widely used in studies of neuronal tracing and of neurite morphology 
. The use of retrograde or anterograde fluorochromes for neurite labeling is a classical approach for visualizing cell morphology and long range projections in nervous tissue. Fluorescent DAB photooxidation improves these methods by eliminating the risk of information loss by photobleaching and by providing images with improved visual contrast for manual or computer-assisted neuronal reconstructions 
. Another main advantage of the DAB products generated by fluorescent photooxidation is that they are electron-dense and osmiophilic, thus allowing for visualization under transmission electron microscopy 
. Previous studies have applied this principle to visualize specific organelles, cellular compartments, lipoproteins and proteins 
. In this context, our approach has a broad range of potential applications, from the study of general cellular morphology, including the reconstruction of neuronal processes and long range mapping of neural projections, to the analysis of cellular and molecular ultrastructure under electron microscopy.