Recent advances in super-resolution microscopy have pushed the resolution limit of light microscopy closer to that of electron microscopy. However, as they invariably rely on fluorescence, light microscopy techniques only visualize whatever gets labeled. On the other hand, while electron microscopy reveals cellular structures at the highest resolution, it offers no specificity. The information gap between the two imaging modalities can only be bridged by correlative light and electron microscopy (CLEM). Previously we have developed a probe (mEos4) whose fluorescence and photoconversion survive 0.5-1% OsO4 fixation, allowing super-resolution visualization of organelles and fused proteins in the context of resinembedded ultrastructure in both transmission EM (TEM) and scanning EM (SEM) [1,2].

Correlative light and electron microscopy (CLEM) combines the power of electron microscopy, with its excellent resolution and contrast, with that of fluorescence imaging, which allows the staining of specific molecules, organelles, and cell populations. Fluorescence imaging is also readily compatible with live cells and behaving animals, facilitating real-time visualization of cellular processes, potentially followed by electron microscopic reconstruction. Super-resolution single-molecule localization microscopy is a relatively new modality that harnesses the ability of some fluorophores to photoconvert, through which localization precision better than Abbe’s diffraction limit is achieved through iterative high-resolution localization of single-molecule emitters. Here we describe our lab’s recent progress in the development of reagents and techniques for super-resolution single-molecule localization CLEM and their applications to biological problems.

To truly understand biological systems, one must possess the ability to selectively manipulate their parts and observe the outcome. (For purposes of this review, we refer mostly to targets of neuroscience; however, the principles covered here largely extend to myriad samples from microbes to plants to the intestine, etc.).

Drugs are the most commonly employed way of introducing such perturbations, but they act on endogenous proteins that frequently exist in multiple cell types, complicating the interpretation of experiments. Whatever the applied stimulus, it is best to introduce optimized exogenous reagents into the systems under studydenabling manipulations to be targeted to speci!c cells and pathways. (It is also possible to target manipulations through other means, such as drugs that acquire cell-type speci!city through targeting via antibodies and/or cell surface receptor ligands, but as far as we are aware, existing reagents fall short in terms of necessary speci!city.) Many types of perturbations are useful in living systems and can be divided into rough categories such as the following: depolarize or hyperpolarize cells, induce or repress the activity of a speci!c pathway, induce or inhibit expression of a particular gene, activate or repress a speci!c protein, degrade a speci!c protein, etc. User-supplied triggers for such manipulations to occur include the following: addition of a small molecule (“chemogenetics”dideally inert on endogenous proteins) [1], sound waves (“sonogenetics”) [2], alteration of temperature (“thermogenetics”d almost exclusively used for small invertebrates) [3], and light (“optogenetics”). There are reports of using magnetic !elds (“magnetogenetics”) [4], but there is no evidence that such effects are reproducible or even physically possible [5,6]. Of these, the most commonly used, for multiple reasons, is light.

Many factors make light an ideal user-controlled stimulus for the manipulation of samples. Light is quickly delivered, and most light-sensitive proteins and other molecules respond quickly to light stimuli, making many optogenetic systems relatively rapid in comparison to, for instance, drug-modulated systems. Light is also quite easy to deliver in localized patterns, allowing for targeted stimulation. Multiple wavelengths can be delivered separately to distinct (or overlapping) regions, potentially allowing combinatorial control of diverse components. Finally, light can be delivered to shallow brain regions (and peripheral sites) relatively noninvasively, and to deeper brain regions with some effort.

However, there are also a number of shortcomings of using light for control. Robust and uniform penetration of light into the sample is the most signi!cant concern. For systems requiring modulation of many cells, particularly at depth, the use of systems controlled by small molecule drugs would generally be recommended instead of optogenetic approaches. When light is delivered through the use of !bers, lenses, or other optical devices, such interventions can produce signi!- cant cellular death, scar formation, and biofouling. The foreign-body response of tissue to objects triggers substantial molecular alterations, the implications of which are incompletely de!ned, but can involve reactive astrogliosis, oxidative stress, and perturbed vascularization. Head-mounted lightdelivery devices can be heavy and/or restrictive, and thus perturb behavior, particularly for small animals (e.g., mouse behavior is much more disrupted than rat behavior). More generally, all light causes tissue heating, which can have dramatic effects on cell health, physiology, and animal behavior. This is most concerning for tiny animals such as "ies. Light itself also damages tissue, most obviously through photochemistry (e.g., oxidation and radicalization) and photobleaching of critical endogenousmolecules. Furthermore, of course, light is ubiquitous, meaning that the sample is never completely unstimulated, despite precautions. Light passes through the eyes into the brain with surprising ease, and even through the skull with modest ef!cacy [7]dwhich can disrupt animal behavior (as can the converse: stimulating light in the brain perceived as a visual stimulus through the back of the eyes.) Light-responsive proteins exist in all samples, particularly in the eyes but to some extent in all tissuesdnotably, deep-brain photoreceptors [8].

The use of optogenetic tools has accelerated research on many fronts in disparate !elds. Additional, perhaps most, limitations on the utility of optogenetics must, however, be placed squarely on the shortcomings of the current suite of tools (and potential inherent limits in their performance.) The vast majority of optogenetic effectors are gated by blue light, which has signi!cant penetration issues and can be phototoxic under high intensity; redder wavelengths would in general be preferred. Furthermore, multiplexing requires tools making use of other parts of the visible spectrum (and redder wavelengths). A related issue is that most chromophores for optogenetic reagents have very broad action spectra (w250 nm bandwidth for retinal; w200 nm bandwidth for "avin), complicating both multiplexing and their use alongside many optical imaging reagentsdnarrower action spectra would be preferred for effectors in most situations. More generally, the current classes of optogenetic effectors are few, mostly limited to (1) channels and pumps (most with poor ion selectivity), (2) dimerizers, and (3) a handful of enzymes. The number of optogenetic tools that perform a very speci!c function in cells is small. Although progress has undeniably been made, much additional research and engineering will be required to dramatically expand the optogenetic toolkit.

Rather than providing a survey of research !ndings, this review covers general considerations of optogenetics experiments, and then focuses largely on molecular tools: the existing suite, their features and limitations, and goals for the creation and validation of additional reagents.