Significance Recently, neuroscience has attracted great attention around the world. To prompt the study of neuroscience, a lot of countries have launched brain projects, in which the development of advanced neural techniques is regarded as the driving force. Optical techniques own the advantages of being less-invasive and high spatial resolution, etc, promising for neural activity recording and manipulation. Compared to traditional electrode stimulation methods, optical stimulation based on optogenetics could selectively excite or inhibit specific neural ensembles, benefiting from the introduction of gene engineering. So far, a variety of opsins have been developed for the activation or inhibition of neural activity. On the other hand, to achieve selective manipulation at single-neuron resolution, the techniques for two-photon optogenetics are emerging. Here, we review various strategies for illumination in two-photon optogenetics. We summarize their technical principles, and discuss their advantages and disadvantages.

Progress Illumination strategies in optogenetics can be classified as conventional illumination strategies based on single-photon absorption and high-spatial-resolution illumination strategies based on two-photon absorption.

In the early days of optogenetics, the wide-field illumination strategy based on single-photon absorption was used to manipulate neurons with opsins. Due to the scattering and absorption of biological tissues, the power of illumination light decreases significantly with the increase of penetration depth in wide-field illumination strategy. For neuron manipulation in deep tissues, fiber-coupled illumination is performed, in which the excitation light is guided through the fiber to the targeted depth. However, specific manipulation of neural activity at single-neuron resolutions is not achieved in neither wide-field illumination nor fiber-coupled illumination, due to the fact that all neurons with opsins in the illumination region would be excited.

For specific manipulation of neural activity at single-neuron resolution or sub-neuron resolution (such as a dendrite or a dendritic spine), the two-photon illumination strategy has to be adopted, which ensures high spatial resolution in three-dimensional (3D) space. Besides, the longer wavelength in two-photon excitation is more robust to scattering and leads to a deeper penetration depth than that of the conventional illumination strategies.

In general, the two-photon illumination strategies can be classified as serial scanning illumination and parallel illumination based on phase modulation. In the former, a single focus is steered to perform spiral scanning on a neuron to open enough ion channels for neuron excitation before being switched to another neuron (Fig. 1). Or, serial scanning with soma-patterned illumination can be employed (Fig. 2). However, the temporal resolution in these methods is low, which makes it only compatible with opsins of slow turn-off time. Besides, parallel illumination can be achieved based on phase modulation (Fig. 3), in which the phase can be calculated by the generalized phase contrast (GPC) method and the computer generated holography (CGH) method. The phase modulation plane of GPC is the conjugate plane of the focal plane, while that of CGH is the conjugate plane of the Fourier plane. The GPC method has good uniformity of excitation patterns but is of a poor energy efficiency. Based on the theory of CGH, two illumination schemes have been developed, i.e. the multi-foci generation combined with the spiral scanning strategy, and the multiple extended-pattern generation strategy. The former has a high energy efficiency and can realize the excitation of 3D distributed neurons, but it is of poor temporal resolution and only works with opsins of slow turn-off response time. The latter can directly generate a multiple expanded-pattern and excites multiple neurons simultaneously. Combined with the temporal focusing technology, this method has high temporal resolution and high axial resolution, however, its energy efficiency is low. At the same time, we summarize the commonly used CGH algorithms for parallel illumination based on phase modulation, which contain superposition algorithm, Gerchberg-Saxton (GS) algorithm, non-convex optimization (NOVO) algorithm, and DeepCGH algorithm. The diagrams of different algorithms are presented (Fig. 4). The basic ideas, advantages, and disadvantages of these algorithms are briefly pointed out.

Conclusion and Prospect Conventional neural manipulation in vivo relies on single-photon illumination, which is not good for specific excitation of neural ensembles at high spatial resolution. To this end, several techniques of two-photon optogenetics have been proposed, and have achieved in vivo neural manipulation at high-spatial-resolution. We summarize the development of two-photon illumination strategies, and compare their advantages and disadvantages. Then we discuss the potential issues in the practical employment of two-photon optogenetics, such as excitation precision and field-of-view. We expect that, the all-optical physiology, in which two-photon imaging and two-photon optogenetics are combined, is promising in neuroscience, benefiting from the simultaneous monitoring and manipulating of neural circuits in vivo.