Fundamentals Series: Deep-Tissue Optical Recording and Stimulation in behaving animals. Part 4: Multiphoton microscopy – an overview

Written by Sias Jordaan

March 25, 2020

​The previous chapter in our Fundamentals Series on deep-tissue imaging focused on the Optical Fiberscope. The next method we’ll discuss is multiphoton microscopy (which includes two- and three-photon configurations). As with other chapters in this series, we intend to briefly:

  • Introduce the basic principles behind the method
  • List some of the main benefits and restrictions of the method
  • Refer to existing commercially available systems and
  • Look at the latest developments and indications of future trends


This technology has developed a lot in recent years, and many variations on the basic method have elevated this to a principle method in the neuroscience laboratory.


1. Basic Principles behind Multiphoton Microscopy


Laser scanning multiphoton microscopy was first described in a 1990 Science paper, where Denk et al explained how the known two-photon absorption effect can be applied to a laser-scanning microscope. This principle was applied to recording three-dimensional fluorescence images with reduced photobleaching and phototoxicity.

The two-photon absorption concept was first described by the Nobel-prize winning physicist, Maria Goeppert Mayer in 1931. She explained how a fluorophore molecule can be excited by the combined photon-energy of two photons – with these photons having longer wavelengths than the fluorophore’s emitted photons. A similar principle applies to (the more recently developed) three-photon absorption, but with the energy from three photons being absorbed by the fluorophore, instead of two.

The two-photon absorption adaptation by Denk described the application of this phenomenon to exciting fluorophores in living tissue, under a microscope with scan mirrors, to control the scan-path.

For this, an infra-red (IR) or near-IR (NIR) pulsed femtosecond laser, capable of generating pulse widths around 100fs is critical. The laser also needs to have a repetition rate of about 80mHz. This beam is directed by scan mirrors via adapted microscope optics, producing a focal point on the focal plane in or on the sample. At this point, illumination intensity is at its highest level. The probability of the fluorophore absorbing the combined energy of two photons for a single event increases exponentially at this point, and under these specific conditions. This absorption generates the Stokes shift necessary to emit photons with a wavelength which can be optically filtered and detected by a sensor (usually a photomultiplier tube (PMT)). The position of the scanning hardware (usually scan mirrors – see below for a discussion) is calibrated with the PMT detection in order to allocate a certain intensity of light detected to a scanned voxel on the sample. With this information, an image can be reconstructed by data acquisition software.

Figure 1. Simplified schematic of the critical components in a multiphoton setup

The core benefits to this method are

  • Since the excitation beam has a longer wavelength, the excitation pulse offers deeper penetration compared to the shorter excitation pulse used with conventional laser scanning microscopy
  • This excitation beam, being pulsed and of longer wavelength, carries less energy – this reduces sample damage in terms of phototoxicity and photobleaching.
  • Since the excitation happens only in the focal plane, it enables the system to obtain z-sections and 3-D information in thicker samples. There’s no need for pinholes (as with conventional confocal laser-scanning microscopy), leading to increased detection efficiency.


The core components (see Figure 1) of a multiphoton microscope are

  • the femtosecond pulsed IR laser,
  • launch optics, launching the laser beam into the scan-optics,
  • scan head, including the scan-optics core components. These enable the user to control the direction of the laser beam
  • a high Numerical Aperture (N.A.) and IR-corrected objective lens
  • the detection unit, including collection optics (to collect emitted photons) and sensitive detectors. Currently, preferred detectors are PMTs. The collection optics include filters that clean up the emission signal(s) and separate it from the excitation beam path
  • and the data acquisition system for reconstruction of the detected image based on intensity/time on the detector(s), calibrated to coincide with the excited voxel position.


The core benefits of the method make the multiphoton microscope the ideal tool for in vivo imaging. However, the challenge of achieving this in behaving rodents has only been successfully addressed in the last decade.

A breakthrough on this front has been the adaptation of a method for behavioural recordings in restrained insects, developed in the 1980’s – first described by the Dahmen lab in Tubingen.  In this paper, they fixed a beetle on the vertex of a Styrofoam sphere. The bottom half of the sphere was placed in a cup with air current vents, resulting in the ball floating on air. The Styrofoam sphere could, therefore, rotate in any direction with very little resistance to the animal’s movement – a multidimensional treadmill. The Dahmen lab also integrated a method for recording the directional movement of the beetle by calibrating movement of the sphere.

This method was adopted by various groups in more recent publications. One notable article from Dombeck et al., 2007 (Neuron) described the use of this air-supported Styrofoam sphere to allow more behavioral freedom whilst performing multiphoton imaging in head-fixed mice (See Figure 2). In a later publication, the same group further enriched the environment by adding virtual reality screens around the animal, investigating neural mechanisms of navigation. They also managed to do multiphoton recordings of place cells in the hippocampus by aspirating cortical tissue-layers and implanting a cannula for chronic recordings. In addition, they recorded with patch electrodes at the same time.

Figure 2: Schematic showing the mouse on the floating Styrofoam ball in the Dombeck et al. multiphoton microscope setup.

The mouse’s movement was recorded by two optical computer mice fitted onto the cup in which the Styrofoam ball was placed. The treadmill environment was further developed by several labs, and notably also by some commercial entities (see “currently available technologies” section below)  

2. Benefits


We referred earlier to the core benefits of the technique for optical recording and stimulation – below, we’ll elaborate and describe these in greater detail.


2.1. Non-Invasive Access and Deep-tissue Imaging


Optimized two-photon microscopes allow the user to record from brain tissue up to 1mm (under very specific conditions, and with limited resolution) below the tissue surface, and down to 1.5mm with three-photon approaches (values used in commercial material from manufacturers). This explains its popularity for studying cortical function across layers of the cortex, typically after creating a cranial window (either by thinning the skull or direct access, which can be minimally invasive) without having to insert an invasive probe or remove brain tissue.


The limit in imaging depth mentioned above is because of photons scattering when traveling through tissue medium. Targeting deeper regions than the 1-1.5mm mentioned above comes at a cost of invasiveness. This is done by tissue aspiration or implanting of optical fibers or lenses – often implanted cannulae or Gradient Index (GRIN) lenses. But these methods are now widely used and well published.  With tissue aspiration approaches (with implanted cannula when doing chronic recordings), the depth is restricted by concerns about lost network connectivity. The dimensions and the working distance of the objective lens are additional limitations. Most lenses are too big to position inside a cannula or aspirated area. The maximum imaging depth is therefore the distance between the desired focal plane and the brain or skull’s surface, or the top edge of the implant (see Figure 3).

Figure 3: Conventional objectives present an obstruction problem when targeting deeper areas with cannulae

A combined-objective approach, (where the imaging objective picks up photons from a GRIN lens implant) allows for deeper penetration, restricted by the length of the GRIN lens. However, the optical properties of the GRIN lens may affect spatial resolution and lead to unwanted aberrations. A more advanced approach to the objective-GRIN combination is the objective-mini-objective combination. Here, a miniaturized compound objective is combined with a regular objective to form an “endoscopic lens” 

Figure 4: The MCI Pryer Endoscopic Objective as an example of the objective-mini-objective combination

This improves image quality. However, compared to the GRIN lens, the mini-objective component (which dips into the tissue or down the cannula) has a wider diameter (at least 2.1mm, depending on the magnification) and cannot be implanted. 

Figure 5: Recording of GCaMP6 in dendritic spines of mouse hippocampus.

The mini-lens component of the Pryer endoscopic objective was lowered down an implanted cannula. Used in conjunction with an Olympus two-photon microscope.

2.2. Decreased Phototoxicity and Photobleaching


As explained in the introduction, when performing multiphoton excitation, the excitation beam has the following characteristics:

  • intermittent pulses, reducing total energy transferred to the tissue
  • longer wavelengths (NIR or IR), with lower energy values
  • excitation of only the focal point – nothing above or below


This means that phototoxicity and photobleaching are generally accepted to be reduced compared to widefield epifluorescence and single-photon confocal imaging. This reduction is particularly noticeable outside the focal point.

Reaching deep areas require higher laser power, so this remains a major consideration during experimental design.


2.3. Eliminating Out-of-Focus Fluorescence by Z-Sectioning


Image quality is greatly improved by the multiphoton method that only excites reporters in the focal point or focal plane when scanning in x and y. Additionally, this method enables 3D imaging, which is of special interest when performing functional imaging. The recordable data becomes even more interesting in terms of functional recordings with the feature of some microscopes to perform fast z-adjustments (see section below)


2.4. Higher Temporal Resolution 


With fast detection electronics, the current bottleneck for recording faster events in multiphoton setups is the scan rates in all three dimensions.

The early multiphoton scanning approaches used two galvanometric (galvo) mirrors (one for the X and one for the Y direction). For adjustment in Z, motorized or piezo-focus were the preferred mechanisms. Since the mirrors can be controlled very precisely, it is possible to reduce the scan area or target several smaller areas. This is done by doing single-line or point scans, for example targeting specific cell-bodies in the focal plane, which increases the scan rate compared to scanning the whole field of view.

Galvo/galvo-scanning is still popular, but the later integration of resonance-scanning mirrors offers increased scan rates in one direction (usually X), with a galvo mirror handling the Y-direction. However, controlling point-to-point scanning on the resonance mirror is more difficult. Therefore, this is not commonly used when the user has several smaller targets in the tissue volume.

Currently, the most advanced method for scanning control features acousto-optic (AO) crystals. Control of acoustic signals entering the crystals causes changes in its optical properties. The result is an instant change in the focal point of the very carefully aligned crystals. This method allows for fast control of the beam focal plane in the X, Y and (more recently) Z planes. In addition, it can also be used for selective scanning of reduced areas of interest, such as point scans, line scans or any other volume. Having access to all three dimensions offers a major advantage for fast, functional work in 3D volume. The development of improved genetically-encoded voltage indicators is bound to further increase demand for this fast 3D functionality.


2.5. Improved Spatial Resolution


Compared to other methods discussed in this Fundamentals blog-series, multiphoton methods offer excellent spatial resolution. This is reduced by scattering when targeting areas in deeper tissue, but sub-micron recordings of superficial dendritic spines are standard.

Objective N.A. is a major determining factor with spatial resolution. There have been interesting developments on this front with purpose-built multiphoton objectives and high-resolution endoscopic objectives (mentioned above).


2.6. Targeted and simultaneous imaging/stimulation


In a multiphoton setup, the laser beam used for imaging can also be used for optical stimulation, depending on the activation wavelength of the actuator/opsin. The more common method is, however, to implement two separate scan paths on the microscope – one for imaging (with a tunable laser) and one for stimulation (with a continuous wave or a femtosecond IR laser, depending on requirements and budget).

It is also relatively easy to integrate widefield single-photon light on a multiphoton setup, or to use a spatial illuminator (such as the Polygon1000) for simultaneous optical stimulation. This simultaneous stimulation and imaging are beneficial as it provides the possibility of not only imaging brain tissue in 3D, but also being able to manipulate specific neuronal activity.


2.7. A Wide Range of Commercially Available Systems


Since the patent (previously held by ZEISS) expired in 2009, several companies have developed innovative solutions. In many cases, this improved affordability and expanded potential applications and combinations with other techniques.

Several companies are targeting the Neuroscience arena specifically, developing turn-key solutions for neuro-specific applications. Some of the leading companies that supply various configurations (galvo/resonant/stimulating), specifically aimed at neuroscience are FemtonicsThorlabsScientifica and Sutter Instruments.  The only commercially available AO scanning systems are currently supplied by Femtonics and (more recently) Karthala.


3. Restrictions


Some of the main challenges or restrictions with multiphoton setups compared to other in vivo recording and stimulation techniques in behaving animals are:


3.1. Head-Fixation


Recording from behaving animals has been an important step forward for multiphoton technology. However, this usually requires head-fixation of an animal. Companies like Phenosys and Neurotar have developed interesting options for virtual reality systems and head-fixed maze navigation. They also commercialised useful accessories for more complex or native behavioural tasks. Still, head-fixation remains a limitation on experimental design and prevents some questions from being investigated using multiphoton systems (e.g. social interaction and grooming). In the section below, on new and future developments, we’ll mention the recent development of a multiphoton-enabled miniscope-hybrid, which will partly overcome this restriction.

On the other hand, the head-fixed approach offers more control and reduces movement artifacts. Despite this additional stability, movement artifacts play a role in most experiments. This becomes even more pronounced when imaging small structures, like neuronal processes, spines or boutons. In these instances, it is essential to build a compensating mechanism into your imaging protocol. Synchronizing imaging with heart rate and breathing is an approach that is sometimes implemented. Fast scanning approaches are also well suited to movement aberration correction, by applying algorithms to the acquisition software to correct for movement.


3.2. Cost


The cost of multiphoton systems is still prohibitive to many labs, with commercially available systems costing anything from £150k upwards. There is a growing library in the literature on how to build your own setup, and many groups or open-hardware projects like Nemonic continue to contribute. The affordability barrier has also been lowered by consulting companies (Cosys and INSS) offering hybrid-DIY solutions. Further developments of laser technology (e.g. fixed wavelength femto-lasers) may lead to further price reductions.


3.3. Spatial requirements (Large Footprint)


This is another area that has improved significantly in recent years, but most systems require a large optical table (typically starting at 2m x 1.5m) for mounting of the system.



4. New and Future developments


Multiphoton microscopy has gradually evolved over the last thirty years. The last decade has been particularly exciting, with many variations and extensions on the technique. This has been partly sparked by open competition between developers after the expiry of the patent. Some of these developments have been referred to in the “Benefits” section, but we’ll specifically describe the most recent developments for each, with a look to the future as well:


4.1. Temporal Resolution; increasing the recording speed for physiological events


The bottleneck with multiphoton temporal resolution is the rate at which you can scan the area of interest. The most established method of speeding up scan rates is to scale down the target area, as mentioned in the section above. Control algorithms for scanners are exploiting this principle for controlling the beam direction on galvo, resonance and AO scanners. AO methods for controlling the beam focal point in X, Y- and Z-directions opens many new possibilities. However, control of acousto-optic deflectors (AODs) presents a difficult engineering challenge, requires sophisticated optical compensation methods and is expensive to build.

To find alternative methods for upgrading scan rates, several recent methods papers used variations on “parallel scan” approaches. These authors divided the excitation beam into multiple beamlets to increase the scanned area. Fluorophores in the target volume of the sample were excited in quick succession (nanosecond scale), and processed with fast detection and acquisition electronics, thereby conserving the spatial resolution. In other instances, images were processed in parallel by using an array detector, using sCMOS or CCD cameras. All these methods compromised on some other aspects when compared to the performance of the AOD scanners.  Nonetheless, many more developments are sure to follow. These methods were summarized well be Spencer LaVere Smith in a recent blog post.


4.2. Purpose-Built Objectives; increasing the field of view, detection depth, and sensitivity


With Multiphoton technology’s wide acceptance over the last 15 years, big-brand microscopy companies could justify the continuous development of high-end, purpose-built objective lenses. This remains a field where it’s difficult for smaller companies to emulate the success of the large companies. It’s therefore not uncommon for new developments on this front to be the main selling point, which is then sold exclusively with the manufacturer’s complete system. The complete system drives revenue, and therefore adds more incentive for further development of objectives (and other competitive advantages). The trend pushes towards higher N.A., longer working distance, lower magnification objectives – ideally with excellent correction of optical aberration, and high transmission characteristics from around 450nm and into the IR.

A few very interesting niche lenses have also seen the light, for example Brad Amos’ mesolens for large area detection. Another example would be the MCI Pryer endoscopic objective (mentioned above…because we’re biased), for detecting deep tissue signals beyond the reach of more conventional objectives.

4.3. Increased Detector Sensitivity; enabling low-photon emission to be detected


Advances in material sciences and the ever-growing wider market for sensitive detectors, continue to push the boundaries with an expanding selection of detectors. Choosing the best option often involves a compromise of sensitivity for the sake of other important (and often dependent) variables. For example, it is essential to also consider the photosensitive surface area, its position within the detector housing, dead time, rise time, noise levels and cost. Ultimately, the goal is to have optimized, system-wide integration for improved signal-to-noise ratio, with careful consideration of the short and longer-term experimental design.

For several years now, the gold standard has been Gallium-Arsenide-Phosphide (GaAsP) detectors. A recent comparison in Optics Express concluded that the latest Silicon PMTs have better signal-to-noise values, particularly under very low-light conditions. With scanning methods speeding up and new genetically encoded voltage reporters likely to push demand for increased sensitivity, this may be a significant development.

Further to the parallel processing methods discussed under Temporal Resolution, array detectors offer an interesting alternative to the conventional “single-pixel” detectors (PMTs and photodiodes). The impressive developments in sCMOS cameras are likely to play a more prominent role as multiphoton detectors.


4.4. Laser Sources


As is the case with detector technology, the bulk of femtosecond laser applications lie outside of the multiphoton space, in micro or precision-fabrication. This wider market adds incentive for the development of these lasers, but it means that some necessary customisation specific to the microscopy niche may be lacking.

Intensity (across the relevant wavelength spectrum) and cost have been the obvious areas for improvement in multiphoton laser technology. Pushing intensity in the longer-wavelength ranges (typically longer tho 1300nm) also improves the potency of three-photon approaches. Three-photon excitation allows for deeper penetration of excitation pulses, due to reduced scattering, and has become much more popular in recent years.

Recent times have seen plenty of interest at trade shows in fixed wavelength lasers as a less expensive alternative to tunable lasers. Despite the apparent slow uptake of these lasers, it is probably a matter of time before the sub-$35k, 920nm femtosecond laser with 2W output power becomes a popular choice (to our knowledge, still unavailable at this stage).

Integration of peripherals (e.g. attenuation or built-in dispersion compensation) is further improving the cost aspect and optical efficiency. Ergonomic considerations also come into play, with the increase in popularity of smaller footprint or fiber-launched lasers.

A few years back, choices were limited to those offered by Newport (Spectra-Physics) and Coherent. This has now changed, with several new competitors surfacing (you still need to scroll down a bit to find these alternatives with a Google search!). Of course, this is a promising development for end-users. Expect more options for increased intensity and specifications in the longer IR-range, for improving imaging efficiency in deeper tissue layers and for three-photon microscopy. Having more intensity output also makes the prospect of splitting the beam into several beamlets more viable, in terms of parallel excitation experiments (as discussed under Temporal Resolution).


4.5. Peripherals; facilitating freedom and control in behaving animals


These developments in many instances originated from DIY improvisations that were very specific to the experimental need. Instances where these improvisations become more reproducible (from a technical and market-demand perspective) lead to production by companies. But the DIY trend of creative, purpose-built designs seems bound to continue its upward trajectory. Part of this is facilitated by the growing open-hardware community and components, as well as maker-community tools like 3D printing and prototype labs.

Other “peripherals” (not specifically aimed at the behavioural freedom dimension) have also grabbed the attention recently. An example of an easy, but useful integration would be digital micromirror devices (DMDs). These devices, commonly used in commercial projectors, offer control over the on/off position of individual mirrors in a high-density array. When integrated into the optical path, projection of light (LEDs or laser light) onto the mirror-array surface gives the user spatial control over light projected onto the sample. This “mask-effect” of DMDs can be implemented easily as a cost-effective alternative to integrating a separate scan path for photostimulation.

Another promising array-technology that has featured in many methods papers over the last decade, has been adaptive optics. This includes deformable mirrors and spatial light modulators. The fast-evolving applications of these devices revolve around control over the wavefront of excitation and emission light in the microscope. This means holographic control when doing 3D photostimulation, correcting for aberrations when imaging in deep tissue, as well as adjusting the focal plane optically (so without mechanical adjustment). The optical scanning option opens many possibilities in imaging systems where it is not possible to adjust focus mechanically, for example with implanted GRIN lenses or endoscopic objectives.



4.6. Ergonomics


Early commercial systems had a massive footprint. Whilst recent systems still can’t be described as compact compared to regular epifluorescence microscopes, options with reduced footprints are available. Multi-tier optical tables also offer improvements in space efficiency.  All these improvements make it easier to integrate with the behavioural system hardware (like jet-ball VR setups) without mechanical interference.

Furthermore, when accessing lateral brain areas in behaving animals, the objective is required to enter at an angle. This type of application has been made easier with commercial options like the Femtonics tilting objective and Thorlabs’ B-Scope,  having integrated robotic objective-arms to adjust the objective angle repeatably and accurately.

Finally, and not quite fitting under the “ergonomics” heading, but worth a fundamentals chapter on its own (keep an eye out for our next chapter), would be the integration of multiphoton capability to the concept of head-mounted miniscopes, or fiberscopes. This development overcomes (or will overcome, since it’s still in its infancy) the obstacle of having to have the animal head-fixed.



5. Conclusion


Multiphoton microscopy has been a popular technique for nearly 30 years and has established itself as a gold standard for in vivo imaging in Neuroscience. It has matured as a technology, but still offers much scope for further breakthrough developments. It has also reached critical numbers in the field to give it enough traction for a further cost reduction, which will give more labs access to this amazing technology.

Another factor in the momentum gain has been the easy piggybacking on industrial technology developments like detector, laser technology and adaptive optics. Imagining an even more exciting few decades ahead in the scope of multiphoton microscopy is, therefore, quite easy to do.

Read our previous Fundamentals chapters on Deep-Tissue Optical Recording and Stimulation:

Chapter 1: An Introduction

Chapter 2: Fiber Photometry

Chapter 3: Optical Fiberscope