Be still my beating brain – reduction of brain micromotion during in vivo two-photon imaging

In light of the incredible complexity of the brain, the study of neural physiology in health and disease presents many challenges. However, with the advancement of powerful imaging tools, such as two-photon excitation fluorescence microscopy (2PEFM), we now have the opportunity to glimpse into the inner workings of the brain with high spatial and temporal resolution (Denk et al. 1990). This advanced imaging technique allows for examination of dynamic interactions and activity of cellular sub-types (e.g. microglia, astrocytes, neurons) and the surrounding brain microvasculature (Helmchen & Denk, 2005). While the advantages are many, a major challenge in capturing undistorted high-resolution images is the inherent rapid micromotion caused by respiration and heartbeat. In a recent issue of The Journal of Physiology, Paukert & Bergles (2012) present a novel methodology to overcome this confounding movement and enhance the 2PEFM imaging resolution of cellular structures.

Previous attempts to correct for brain micromotion, due to vital physiological functions, have included complex and invasive techniques such as muscle paralysis, cardiopulmonary bypass, and animal intubation. While these approaches may reduce micromotion they have profound effects on normal physiological function – ultimately confounding the very purpose of functional studies. In addition to physical attempts to improve specimen stability, post-imaging analysis is frequently incorporated, involving complex image manipulation. For example, many studies fit images to a Hidden Markov Model, a common statistical tool used in a variety of biomedical imaging contexts to extract data from noisy images (Dombeck et al. 2007; Chen et al. 2010). Statistical analyses such as these are highly useful, although they frequently require mathematical assumptions and don't reach the authenticity of pure image data. Moreover, not all motion artifacts can be corrected with existing statistical models, as micromotion displacements that occur during the acquisition of individual frames (such as those caused by heartbeats) are difficult to correct post hoc. More recently, Laffray et al. (2011) developed an approach to eradicate micromotion by using an optical stabilization sensor to dynamically refocus the objective onto the plane of interest, while monitoring the position of the animal. Although this may improve image stability, the device is costly and has yet to be implemented in commercial systems. As such, Paukert & Bergles (2012) sought to design and develop a cost-effective imaging approach to reduce in-frame variability, improve overall resolution and increase image stability.

Transgenic anaesthetized mice were used to image the cellular architecture of neuronal processes in cortical regions. 2PEFM imaging of dendrites was first performed by rapidly scanning one focal plane, as traditionally done. The authors found that motion artifacts were still present even when commonly used post hoc image registration was performed. Such artifacts led to movements of entire dendritic spines up to 1.5 μm, or shifts of ∼10% of the entire field of view, which produces profound limitations when imaging finer details of neural dendrites (Chen et al. 2011). Interestingly, comparing images that were acquired at the same time in the cardiac cycle significantly reduced the image distortions. These findings suggest that the beating of the heart is the major contributor to image distortion.

In order to control for brain micromotion caused by the beating of the heart, acquisition of individual frame scans were synchronized to the cardiac cycle. Specifically, each scan was triggered when an electrocardiogram (ECG) signal exceeded a set voltage, which occurred only during the R-wave of the cardiac cycle. Using this approach, the authors demonstrate that the dissimilarity between consecutively taken images was significantly reduced by R-wave-triggered scanning, compared to traditional non-discriminant scanning. This technique was also beneficial in reducing image distortion when consecutive sub-frames (1/10 of a full frame due to heartbeat triggered scanning) were interlaced and image stacks were constructed from multiple focal planes. Removing frames that were acquired during respiratory movement did not further improve image quality, supporting the claim that micro-displacements of dendritic spines are primarily due to the beating heart. Collectively, the findings of Paukert & Bergles (2012) demonstrate that syncing 2PEFM acquisition with cardiac activity improves image stability and spatial resolution.

In addition to demonstrating that the majority of in-frame neural tissue movements are due to heartbeat-induced distortions, this methodological approach may help facilitate the use of 2PEFM in a number of major areas. First, the novelty of this approach is that reliance on probabilistic assumptions is reduced, increasing the objectivity of highly sensitive data. Paukert and Bergles demonstrate that the image stability of small neural structures (dendritic spines) can be enhanced through the use of cardiac-triggered 2PEFM acquisition. In this regard, 2PEFM offers the ability to track morphological changes of individual dendritic spines over time periods of minutes to months (Hofer et al. 2008). Coupling ECG-based scanning and enhanced imaging stability, with studies of the synaptic structural plasticity, may greatly enhance our understanding of the workings of neuronal processes. Indeed, alterations in the reorganization and dynamic turnover of synapses has been linked to the development of the nervous system, to processing of experiences (e.g. learning) and neurodegenerative conditions, such as memory loss (Pan & Gan, 2008). Furthermore, 2PEFM absorption technology allows for micrometre scale disruption of biological processes, permitting the removal of sub-micrometre structures. However, ablation and photolysis rely on focusing the laser beam to a fixed point in space and are therefore highly sensitive to tissue movements. Cardiac-triggered ablation techniques, implementing Paukert and Bergles’ methodology, may facilitate and stabilize a wide range of ablation experiments, thereby enhancing our understanding of physiological/pathophysiological structures and processes.

Aside from enhancing image stability, the approaches employed by Paukert & Bergles (2012) may also aid studies focused on the functional aspects of single neuronal synapses. For example, 2PEFM photostimulation and uncaging of neurotransmitters (e.g. glutamate), permits the stimulation of individual synapses. Moreover, studies designed to investigate network activity rely on measuring calcium currents as a functional output. While the majority of these studies have been performed in vitro due to image stability issues, advancements have been made in vivo (e.g. Noguchi et al. 2011). Importantly, it is well known that anaesthesia has a profound effect on neural activity and it is evident that the design of studies will have to evolve to include conscious animals. In this regard, Dombeck et al. (2007) demonstrated an awake 2PEFM imaging preparation, which allows for investigation of neuronal structures and calcium imaging under resting and running conditions. While considered a significant advancement, a major hurdle regarding this particular animal preparation will be overcoming mechanical movement artifacts. The authors themselves noted that this preparation could work well in conjunction with motion-correcting software. In short, as optical techniques and in vivo preparations continue to advance, consideration of the beating of the heart on functional outcomes will need to be addressed and as such, Paukert and Bergles’ approach may prove to be beneficial.

While this methodology advances the potential uses of 2PEFM, triggering imaging scans by cardiac activity inherently limits temporal resolution. Paukert & Bergles (2012) worked with a high frame acquisition rate of 6.2 Hz on mice with a heart rate around 8.7 Hz. As such, cardiac triggered image acquisition had to be interrupted mid-scan. As mentioned by the authors, faster scanning techniques and interlacing subsequent scans to create an entire image frame may reduce this issue, but information will still unavoidably be lost during cardiac activity. While this is likely not to be an issue for more stable structures (dendrites) or even dynamic cellular movements (microglia), it may be a concern when imaging fast, beat-to-beat physiological changes. Indeed, 2PEFM has revealed pulsatile changes in cortical blood flow, coinciding with the cardiac cycle (Santisakultarm et al. 2012). Therefore, 2PEFM examinations of cerebral microvascular flow and/or neural–vascular coupling may not warrant the use of heartbeat-focused image acquisition, as crucial data may be lost.

Beyond the brain, continual technological advancements of 2PEFM have demonstrated the utility of this technique for in vivo imaging, from the spinal cord to pancreatic β cells (Dunn & Sutton, 2008). While each examination presents unique challenges, image stability remains a confounding factor for all of them. The brain in fact contends with the least amount of micromotion due to the structural confinements of the skull, whereas other areas are mechanically driven by respiratory and cardiac movements and are susceptible to image shifts of up to tens of micrometres. Indeed, 2PEFM studies examining mitochondrial function, physiological coupling and calcium currents in cardiac myocytes have been performed in Langendorff-perfused heart preparations to reduce the interference of physiological processes. Similarly, ex vivo preparations have been used to study other organs, such as the lung (Dunn & Sutton, 2008). It is therefore clear that the ECG-based acquisition of image frames, as outlined by Paukert & Bergles (2012), may prove to be beneficial for future examinations deeper within and outside of cortical brain regions.

In summary, Paukert & Bergles (2012) have provided an effective methodology to still the beating brain. The use of this technique, in conjunction with other movement correction strategies, may have wide ranging applicability and aid in our understanding of the complex physiology of the brain and beyond in both health and disease.