A novel application of motion analysis for detecting stress responses in embryos at different stages of development

Bioimaging and the future of whole-organismal developmental physiology

While omics has transformed the study of biology, concomitant advances made at the level of the whole organism, i.e. the phenome, have arguably not kept pace with lower levels of biological organisation. In this personal commentary we evaluate the importance of imaging as a means of measuring whole organismal developmental physiology. Image acquisition, while an important process itself, has become secondary to image analysis as a bottleneck to the use of imaging in research. Here, we explore the significant potential for increasingly sophisticated approaches to image analysis, including deep learning, to advance our understanding of how developing animals grow and function. Furthermore, unlike many species-specific methodologies, tools and technologies, we explore how computer vision has the potential to be transferable between species, life stages, experiments and even taxa in which embryonic development can be imaged. We identify what we consider are six of the key challenges and opportunities in the application of computer vision to developmental physiology carried out in our lab, and more generally. We reflect on the tangibility of transferrable computer vision models capable of measuring the integrative physiology of a broad range of developing organisms, and thereby driving the adoption of phenomics for developmental physiology. We are at an exciting time of witnessing the move from computer vision as a replacement for manual observation, or manual image analysis, to it enabling a fundamentally more powerful approach to exploring and understanding the complex biology of developing organisms, the quantification of which has long posed a challenge to researchers.

LabEmbryoCam: An opensource phenotyping system for developing aquatic animals

Phenomics is the acquisition of high-dimensional data on an individual-wide scale and is proving transformational in areas of biological research related to human health including medicine and the crop sciences. However, more broadly, a lack of accessible transferrable technologies and research approaches is significantly hindering the uptake of phenomics, in contrast to molecular-omics for which transferrable technologies have been a significant enabler. Aquatic embryos are natural models for phenomics, due to their small size, taxonomic diversity, ecological relevance, and high levels of temporal, spatial and functional change. Here, we present LabEmbryoCam, an autonomous phenotyping platform for timelapse imaging of developing aquatic embryos cultured in a multiwell plate format, and while optimised for embryos, the instrument is extremely versatile. The LabEmbryoCam capitalises on 3D printing, single board computers, consumer electronics and stepper motor enabled motion. We combine these into a compact and modular laboratory insturment to provide X, Y and Z motion of a camera and lens, a web application streamlined for rapid setup of experiments, user email notifications and a humidification chamber to reduce evaporation over prolonged acquisitions. Downstream analyses are provided, enabling automated embryo segmentation, heartrate measurement, motion tracking, and energy proxy trait (EPT) measurement. The LabEmbryoCam is a scalable, and flexible laboratory instrument, that leverages embryonic and early life stage organisms to tackle key global challenges including biological sensitivity assessment, toxicological screening, but also to support broader engagement with the earliest stages of life.

Mild cognitive impairment prediction based on multi-stream convolutional neural networks

BACKGROUND

Mild cognitive impairment (MCI) is the transition stage between the cognitive decline expected in normal aging and more severe cognitive decline such as dementia. The early diagnosis of MCI plays an important role in human healthcare. Current methods of MCI detection include cognitive tests to screen for executive function impairments, possibly followed by neuroimaging tests. However, these methods are expensive and time-consuming. Several studies have demonstrated that MCI and dementia can be detected by machine learning technologies from different modality data. This study proposes a multi-stream convolutional neural network (MCNN) model to predict MCI from face videos.

RESULTS

The total effective data are 48 facial videos from 45 participants, including 35 videos from normal cognitive participants and 13 videos from MCI participants. The videos are divided into several segments. Then, the MCNN captures the latent facial spatial features and facial dynamic features of each segment and classifies the segment as MCI or normal. Finally, the aggregation stage produces the final detection results of the input video. We evaluate 27 MCNN model combinations including three ResNet architectures, three optimizers, and three activation functions. The experimental results showed that the ResNet-50 backbone with Swish activation function and Ranger optimizer produces the best results with an F1-score of 89% at the segment level. However, the ResNet-18 backbone with Swish and Ranger achieves the F1-score of 100% at the participant level.

CONCLUSIONS

This study presents an efficient new method for predicting MCI from facial videos. Studies have shown that MCI can be detected from facial videos, and facial data can be used as a biomarker for MCI. This approach is very promising for developing accurate models for screening MCI through facial data. It demonstrates that automated, non-invasive, and inexpensive MCI screening methods are feasible and do not require highly subjective paper-and-pencil questionnaires. Evaluation of 27 model combinations also found that ResNet-50 with Swish is more stable for different optimizers. Such results provide directions for hyperparameter tuning to further improve MCI predictions.

Spectral phenotyping of embryonic development reveals integrative thermodynamic responses

Background

Energy proxy traits (EPTs) are a novel approach to high dimensional organismal phenotyping that quantify the spectrum of energy levels within different temporal frequencies associated with mean pixel value fluctuations from video. They offer significant potential in addressing the phenotyping bottleneck in biology and are effective at identifying lethal endpoints and measuring specific functional traits, but the extent to which they might contribute additional understanding of the phenotype remains unknown. Consequently, here we test the biological significance of EPTs and their responses relative to fundamental thermodynamic principles. We achieve this using the entire embryonic development of Radix balthica, a freshwater pond snail, at different temperatures (20, 25 & 30 °C) and comparing responses against predictions from Arrhenius’ equation (Q₁₀ = 2).

Results

We find that EPTs are thermally sensitive and their spectra of frequency response enable effective high-dimensional treatment clustering throughout organismal development. Temperature-specific deviation in EPTs from thermodynamic predictions were evident and indicative of physiological mitigation, although they differed markedly in their responses from manual measures. The EPT spectrum was effective in capturing aspects of the phenotype predictive of biological outcomes, and suggest that EPTs themselves may reflect levels of energy turnover.

Conclusions

Whole-organismal biology is incredibly complex, and this contributes to the challenge of developing universal phenotyping approaches. Here, we demonstrate the biological relevance of a new holistic approach to phenotyping that is not constrained by preconceived notions of biological importance. Furthermore, we find that EPTs are an effective approach to measuring even the most dynamic life history stages.

Supplementary information

The online version contains supplementary material available at 10.1186/s12859-021-04152-1.

Development of cardiovascular and neurodevelopmental metrics as sublethal endpoints for the Fish embryo toxicity test

The fathead minnow fish embryo toxicity (FET) test has been proposed as a more humane alternative to current toxicity testing methods as younger organisms are thought to experience less distress during toxicant exposure. However, the FET test protocol does not include endpoints that allow for the prediction of sublethal adverse outcomes, limiting its utility relative to other test types. Researchers have proposed the development of sublethal endpoints for the FET test to increase its utility. The present study 1) developed methods for previously unmeasured sublethal metrics in fathead minnows (i.e., spontaneous contraction frequency and heart rate) and 2) investigated the responsiveness of several sublethal endpoints related to growth (wet wt, length, and growth-related gene expression), neurodevelopment (spontaneous contraction frequency, eye size, and neurodevelopmental gene expression), and cardiovascular function and development (pericardial area, heart rate, and cardiovascular system-related gene expression) as additional FET test metrics using the model toxicant 3,4-dichloroaniline. Of the growth, neurological, and cardiovascular endpoints measured, length, eye size, and pericardial area were found to be more responsive than the other endpoints evaluated. Future studies linking alterations in these endpoints to longer-term adverse impacts are needed to fully evaluate the predictive power of these metrics in chemical and whole-effluent toxicity testing. Environ Toxicol Chem 2018;37:2530-2541. © 2018 SETAC.

Multi-generational responses of a marine polychaete to a rapid change in seawater pCO2

Little is known of the capacity that marine metazoans have to evolve under rapid p CO ₂ changes. Consequently, we reared a marine polychaete, Ophryotrocha labronica, previously cultured for approximately 33 generations under a low/variable pH regime, under elevated and low p CO ₂ for six generations. The strain used was found to be tolerant to elevated p CO ₂ conditions. In generations F1 and F2 females' fecundity was significantly lower in the low p CO ₂ treatment. However, from generation F3 onwards there were no differences between p CO ₂ treatments, indicating that trans-generational effects enabled the restoration and maintenance of reproductive output. Whilst the initial fitness recovery was likely driven by trans-generational plasticity (TGP), the results from reciprocal transplant assays, performed using F7 individuals, made it difficult to disentangle between whether TGP had persisted across multiple generations, or if evolutionary adaptation had occurred. Nonetheless, both are important mechanisms for persistence under climate change. Overall, our study highlights the importance of multi-generational experiments in more accurately determining marine metazoans' responses to changes in p CO ₂, and strengthens the case for exploring their use in conservation, by creating specific p CO ₂ tolerant strains of keystone ecosystem species.

An embryonic transcriptome of the pulmonate snail Radix balthica

The pond snail, Radix balthica (Linnaeus 1758), is an emerging model species within ecological developmental biology. While its development has been characterised in detail, genomic resources for embryonic stages are lacking. We applied Illumina MiSeq RNA-seq to RNA isolated from pools of embryos at two points during development. Embryos were cultured in either the presence or absence of predator kariomones to increase the diversity of the transcripts assembled. Sequencing produced 47.2M paired-end reads, assembled into 54,360 contigs of which 73% were successfully annotated. This transcriptome provides an invaluable resource to build a mechanistic understanding of developmental plasticity.

Challenges and opportunities in developmental integrative physiology

This review explores challenges and opportunities in developmental physiology outlined by a symposium at the 2014 American Physiological Society Intersociety Meeting: Comparative Approaches to Grand Challenges in Physiology. Across animal taxa, adverse embryonic/fetal environmental conditions can alter morphological and physiological phenotypes in juveniles or adults, and capacities for developmental plasticity are common phenomena. Human neonates with body sizes at the extremes of perinatal growth are at an increased risk of adult disease, particularly hypertension and cardiovascular disease. There are many rewarding areas of current and future research in comparative developmental physiology. We present key mechanisms, models, and experimental designs that can be used across taxa to investigate patterns in, and implications of, the development of animal phenotypes. Intraspecific variation in the timing of developmental events can be increased through developmental plasticity (heterokairy), and could provide the raw material for selection to produce heterochrony--an evolutionary change in the timing of developmental events. Epigenetics and critical windows research recognizes that in ovo or fetal development represent a vulnerable period in the life history of an animal, when the developing organism may be unable to actively mitigate environmental perturbations. 'Critical windows' are periods of susceptibility or vulnerability to environmental or maternal challenges, periods when recovery from challenge is possible, and periods when the phenotype or epigenome has been altered. Developmental plasticity may allow survival in an altered environment, but it also has possible long-term consequences for the animal. "Catch-up growth" in humans after the critical perinatal window has closed elicits adult obesity and exacerbates a programmed hypertensive phenotype (one of many examples of "fetal programing"). Grand challenges for developmental physiology include integrating variation in developmental timing within and across generations, applying multiple stressor dosages and stressor exposure at different developmental timepoints, assessment of epigenetic and parental influences, developing new animal models and techniques, and assessing and implementing these designs and models in human health and development.

Combining motion analysis and microfluidics--a novel approach for detecting whole-animal responses to test substances

Small, early life stages, such as zebrafish embryos are increasingly used to assess the biological effects of chemical compounds in vivo. However, behavioural screens of such organisms are challenging in terms of both data collection (culture techniques, drug delivery and imaging) and data evaluation (very large data sets), restricting the use of high throughput systems compared to in vitro assays. Here, we combine the use of a microfluidic flow-through culture system, or BioWell plate, with a novel motion analysis technique, (sparse optic flow - SOF) followed by spectral analysis (discrete Fourier transformation - DFT), as a first step towards automating data extraction and analysis for such screenings. Replicate zebrafish embryos housed in a BioWell plate within a custom-built imaging system were subject to a chemical exposure (1.5% ethanol). Embryo movement was videoed before (30 min), during (60 min) and after (60 min) exposure and SOF was then used to extract data on movement (angles of rotation and angular changes to the centre of mass of embryos). DFT was subsequently used to quantify the movement patterns exhibited during these periods and Multidimensional Scaling and ANOSIM were used to test for differences. Motion analysis revealed that zebrafish had significantly altered movements during both the second half of the alcohol exposure period and also the second half of the recovery period compared to their pre-treatment movements. Manual quantification of tail flicking revealed the same differences between exposure-periods as detected using the automated approach. However, the automated approach also incorporates other movements visible in the organism such as blood flow and heart beat, and has greater power to discern environmentally-driven changes in the behaviour and physiology of organisms. We suggest that combining these technologies could provide a highly efficient, high throughput assay, for assessing whole embryo responses to various drugs and chemicals.