1 Introduction

1.1 Overview

Traumatic brain injury (TBI), a major cause of death and permanent disability, has been recognized as a serious public health concern in recent decades. In the USA, approximately 1.7 million individuals are diagnosed with TBI annually [1]. These types of injuries are frequently associated with falls, contact sports, and motor vehicle accidents among civilians. Blast-induced TBI (bTBI) has emerged as the signature injury of service members who participated in the conflicts in Iraq and Afghanistan, where studies suggest over 50% of TBIs are associated with explosions [2]. Since 2000, TBI has been diagnosed in more than 350,000 service members [3].

TBI causes varying degrees of neuronal deterioration and dysfunction, often leading to attenuation of cognitive and behavioral function. Despite significant investment in TBI research over the past two decades, the molecular and cellular etiology underlying TBI remains elusive. Studies of TBI using both experimental models and clinical data have faced significant challenges in scalability and reproducibility in humans, reflecting the complex nature of this disease. To date, no reliable biomarkers for TBI diagnosis and prognosis have been identified and every TBI drug tested in late-stage clinical trials has failed [4].

While cadaveric and surrogate systems that use instrumented anthropomorphic representations of humans provide good mechanical responses to blast dose, ultimate measures of acute biological injury response are difficult to identify at a molecular level. Existing in vitro biological models frequently employ a singular input for tension, shear, or pressure and induce loading directly onto ex vivo tissue or cell culture samples. While the assessments performed can provide a glimpse of a graded biological response to varying inputs, these models fail to accurately recreate a biofidelic model of trauma for blast injury that accounts individually for mechanical input interface, scale, and boundary conditions [5]. Additionally, these in vitro systems may allow examination of single cell responses but are unable to reproduce larger-scale injury cascades that occur within in vivo systems, including multicellular responses that result in longer-term alterations in genomic and protein expression. A holistic approach that considers multiple aspects of the physiological response of more complete biological systems in conjunction with graded biomechanical input–output relationships is needed to improve understanding of injury mechanisms and thresholds. Further, a greater understanding of the molecular events underlying the acute and chronic responses to blast will facilitate development of diagnostic and prophylactic treatments to ultimately mitigate the severity of bTBI.

1.2 An alternative paradigm for traumatic brain injury studies

A number of in vivo and in vitro models have been developed to help elucidate the molecular and cellular mechanisms that underlie the pathophysiological changes associated with bTBI (reviewed in [4]). Rodents are the most commonly used in vivo models of bTBI due to their small size, standardized outcome measures, and relatively low cost; however, results obtained in these models often do not scale to humans. To circumvent the issue of scalability, larger animals, such as pigs, are sometimes used. Using large animals is very expensive and, unlike for rodents, well-established functional tests of physiological and behavioral outcomes do not exist. Further, in vivo studies involve relatively small sample sizes and are not amenable to high-throughput screening. In vitro models have been developed as a low-cost high-throughput alternative to in vivo models. In vitro models include explants of central nervous system (CNS) tissue, organotypic cultures, dissociated primary cells, and immortalized cell lines (reviewed in [6]). These models maintain varying degrees of normal cellular complexity and have proved somewhat useful for ultrastructural, molecular, and biochemical studies; however, cells and tissues isolated in culture may differ in cell–cell interactions, cell signaling, and gene expression when compared to in vivo models. To date, a high-throughput animal model of bTBI with highly standardized injury exposures has not been developed.

The roundworm Caenorhabditis elegans has been used in biomedical research to model a number of human diseases, including several neurodegenerative disorders [7]. C. elegans is useful as a model of human disease due to the high degree of conservation between the human and C. elegans genomes; the majority of human disease genes and disease pathways are present in C. elegans [8]. In addition to this high degree of genomic conservation, C. elegans can be easily cultivated in the laboratory and has a short generation time, and a single self-fertilizing hermaphrodite can produce approximately 300 progeny. These attributes make C. elegans amenable to high-throughput experiments at low cost [9]. Despite its apparent simplicity, C. elegans is sufficiently complex to be a valuable model to study bTBI, having a highly developed and well-characterized nervous system that consists of 302 neurons and 50 glial cells, with the location and connectivity of each of these cells known [10].

Although tremendous efforts have been performed to advance the state of the science in human surrogates and finite element models, a gap remains in understanding the link between these systems and the thresholds at which acute injury and chronic responses occur. Additionally, current personal protective equipment (PPE) is primarily designed to protect the warfighter against penetrating and blunt injuries, but their efficacy in preventing acute and delayed neurological deficits (i.e., bTBI) from exposure to blast remains unknown. There is a significant need for biological data which link mechanical correlates, such as blast dose, to neurological injury outcomes using a device that can both represent a human and evaluate the efficacy of PPE for bTBI mitigation [4].

Herein, we describe a novel model of bTBI that incorporates a relevant biological model system embedded in an instrumented head surrogate that approximates biophysical features of the human brain, skull, and neck. This biologically integrated surrogate head model is comprised of materials that have physical properties representative of skin, bone, and brain tissue. Combined with the ability to insert and recover living material, this allows characterization of biomolecular, functional, and genomic responses to varying blast overpressure levels and links biomechanical sensor information with observed biological outcomes. Additionally, the flexibility of this model permits high-throughput testing in ways that traditional in vitro and in vivo models of blast exposure do not. These data may enable the development of transfer functions relating tissue-level biomechanical response measures (e.g., intracranial pressure, skull and brain kinematics [11]) with molecular and cellular responses, and by extension, risk of neurological injury. Such a combined model approach has potential to address significant gaps in understanding of bTBI and aid the evaluation and development of corresponding injury mitigation strategies and diagnostic measures.

Fig. 1
figure 1

Development of a biofidelic headform surrogate. Using a previously established headform design, we designed a new model with pressure sensors near the glabella (forehead midpoint), as well as the interior of the PDMS brain simulant (inset panel). Exposure to blast overpressure produced results similar to a previously demonstrated, highly sensorized version, with results for surface pressure in (a) and intracranial pressure in (b) at various driver pressures

Full size image

2 Methods

2.1 Biofidelic headform surrogate

An advanced anthropomorphic human head surrogate was developed to enable the evaluation of biological responses of integrated living systems to complex loading scenarios, including overpressure, bulk tissue movement, and tissue strain. The design of the surrogate headform was based on a previous biofidelic and highly sensorized model, the Johns Hopkins Applied Physics Laboratory’s Human Surrogate Head Model (HSHM), that was developed for blast injury testing [12,13,14,15]. Keeping with the previous surrogate design, biosimulant materials were incorporated into the headform to ensure that mechanical loading and material movement would better emulate the responses seen in a purely biological model. The biosimulant materials used with these molds included a glass/epoxy mixture (skull) [16], Sylgard silicone gel (brain), and polycarbonate (facial structure). The headform was designed and produced at JHU/APL using both additive manufacturing (3D printing) for the complex skull, facial, and brain structures, with traditional fabrication approaches used to apply the external layer representative of skin and other connective tissue.

This anthropomorphic headform design permits the assessment of a number of commonly tested loading conditions, including blast exposures from multiple directions and complex exposures that include combined blast and blunt trauma, as well as assessments with donned helmets and other head protective equipment. Biomechanical response has been characterized in blast-relevant loading environments using the precursor HSHM [12, 13, 17], and to enable comparison with data collected by this derivative headform, geometries and key sensor types and locations were maintained to be common. For each headform, a flat-profile surface pressure transducer is located at the glabella (forehead midpoint, Endevco part number 8515CM35-502, Fig. 1, inset), and an in-line intracranial pressure sensor was placed in the inferior region of the brain along the midsagittal line (TE Sensor Solutions, EPIH-300PS-X00003, Fig. 1b).

To permit the introduction of biological samples into the brain simulant material of the headform, the posterior and superior quarter half of the head was made detachable (Fig. 2a). The impact of the mechanical boundary created by this detachable component within the brain simulant material was minimized by the surface properties and bonding affinity of the incompletely cured poly-dimethylsiloxane (PDMS) [16] used as the matrix within the cranial cavity. The ability to introduce living biological samples within the simulated tissues of the surrogate enables blast-relevant mechanical forces to be imparted to the sample at the tissue and cell level. Additionally, this detachable design enables the rapid introduction and extraction of a variety of biological samples, including encapsulated cell cultures, organotypic tissue slices, and, in the case of this study, whole animals (Fig. 2b, c).

Fig. 2
figure 2

Insertion of biological materials into headform surrogate. CAD renderings of the biofidelic headform surrogate are shown in (a). The removable skull plate allows for insertion of tissue sample reservoirs as indicated in (a) and is additionally visible in (b). After insertion of biological material, the headform is secured using fasteners that create a continuous structure for both the brain simulant material and skull plate (c)

Full size image

2.2 Exposure of animals to blast overpressure

Blast overpressure was produced using a shock tube system with a 15-cm diameter containing a 94-cm driver section and 495-cm driven section. All data from the shock tube and headform were collected using a DEWE system (Dewetron, Austria) with a 1 MHz sampling rate and 16 bit depth, with all data analyzed subsequently in MATLAB. Headform experiments utilized a Hybrid III neck attached to a weighted sled, and all tests were conducted with the head positioned external to the shock tube exit with a standoff of 30-cm. While this test condition is not fully representative of a free-field blast exposure in that it results in a greater pressure impulse being imparted to the head, it does maintain key potential mechanisms of blast injury including pressure propagation through the brain followed by overall motion of the headform. Animals exposed to blast overpressure outside of the headform (referred to throughout as “direct” exposure) were inserted into the shock tube in line with the terminal pressure sensor. Physical limitations prohibited identical loading locations for petri dishes and the headform in our system.

Synchronized C. elegans egg preparations were prepared 36 h before exposure to blast overpressure. Prior to exposure, animals were verified by microscopy to be in the L4/early adult stage of their life cycle. All animals were transferred by picking to either new 6-cm culture plates for the direct exposure model or into a Sylgard well suitable for insertion into the headform. After exposure to sham or blast conditions, animals were either collected by cold centrifugation in M9 buffer and frozen on dry ice (\(t=0\) h time point) or placed back onto seeded 6-cm culture plates until collection and freezing (\(t=6\) h time point). For each condition, 20–50 animals were tested in unison. Harvested animals were stored at −80 \(^{\circ }\)C until processing for transcriptional profiling.

2.3 Sample preparation for transcriptomic profiling

Frozen animals were thawed on ice prior to disruption with a handheld microfuge homogenizer. RNA was isolated from resulting extracts using Agencourt RNAdvance (Beckman Genomics; A36245), with total RNA quantified using the Qubit RNA HS reagent (Life Technologies; Q32852). Following purification, 100 ng of RNA was input into the NEBNext Poly(A) mRNA Magnetic Isolation Module (New England Biolabs; E7490). Enriched mRNA was then used to generate sequencing libraries using NEBNext Ultra RNA library preparation reagents according to manufacturer’s instructions (New England Biolabs; E7530) and was multiplexed using NEBNext Multiplex Oligos for Illumina (New England Biolabs; E7335). Resulting sequencing libraries were quantified using high sensitivity picogreen reagents (Life Technologies; Q32854), pooled by combining equal mass libraries, and qualified using a bioanalyzer D1000 chip (Agilent; 5067-1504). The final library generated a 34.1 nM pool with an average insert size of 369 base pairs.

2.4 Sequencing data production and analysis

Transcriptomic sequencing data were generated using the Illumina HiSeq platform with a 50-base pair read length. A minimum of 39 million reads was produced for each sample. Raw reads were first filtered using Trim Galore! [18] using an adapter sequence of CTGTCTCTTATACACATCT and a quality score cutoff of 30. Quality-trimmed reads were then aligned to the C. elegans reference genome (PRJNA13758, WS247 build) using TopHat version 2.0.14, which used Bowtie version 2.2.6.0:

  • tophat2 --no-novel-juncs --no-coverage-search --output-dir <out-path> --GTF <WS247.gtf> <WS247.bwt>

    <sample.fastq>‘

as well as a GFF3 annotation file (PRJNA13758, WS247 build) obtained from WormBase. Aligned reads were then assembled into transcripts using Cufflinks version 2.1.1:

  • cufflinks --library-type fr-firststrand --no-effective-length-correction -b <WS247.fasta> -G <WS247.gff3>

    -o <out-path> <tophat2-accepted_hits.sam>‘

and those transcripts were quantified using Cuffquant version 2.2.1:

  • cuffquant --output-dir <out-path> --frag-bias-correct <WS247.fasta> <WS247.gtf>

    <tophat2-accepted_hits.bam>‘

These quantified transcripts were then combined into a normalized expression table using Cuffnorm version 2.2.1:

  • cuffnorm --output-dir <out-path> --total-hits-norm <WS247.gtf> <abundances1.cxb>,<abundances2.cxb>,

which was then subsequently processed through the “topGO” and “genefilter” packages in R after conversion to an ExpressionSet datatype. Cuffquant-quantified transcripts were also compared using Cuffdiff version 2.2.1:

  • cuffdiff --output-dir <out-path> <WS247.gtf> <tophat2-accepted_hits1.bam> <tophat2-accepted_hits2.bam>

This was performed for both blast–sham pairs (tube and headform), which were then analyzed using the “cummeRbund” R package. Significance testing results in “yes” or “no,” depending on whether p is greater then the FDR after Benjamini–Hochberg correction for multiple-testing.

3 Results

3.1 In vivo detection of acute phase transcriptional alterations after exposure to blast overpressure

Molecular events induced by blast injury are challenging to capture experimentally. To better understand these acute phase cellular responses, we used two experimental paradigms that incorporated age-matched C. elegans exposed to 36.6 ± 1.1 psi dynamic overpressure generated by a shock tube (Fig. 3a). In addition to the headform model developed for this study, we exposed worms directly to dynamic overpressure without protection from end jet effects or other environmental variables for comparison. After exposure, animals were collected both immediately (\(t=0\)) and after 6 h (\(t=6\)) to analyze changes in messenger RNA expression profiles; we chose these time points to establish a baseline condition as well as a stable cellular state following exposure.

Fig. 3
figure 3

Measurements for direct and headform blast overpressure exposures. Blast overpressure was produced using a 15-cm shock tube (a) with samples inserted both directly in the tube with no protection (blue rectangle) and just outside the tube encased in the biofidelic headform. Representative pressure wave propagation measurements for a single exposure are displayed in (b), with an average maximum overpressure of \(36.6\,\pm \,1.1\) psi observed at the terminal pressure sensor (green arrow) for all experiments in this study

Full size image
Table 1 Differentially expressed transcripts observed after exposure to blast overpressure
Full size table

When compared to matched sham controls, we observed a total of 205 significant gene expression changes over both conditions (Fig. 4a, Table 1), including 51 common to both (Fig. 4b). Direct exposure resulted in 122 significant transcriptional alterations that were not observed in the headform. Similarly, blast exposure within the headform resulted in 32 significant transcriptional alterations that were not observed in the direct exposure model. Two major gene classes were altered by the blast overpressure insult, regardless of direct or headform loading. These classes can be broadly categorized as transcription factors and genes involved in protein degradation based on gene ontology analysis. All conditions identified uncharacterized gene products that do not have gene ontology information available.

Fig. 4
figure 4

Transcriptional alterations in C. elegans after exposure to blast overpressure. Animals were surveyed for global transcriptional alterations after exposure to blast overpressure at \(t=0\) h and \(t=6\) h. Significant differential expression was observed for a total of 205 transcripts, with 51 overlapping in both direct and headform exposure scenarios (a). Differential expression for the 51 transcripts observed in both conditions observed in (b)

Full size image

Many of the differentially expressed transcripts observed only after direct blast exposure were F-box proteins, collagens, and other structural proteins. These results suggest increased physical trauma at the cellular level despite no visible changes in gross morphology or behavior. Conversely, transcriptional alterations observed only after blast exposure in the headform include G protein-coupled receptors and genes expressed primarily in neurons, as well as a number of cytochrome P450-related gene products. These gene families suggest potential cellular damage or inflammatory response induced after blast in the headform.

4 Discussion and conclusions

Our experimental system serves as a starting point to understand acute phase biological responses to blast injury. To our knowledge, this is the first study to characterize transcriptional regulation in the nematode C. elegans after exposure to blast overpressure. Additionally, our biofidelic headform is designed to approximate mechanical conditions that would occur within the human brain, such as shear strain and global head movement, that are not accurately captured by most existing in vitro or in vivo models. This capability allows assessment of the molecular and cellular sequelae induced by blast exposure, leading to a more comprehensive assessment of the physiological response and potentially providing mechanisms by which injury severity can be accurately and reproducibly gauged.

Transcriptional alterations of genes involved in protein degradation and endoplasmic reticulum (ER) stress response, such as proteasome or unfolded protein response (UPR) components, are compelling given the links between these processes to TBI and neurodegeneration [19,20,21]. The UPR has recently been shown to play a role in recovery after peripheral nerve injury in vivo [22] and altering the ER stress response has been shown to improve recovery from traumatic brain injury [23,24,25]. While many studies focusing on TBI therapeutics have utilized rodent models, the use of C. elegans provides the ability to rapidly assess different blast insult intensities, expose animals to potential therapeutics, and perform global transcriptional profiling with modest experimental costs given the amount and quality of data generated.

Comparison of a direct exposure event to that of our biofidelic headform generated a significant number of non-overlapping transcriptional regulation events. Due to differences in blast exposure, alterations of boundary conditions, and animal handling during transfer into and out of the headform, any link between these transcriptional changes requires further investigation to definitively attribute to features unique to experimental conditions introduced by our system. Determining the biological mechanism that causes these differences will require additional experiments in which variables, such as blast wind and shear stress, are tightly controlled. In addition, future efforts will further characterize the transcriptional alterations observed in the C. elegans model system and identify conserved mechanisms in cultured mammalian cells. Due to the suitability of the biofidelic headform for pairing with in vitro models, we anticipate that global transcriptional profiling, enabled by high-throughput sequencing technologies, can rapidly identify genes involved in response to blast injury and potentially underlying bTBI progression. Results of these studies can be extended to other model organisms as well as human tissue, which is only available for testing postmortem. Ultimately, these findings have potential to identify new biomarkers that cannot be easily identified using current model systems or clinical intervention strategies.