Feeds:
Posts
Comments

Archive for the ‘Genome Biology’ Category


Deep Learning extracts Histopathological Patterns and accurately discriminates 28 Cancer and 14 Normal Tissue Types: Pan-cancer Computational Histopathology Analysis

Reporter: Aviva Lev-Ari, PhD, RN

Pan-cancer computational histopathology reveals mutations, tumor composition and prognosis

Yu Fu1, Alexander W Jung1, Ramon Viñas Torne1, Santiago Gonzalez1,2, Harald Vöhringer1, Mercedes Jimenez-Linan3, Luiza Moore3,4, and Moritz Gerstung#1,5 # to whom correspondence should be addressed 1) European Molecular Biology Laboratory, European Bioinformatics Institute (EMBL-EBI), Hinxton, UK. 2) Current affiliation: Institute for Research in Biomedicine (IRB Barcelona), Parc Científic de Barcelona, Barcelona, Spain. 3) Department of Pathology, Addenbrooke’s Hospital, Cambridge, UK. 4) Wellcome Sanger Institute, Hinxton, UK 5) European Molecular Biology Laboratory, Genome Biology Unit, Heidelberg, Germany.

Correspondence:

Dr Moritz Gerstung European Molecular Biology Laboratory European Bioinformatics Institute (EMBL-EBI) Hinxton, CB10 1SA UK. Tel: +44 (0) 1223 494636 E-mail: moritz.gerstung@ebi.ac.uk

Abstract

Pan-cancer computational histopathology reveals mutations, tumor composition and prognosis

Here we use deep transfer learning to quantify histopathological patterns across 17,396 H&E stained histopathology image slides from 28 cancer types and correlate these with underlying genomic and transcriptomic data. Pan-cancer computational histopathology (PC-CHiP) classifies the tissue origin across organ sites and provides highly accurate, spatially resolved tumor and normal distinction within a given slide. The learned computational histopathological features correlate with a large range of recurrent genetic aberrations, including whole genome duplications (WGDs), arm-level copy number gains and losses, focal amplifications and deletions as well as driver gene mutations within a range of cancer types. WGDs can be predicted in 25/27 cancer types (mean AUC=0.79) including those that were not part of model training. Similarly, we observe associations with 25% of mRNA transcript levels, which enables to learn and localise histopathological patterns of molecularly defined cell types on each slide. Lastly, we find that computational histopathology provides prognostic information augmenting histopathological subtyping and grading in the majority of cancers assessed, which pinpoints prognostically relevant areas such as necrosis or infiltrating lymphocytes on each tumour section. Taken together, these findings highlight the large potential of PC-CHiP to discover new molecular and prognostic associations, which can augment diagnostic workflows and lay out a rationale for integrating molecular and histopathological data.

SOURCE

https://www.biorxiv.org/content/10.1101/813543v1

Key points

● Pan-cancer computational histopathology analysis with deep learning extracts histopathological patterns and accurately discriminates 28 cancer and 14 normal tissue types

● Computational histopathology predicts whole genome duplications, focal amplifications and deletions, as well as driver gene mutations

● Wide-spread correlations with gene expression indicative of immune infiltration and proliferation

● Prognostic information augments conventional grading and histopathology subtyping in the majority of cancers

 

Discussion

Here we presented PC-CHiP, a pan-cancer transfer learning approach to extract computational histopathological features across 42 cancer and normal tissue types and their genomic, molecular and prognostic associations. Histopathological features, originally derived to classify different tissues, contained rich histologic and morphological signals predictive of a range of genomic and transcriptomic changes as well as survival. This shows that computer vision not only has the capacity to highly accurately reproduce predefined tissue labels, but also that this quantifies diverse histological patterns, which are predictive of a broad range of genomic and molecular traits, which were not part of the original training task. As the predictions are exclusively based on standard H&E-stained tissue sections, our analysis highlights the high potential of computational histopathology to digitally augment existing histopathological workflows. The strongest genomic associations were found for whole genome duplications, which can in part be explained by nuclear enlargement and increased nuclear intensities, but seemingly also stems from tumour grade and other histomorphological patterns contained in the high-dimensional computational histopathological features. Further, we observed associations with a range of chromosomal gains and losses, focal deletions and amplifications as well as driver gene mutations across a number of cancer types. These data demonstrate that genomic alterations change the morphology of cancer cells, as in the case of WGD, but possibly also that certain aberrations preferentially occur in distinct cell types, reflected by the tumor histology. Whatever is the cause or consequence in this equation, these associations lay out a route towards genomically defined histopathology subtypes, which will enhance and refine conventional assessment. Further, a broad range of transcriptomic correlations was observed reflecting both immune cell infiltration and cell proliferation that leads to higher tumor densities. These examples illustrated the remarkable property that machine learning does not only establish novel molecular associations from pre-computed histopathological feature sets but also allows the localisation of these traits within a larger image. While this exemplifies the power of a large scale data analysis to detect and localise recurrent patterns, it is probably not superior to spatially annotated training data. Yet such data can, by definition, only be generated for associations which are known beforehand. This appears straightforward, albeit laborious, for existing histopathology classifications, but more challenging for molecular readouts. Yet novel spatial transcriptomic44,45 and sequencing technologies46 bring within reach spatially matched molecular and histopathological data, which would serve as a gold standard in combining imaging and molecular patterns. Across cancer types, computational histopathological features showed a good level of prognostic relevance, substantially improving prognostic accuracy over conventional grading and histopathological subtyping in the majority of cancers. It is this very remarkable that such predictive It is made available under a CC-BY-NC 4.0 International license. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. bioRxiv preprint first posted online Oct. 25, 2019; doi: http://dx.doi.org/10.1101/813543. The copyright holder for this preprint signals can be learned in a fully automated fashion. Still, at least at the current resolution, the improvement over a full molecular and clinical workup was relatively small. This might be a consequence of the far-ranging relations between histopathology and molecular phenotypes described here, implying that histopathology is a reflection of the underlying molecular alterations rather than an independent trait. Yet it probably also highlights the challenges of unambiguously quantifying histopathological signals in – and combining signals from – individual areas, which requires very large training datasets for each tumour entity. From a methodological point of view, the prediction of molecular traits can clearly be improved. In this analysis, we adopted – for the reason of simplicity and to avoid overfitting – a transfer learning approach in which an existing deep convolutional neural network, developed for classification of everyday objects, was fine tuned to predict cancer and normal tissue types. The implicit imaging feature representation was then used to predict molecular traits and outcomes. Instead of employing this two-step procedure, which risks missing patterns irrelevant for the initial classification task, one might directly employ either training on the molecular trait of interest, or ideally multi-objective learning. Further improvement may also be related to the choice of the CNN architecture. Everyday images have no defined scale due to a variable z-dimension; therefore, the algorithms need to be able to detect the same object at different sizes. This clearly is not the case for histopathology slides, in which one pixel corresponds to a defined physical size at a given magnification. Therefore, possibly less complex CNN architectures may be sufficient for quantitative histopathology analyses, and also show better generalisation. Here, in our proof-of-concept analysis, we observed a considerable dependence of the feature representation on known and possibly unknown properties of our training data, including the image compression algorithm and its parameters. Some of these issues could be overcome by amending and retraining the network to isolate the effect of confounding factors and additional data augmentation. Still, given the flexibility of deep learning algorithms and the associated risk of overfitting, one should generally be cautious about the generalisation properties and critically assess whether a new image is appropriately represented. Looking forward, our analyses revealed the enormous potential of using computer vision alongside molecular profiling. While the eye of a trained human may still constitute the gold standard for recognising clinically relevant histopathological patterns, computers have the capacity to augment this process by sifting through millions of images to retrieve similar patterns and establish associations with known and novel traits. As our analysis showed this helps to detect histopathology patterns associated with a range of genomic alterations, transcriptional signatures and prognosis – and highlight areas indicative of these traits on each given slide. It is therefore not too difficult to foresee how this may be utilised in a computationally augmented histopathology workflow enabling more precise and faster diagnosis and prognosis. Further, the ability to quantify a rich set of histopathology patterns lays out a path to define integrated histopathology and molecular cancer subtypes, as recently demonstrated for colorectal cancers47 .

Lastly, our analyses provide It is made available under a CC-BY-NC 4.0 International license. (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.

bioRxiv preprint first posted online Oct. 25, 2019; doi: http://dx.doi.org/10.1101/813543.

The copyright holder for this preprint proof-of-concept for these principles and we expect them to be greatly refined in the future based on larger training corpora and further algorithmic refinements.

SOURCE

https://www.biorxiv.org/content/biorxiv/early/2019/10/25/813543.full.pdf

 

Other related articles published in this Open Access Online Scientific Journal include the following: 

 

CancerBase.org – The Global HUB for Diagnoses, Genomes, Pathology Images: A Real-time Diagnosis and Therapy Mapping Service for Cancer Patients – Anonymized Medical Records accessible to anyone on Earth

Reporter: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2016/07/28/cancerbase-org-the-global-hub-for-diagnoses-genomes-pathology-images-a-real-time-diagnosis-and-therapy-mapping-service-for-cancer-patients-anonymized-medical-records-accessible-to/

 

631 articles had in their Title the keyword “Pathology”

https://pharmaceuticalintelligence.com/?s=Pathology

 

Read Full Post »


NSPR1 and DEC2 genes: Survival on 4.5 hours of Sleep per night: A mutation in the β1-adrenergic receptor gene in humans who require fewer hours of sleep than most, ADRB1 + neurons are active during rapid eye movement (REM) sleep and wakefulness

 

Reporter: Aviva Lev-Ari, PhD, RN

 

10/2019 RESEARCH ARTICLE SLEEP

Mutant neuropeptide S receptor reduces sleep duration with preserved memory consolidation

 See all authors and affiliations

Science Translational Medicine  16 Oct 2019:
Vol. 11, Issue 514, eaax2014
DOI: 10.1126/scitranslmed.aax2014

Abstract

Sleep is a crucial physiological process for our survival and cognitive performance, yet the factors controlling human sleep regulation remain poorly understood. Here, we identified a missense mutation in a G protein–coupled neuropeptide S receptor 1 (NPSR1) that is associated with a natural short sleep phenotype in humans. Mice carrying the homologous mutation exhibited less sleep time despite increased sleep pressure. These animals were also resistant to contextual memory deficits associated with sleep deprivation. In vivo, the mutant receptors showed increased sensitivity to neuropeptide S exogenous activation. These results suggest that the NPS/NPSR1 pathway might play a critical role in regulating human sleep duration and in the link between sleep homeostasis and memory consolidation.

It is possible that drugs could be developed to target either the NSPR1 or DEC2 genes, as a treatment for insomnia or other sleep disorders. However, further understanding of exactly how these genes function would be required before this stage. Both are involved in brain function, so targeting them could lead to negative neural side effects.

 

Neuron

Volume 103, Issue 6, 25 September 2019, Pages 1044-1055.e7

A Rare Mutation of β1-Adrenergic Receptor Affects Sleep/Wake Behaviors

Highlights

  • A mutation in ADRB1 leads to natural short sleep trait in humans
  • Mice engineered with same mutation have similar short sleep behavior as humans
  • Activity of dorsal pons ADRB1 + neurons associates with REM sleep and wakefulness
  • Mutation increases the population activity of dorsal pons ADRB1 + neurons

Summary

Sleep is crucial for our survival, and many diseases are linked to long-term poor sleep quality. Before we can use sleep to enhance our health and performance and alleviate diseases associated with poor sleep, a greater understanding of sleep regulation is necessary. We have identified a mutation in the β 1-adrenergic receptor gene in humans who require fewer hours of sleep than most. In vitro, this mutation leads to decreased protein stability and dampened signaling in response to agonist treatment. In vivo, the mice carrying the same mutation demonstrated short sleep behavior. We found that this receptor is highly expressed in the dorsal pons and that these ADRB1 + neurons are active during rapid eye movement (REM) sleep and wakefulness. Activating these neurons can lead to wakefulness, and the activity of these neurons is affected by the mutation. These results highlight the important role of β 1-adrenergic receptors in sleep/wake regulation.

Keywords

Additional SOURCES

Second Gene Mutation that Lets People Survive on Less Sleep

 

Other related articles on Circadian Rhythm and Sleep published in this Open Access Online Scientific Journal include the following:

 

2017 Nobel Prize in Physiology or Medicine jointly to Jeffrey C. Hall (ex-Brandeis, University of Maine), Michael Rosbash (Brandeis University) and Michael W. Young (Rockefeller University in New York) for their discoveries of molecular mechanisms controlling the circadian rhythm

Curator: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2017/10/02/2017-nobel-prize-in-physiology-or-medicine-jointly-to-jeffrey-c-hall-michael-rosbash-and-michael-w-young-for-their-discoveries-of-molecular-mechanisms-controlling-the-circadian-rhythm/

 

Patient-Reported Outcomes Study, Presented at SLEEP 2018, Provides Confirmatory Real-World Evidence of the Previously Presented 7-hour Action of REMfresh®, the First Continuous Release and Absorption Melatonin™

Reporter: Gail S. Thornton, PhD(c)

https://pharmaceuticalintelligence.com/2018/06/10/patient-reported-outcomes-study-presented-at-sleep-2018-provides-confirmatory-real-world-evidence-of-the-previously-presented-7-hour-action-of-remfresh-the-first-continuous-release-and-absorp/

 

Clinically Studied, Continuous Release and Absorption Melatonin, REMfresh, Designed to Give Patients Up to 7 Hours of Sleep Support

Reporter: Gail S. Thornton, M.A.

https://pharmaceuticalintelligence.com/2019/06/19/clinically-studied-continuous-release-and-absorption-melatonin-remfresh-designed-to-give-patients-up-to-7-hours-of-sleep-support/

 

2017 award recipients including Thomas S. Kilduff, PhD, Director, Center for Neuroscience at SRI International in Menlo Park, California

Reporter: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2017/04/28/sleep-research-society-announces-2017-award-recipients-including-thomas-s-kilduff-phd-director-center-for-neuroscience-at-sri-international-in-menlo-park-california/

 

Sleep and Memory

Curator: Larry H. Bernstein, MD, FCAP

https://pharmaceuticalintelligence.com/2016/03/26/sleep-and-memory/

 

Sleep Science

Curator: Larry H. Bernstein, MD, FCAP

https://pharmaceuticalintelligence.com/2016/03/16/sleep-science/

 

Genetic Link to Sleep and Mood Disorders

Curator: Larry H. Bernstein, MD, FCAP

https://pharmaceuticalintelligence.com/2016/02/27/genetic-link-to-sleep-and-mood-disorders/

 

Sleep Apnea Insular Glutamate and GABA Levels

Curator: Larry H. Bernstein, MD, FCAP

https://pharmaceuticalintelligence.com/2016/02/12/sleep-apnea-insular-glutamate-and-gaba-levels/

 

Fat, Sleep and the Gut

Curator: Larry H. Bernstein, MD, FCAP

https://pharmaceuticalintelligence.com/2016/02/06/fat-sleep-and-the-gut/

 

23andMe Genome-Wide Association Study on Human propensity to Get up early or Sleep in the Morning

Reporter: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2016/02/02/23andme-genome-wide-association-study-on-human-propensity-to-get-up-early-or-sleep-in-the-morning/

 

Sleep Quality, Amyloid and Cognitive Decline

Curator: Larry H. Bernstein, MD, FCAP

https://pharmaceuticalintelligence.com/2015/10/31/sleep-quality-amyloid-and-cognitive-decline/

 

Study Shows Learning Is Best Enhanced During Sleep – Jewish Business News

Reporter: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2014/09/02/study-shows-learning-is-best-enhanced-during-sleep-jewish-business-news/

 

Beta-Blockers Cause Lack Of Restful Sleep – Life Extension

Reporter: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2014/08/04/beta-blockers-cause-lack-of-restful-sleep-life-extension/

 

Topical Antispasmodics conducive for Uninterrupted Sleep – A Potential Cardiovascular Chrono-therapeutics

Reporter: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2014/02/13/topical-antispasmodics-conducive-for-uninterrupted-sleep-a-potential-cardiovascular-chrono-therapeutics/

 

Prolonged Wakefulness: Lack of Sufficient Duration of Sleep as a Risk Factor for Cardiovascular Diseases – Indications for Cardiovascular Chrono-therapeutics

Curator: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2014/02/02/prolonged-wakefulness-lack-of-sufficient-duration-of-sleep-as-a-risk-factor-for-cardiovascular-diseases-indications-for-cardiovascular-chrono-therapeutics/

 

Sleep Apnea and Non-invasive positive Pressure Breathing

Curator: Larry H. Bernstein, MD, FCAP

https://pharmaceuticalintelligence.com/2013/06/11/sleep-apnea-and-non-invasive-positive-pressure-breathing/

 

How Might Sleep Apnea Lead to Serious Health Concerns like Cardiac and Cancer?

Author: Larry H Bernstein, MD, FCAP

https://pharmaceuticalintelligence.com/2013/03/20/how-might-sleep-apnea-lead-to-serious-health-concerns-like-cardiac-and-cancers/

 

2019 Warren Alpert Foundation Award goes to Four Scientists for Seminal Discoveries in OptoGenetics – Illuminating the Human Brain

Reporter: Aviva Lev-Ari, PhD, RN

https://pharmaceuticalintelligence.com/2019/07/18/2019-warren-alpert-foundation-award-goes-to-four-scientists-for-seminal-discoveries-in-optogenetics-illuminating-the-human-brain/

 

 

 

 

 

Read Full Post »


Nano-guided cell networks as conveyors of molecular communication

Nature Communications
6,
Article number:
8500
doi:10.1038/ncomms9500
Received
07 March 2015
Accepted
28 August 2015
Published
12 October 2015

Abstract

Advances in nanotechnology have provided unprecedented physical means to sample molecular space. Living cells provide additional capability in that they identify molecules within complex environments and actuate function. We have merged cells with nanotechnology for an integrated molecular processing network. Here we show that an engineered cell consortium autonomously generates feedback to chemical cues. Moreover, abiotic components are readily assembled onto cells, enabling amplified and ‘binned’ responses. Specifically, engineered cell populations are triggered by a quorum sensing (QS) signal molecule, autoinducer-2, to express surface-displayed fusions consisting of a fluorescent marker and an affinity peptide. The latter provides means for attaching magnetic nanoparticles to fluorescently activated subpopulations for coalescence into colour-indexed output. The resultant nano-guided cell network assesses QS activity and conveys molecular information as a ‘bio-litmus’ in a manner read by simple optical means.

At a glance

Figures

View all figures

left

  1. Nano-guided cell networks for processing molecular information.
    Figure 1
  2. Cells express functional, interchangeable protein components indicating both fluorescence and ability for streptavidin-linked surface coupling.
    Figure 2
  3. Cells equipped with magnetic nanoparticles (mNPs) via streptavidin-mediated interaction with surface-expressed proteins.
    Figure 3
  4. Affinity-based probing for functional analysis of AI-2-induced protein expression.
    Figure 4
  5. Single and multi-population cell responses to autoinducer-2.
    Figure 5
  6. Binning molecular information through cell-based parallel processing and magnetically focusing fluorescence into collective consensus output.
    Figure 6
  7. Extension of nano-guided cell networks for hypothetical regulatory structures.
    Figure 7

right

Introduction

It has become increasingly apparent that a wealth of molecular information exists, which, when appropriately accessed, can provide feedback on biological systems, their componentry and their function. Thus, there is a developing niche that transcends length scales to concurrently recognize molecular detail and at the same time provide understanding of the overall system1, 2. An emerging scheme is to develop nano- to micro-scaled tools that intimately engage with biological systems through monitoring and interacting at the molecular level, with synthetic biology being one such tool3, 4, 5, 6, 7.

While synthetic biology is often viewed as an innovative means for ‘green’ product synthesis through the genetic rearrangement of cells, their biosynthetic capabilities and their regulatory networks can instead be tuned for executive function8, 9, 10. That is, cells can be rewired to survey molecular space3, 11, 12 as they have sophisticated capabilities to recognize, amplify and transduce chemical information13. Further, they provide a means to connect biological systems with traditional microelectronic devices and in doing so present a potential interface between chemically based biomolecular processing and conventional vectors of information flow, such as electrons and photons14, 15, 16. Specifically, through engineered design, cell-based molecular processing can be further coupled to enable external abiotic responses. Cells, then, represent a versatile means for mediating the molecular ‘signatures’ common in complex environments, or in other words, they are conveyors of molecular communication17, 18, 19.

Further, beyond clonal cell-based sensors, there is an emerging concept of population engineering to establish microorganisms in deliberate networks that enable enriched system identification through a combination of distinctive yet coexistent behaviours, including, perhaps, competitive or cooperative features8, 20, 21, 22, 23, 24, 25. We posit the use of cell populations assembled in parallel¸ where multiple microbes with distinct molecular recognition capabilities work congruently. An advantage is that populations, as opposed to few cells, can facilitate thorough sampling since the presence of many cells increases their spatial breadth and per-cell data contributions (Fig. 1a). Each cellular unit undergoes independent decision-making and contributes a datum to its entire constituency. The prevalence of data provided within the population, then, substantiates a collective output by the system based on the molecular landscape. As follows in a multi-population system, molecular input thus influences the outcomes of each population, and elicits plural responses when the molecular input ranges overlap the ranges of the sensing populations21, which can define classification boundaries (Fig. 1a). Cell-mediated classification was posited in silico by Didovyk et al.21, where reporter libraries with randomized sensitivities to a molecular cue elicit concentration-dependent fluorescent patterns and these are elucidated by population screening . In the present construct, multiple populations enable multiplexed analysis, resulting, here, in a response gradation that is designed to index the molecular input ‘signature’. Consequently, the fed-back information becomes transfigured beyond a dose-dependent cell-by-cell analysis. That is, the output is predicated by the comparison between the populations rather than accumulation of response within a total population.

Figure 1: Nano-guided cell networks for processing molecular information.

Nano-guided cell networks for processing molecular information.

(a) Biotic (multicellular) processing is facilitated by cell recognition, signal transduction and genetic response. The genetically encoded response reflects the identity and prevalence of the target molecule(s). Biotic processing includes both increased cell number of responders and their genetically tuned response patterns. (b) Abiotic processing, used in conjunction with biotic processing, adds dimensionality to cell-based output by modifying through a physical stimulus (in our example, magnetic focusing). (c) Schematic of a cell population and nanomaterial-based network comprising both biotic (green/red axis) and abiotic (black axis) processing mechanisms. This conceptual system interprets molecular information by intercepting diverse molecular inputs, processes them autonomously through independent cell units within the system and refines output to include positive responders that are viewed via orthogonal means (visual classification). The system’s hierarchical structure allows molecular information to be refined into categorized collective outputs.

With population engineering as a premise for enriched molecular information processing, we engineered cell species, each to achieve an appropriate output through genetic means. There is conceptual basis for incorporation into networks, such as through mobile surveillance and position-based information relay26, 27. Hence, it is conceivable that, in addition to autonomous molecular recognition and processing afforded by synthetic biology, the use of physical stimuli to enable cell response could confer similar networking properties28, 29. For example, the complete information-processing ‘repertoire’ can be expanded beyond specific cell responses by the integration of external stimuli that serve to collate cell populations30. Specifically, we envision integration of nanomaterials that enable co-responses to molecular inputs, such that cell populations employ traditional reporting functions, that is, fluorescence marker expression, as well as responses that enable additional processing via the integration of stimuli-responsive abiotic materials (Fig. 1b).

In our example, cells are engineered to respond by permitting the attachment of magnetic nanoparticles (mNPs), such that each fluorescent cell becomes receptive to a magnetic field. Thus, the combination of cell-nanoparticle structures provides further dimensionality for the conveyance of molecular information (via magnetic stimulation). That is, without magnetic collation the fully distributed system would harbour diffuse responses; a magnetically stimulated system results in acute output due to a filtering and focusing effect (Fig. 1b)31, 32, allowing binned information to be readily, and fluorescently, conveyed.

The detection and interpretation of signalling molecules in our example is based on a microbial communication process known as quorum sensing (QS). The molecules, autoinducers (AIs), are secreted and perceived within a microbial community; once accumulated, the AI level indicates that the population size has reached a ‘quorum’33, 34. By surpassing a threshold concentration, the AI signalling coordinates population-wide phenotypic changes35. We have designed a QS information processor that utilizes two cell populations to independently interrogate natural microbial communities and generate information about QS activity by accessing AI-2 (ref. 36). Each cell population becomes ‘activated’ in response to a characteristic AI-2 level by expressing a fluorescent marker and a streptavidin-binding peptide (SBP) on the outer membrane38. SBP provides a means for collating data by binding mNPs that are introduced into the community. Using a post-processing magnetic sweep, the system as a whole interprets a molecular landscape and refines output into colour-categorized, or ‘binned,’ states (no fluorescence, red, or red and green) through (1) parallel population processing and (2) acute focusing (Fig. 1c).

The use of engineered cells as data-acquiring units and selectively equipping each with functional nanomaterials to form a redistributable processing system merges two paradigms: decentralized, active probing at a molecular scale and self-organization of units through structured dependencies on stimuli42. The population-based system overall contributes categorized feedback about a biological environment.

Results

Surface expression of SBP and fluorescent protein fusions

First, we established expression of a fusion protein consisting of a fluorescent marker (enhanced green fluorescent protein (eGFP) and variants) and SBP. Importantly, for SBP to function as a coupling agent between cells and mNPs, we used AIDAc (kindly shared by J. Larssen)40 to export the chimeric protein to Escherichia coli’s outer surface. Translocation to a cell’s surface utilizes a signal peptide (for inner membrane translocation) and AIDAc as an outer membrane autotransporter pore38, 39, 40, 41, with the passenger protein linked to each. In Fig. 2a, we depict expression of three different constructs using Venus, eGFP and mCherry for optical transmission, and the AIDAc translocator domain for surface localization. These constructs are mapped inSupplementary Fig. 1. After induction with isopropyl B-D-1-thiogalactopyranoside (IPTG), cultures were probed for surface expression of the SBP portion of the tagged fluorescent protein. Cells were incubated with fluorescently labelled streptavidin; the fluorophore of the streptavidin probe was orthogonal to the expressed fluorescent protein. The multiple fluorescence emissions were analysed by confocal microscopy without spectral overlap. The fraction of cells (fc) that exhibit colocalized fluorescent protein and the fluorescently-labeled streptavidin is reported in Fig. 2b, showing that SBP–Venus cells bound streptavidin at a slightly lower frequency than SBP–mCherry and SBP–eGFP, which exhibited statistically similar fractions (fc=0.7).

Figure 2: Cells express functional, interchangeable protein components indicating both fluorescence and ability for streptavidin-linked surface coupling.

Cells express functional, interchangeable protein components indicating both fluorescence and ability for streptavidin-linked surface coupling.

(a) A T7 cassette was used to express chimeric proteins consisting of a membrane autotransporter domain (AIDAc), one of several fluorescent proteins and a streptavidin-binding peptide (SBP). Fluorophore-tagged streptavidin (SA) was used to bind SBP. (b) Of cells expressing fluorescent proteins (FP), those also marked by SBP coupling are represented as a ‘colocalized fraction (fc),’ plotted with image analysis-based s.d. of at least five replicates. The asterisk ‘*’ denotes fc that +SBP–eGFP and+SBP–mCherry are statistically equivalent (fc~0.7) by t-test and greater than +SBP–Venus cells. (c) Composite images show cell fluorescence (Column I) from the fluorescent protein (FP); labelled streptavidin using orthogonal filter sets (Column II); and an overlay of both (Column III). Arrows indicate representative cells with strong colocalization. Plotted in Column IV are the fluorescence mean grey values (y-axis) from a representative horizontal slice of the composite image (x-axis). Vertical bars displayed between Columns III and IV identify the position of each analysed slice. Arrows indicate peaks that match the highlighted cells in Column III. fc values are noted. Fluorophores with non-overlapping spectra were paired. Row 1, Venus expression (yellow-green) was paired with Dylight405-labelled SA (blue). Row 2, eGFP expression (green) was paired with Alexafluor594-labeled SA (red). Row 3, mCherry expression (red) was paired with Alexafluor488-labeled SA (green). Scale bar in lower left, 50μm.

That is, microscopy results related to the colocalization analysis are depicted for pairings of Venus and blue-streptavidin (SA), eGFP and red-SA, and mCherry and green-SA (Fig. 2c). Strong signals were observed in both filter sets (the fluorescent protein (Column I) and the labelled streptavidin (Column II)). Overlaying each image reveals colocalization, as indicated in Column III, where arrows point to examples of strong colocalization. In addition, Column IV plots fluorescence intensities across horizontal sections of the images, where cells that exhibit colocalized fluorescence are indicated by superimposed peaks. For +pSBP–Venus cells, those with both a blue and yellow signal are observed as pale blue–violet in the overlaid image. Cells with +pSBP–eGFP and +pSBP–mCherry and labelled streptavidin emit both green and red signals; their colocalization appears yellow. Controls shown in Supplementary Fig. 2, verify that fluorescent streptavidin (all colours) has specificity for only SBP-expressing cells over negative controls. Colocalization indicates that not only are both components of the fusion, SBP and the fluorescent protein, expressed, but that SBP is accessible to bind streptavidin on the cell’s surface. This is the first use of AIDAc for cell surface anchoring of fluorescent proteins, each having been functionalized with an affinity peptide.

Cell hybridization via mNPs

Given that expression of a fluorescent protein tagged with SBP enabled external binding of streptavidin, we employed this interaction for fastening streptavidin-functionalized materials directly to the cell surface. We chose streptavidin-conjugated mNPs, 100nm in diameter (an order of magnitude smaller than a cell), for binding to a cell surface (Fig. 3a) to impart the abiotic magnetic properties. Scanning electron microscopy (SEM) was used to observe surface interaction between cell surface-expressing SBP and streptavidin-functionalized mNPs. Supplementary Fig. 3a,bshows electron micrographs of E. coli cells (dimensions 1.5–2μm in length) and the mNPs (~100nm in diameter). The SEM image in Fig. 3b, shows a magnetically isolated SBP-expressing cell with streptavidin-mNPs. The sample was prepared by mixing SBP-expressing cells with streptavidin-mNPs, then collecting or ‘focusing’ into a magnetized pellet via magnetic field, then separating from unbound cells in the supernatant. The cells were then washed and resuspended. In Fig. 3b, clusters of surface-bound mNPs are observed. In addition, the elemental composition was analysed with energy-dispersive X-ray spectroscopy, shown in Fig. 3c by an element map superimposed with carbon (red) and iron (green). While the cell appears to be of a uniform carbon composition, the particles localized at the cell surface (highlighted with arrows) were found having a strong iron composition; thus, elemental analysis confirmed particle identity as iron oxide mNPs. Additional characterization of magnetic functionality, including detailed SEM and fluorescent microscopic analysis prior to and after application of magnetic fields, is described in theSupplementary Information (Supplementary Fig. 3).

Figure 3: Cells equipped with magnetic nanoparticles (mNPs) via streptavidin-mediated interaction with surface-expressed proteins.

Cells equipped with magnetic nanoparticles (mNPs) via streptavidin-mediated interaction with surface-expressed proteins.

(a) Cell surface binding of streptavidin-conjugated magnetic nanoparticles occurs via surface-anchored streptavidin-binding peptide (SBP). The fusion of T7-expressed SBP-fluorescent protein (FP)-AIDAc enables the cell surface accessibility. (b) Scanning electron micrograph of an E. coli cell with surface-bound particles. (c) Element map of carbon (red) and iron (green) through energy-dispersive spectroscopy.

In sum, the well-known affinity interaction between streptavidin and the peptide SBP is harnessed to endow cells with non-natural abiotic properties. Here coupling a functionalized nanomaterial to the surface-displayed peptide physically extends the fusion protein and also adds physical (magnetic) functionality to the cell.

Linking expression to AI-2 recognition

The expression system for pSBP–Venus was then put under AI-2 control so that the protein is expressed in the presence of AI-2 instead of IPTG. That is, we coupled the native QS signal transduction circuitry to the reporter cassette. To ensure ample expression (as the native operon is fairly weak), we placed expression of T7 RNA polymerase under control of the natural QS circuitry43. Phosphorylated AI-2 activates the system through derepression of the regulator LsrR, naturally upregulating AI-2 import and phosphorylation44, and, by design, the T7 RNA polymerase on a sensor plasmid43. When sbp–Venus is included downstream of a T7 promoter region on a second plasmid, expression is then triggered by AI-2 uptake (Supplementary Fig. 4a). Then, we used two host sensor strains engineered to provide varied AI-2 sensitivity (denoted responders ‘A’ and ‘B’). In ‘A’, lsrFG, genes required for internally phosphorylated AI-2 degradation45, 46 are deleted. Also, both strains lack the terminal AI-2 synthase, luxS, so they cannot produce AI-2 and, instead, must ‘receive’ AI-2 from an external source (Supplementary Fig. 4a). The phenotypic difference between A and B is the threshold level of AI-2 that activates the genetic response47, 48. Fully constructed, these cells are designed to take up and process AI-2 to generate fluorescence output (that co-functions with streptavidin binding).

We next evaluated the kinetics of surface-fusion protein expression and effects on cell growth. The AI-2-induced expression for AIDAc-linked and SBP-tagged fluorescent proteins did not alter growth kinetics for either cell type (Supplementary Fig. 4b,c). Expression efficacy was also evaluated via immunoassay of the outer membrane, probing for AI-2-induced surface display. After induction with 20μM AI-2, extracts from cell types A and B were size-separated and blotted using alkaline phosphatase-conjugated streptavidin to probe for the SBP-tagged protein fusion (Supplementary Fig. 5). The 88kDa AIDAc–Venus–SBP protein was only found in the membrane-containing pellet fraction (Fig. 4a). Analogously, protein orientation was assessed by immunolabeling the fluorescent protein. Cell type B transformed with pSBP–eGFP was induced with 20μM AI-2 overnight; cell surfaces were then probed for eGFP using a mouse anti-GFP primary antibody and red-labelled secondary anti-mouse IgG. Simultaneously, cells were observed using phase contrast and fluorescence confocal microscopy. We noted a punctate pattern for eGFP, which was in one-to-one correspondence with red immunostaining of the surface-expressed protein. The positive staining of eGFP-expressing cells for red fluorescence, contrasted by the absence of negative control immunostaining indicated surface exposure of the fusion (Supplementary Fig. 6). Confocal microscopy confirmed precise colocalization of the eGFP and red-labelled antibodies within the confines of individual cells (Fig. 4b). Therefore, efficient transport of this functionality to the membrane under AI-2 induction was demonstrated in each host.

Figure 4: Affinity-based probing for functional analysis of AI-2-induced protein expression.

Affinity-based probing for functional analysis of AI-2-induced protein expression.

(a) 64–82kDa region of western blot for pelleted (P) and supernatant (S) protein fractions isolated from Type A and B cells. Alkaline phosphatase-conjugated streptavidin was used to target AIDAc–Venus–SBP at expression timepoints. Arrows indicate the expected position of the full fusion protein. (b) Immunostaining for assessment of the fluorescent protein surface accessibility. The external surfaces of cells expressing AIDAc–eGFP–SBP were probed with an anti–eGFP and Alexafluor594-labelled antibody pair. A representative overlaid fluorescence and phase contrast image is shown along with fluorescence images of the green (G) and red (R) filters for the boxed-in region. Scale bar, 2μM.

Establishing molecular ranges for cell interrogation

Importantly, the engineered cells each provide a characteristic response to the level of AI-2. Recently, we showed that AI-2 level influences the quorum size of responding engineered populations but does not alter the expression level within each quorum47. Here we evaluated our engineered AI-2 responders, again for quorum size (or in other words, percentage of AI-2-responsive cells in the population), this time varying the compositions of molecular input and the configuration of responders (Fig. 5a). First, we added AI-2, synthesized in vitro, to each of the two responder populations (Fig. 5b). We also added conditioned medium (CM), the spent medium from an AI-2 producer culture containing metabolic byproducts, as well as AI-2 (refs 36, 49; Fig. 5c). We also mixed the responder populations and added AI-2 to gauge responses in complex cultures (Fig. 5d).

Figure 5: Single and multi-population cell responses to autoinducer-2.

Single and multi-population cell responses to autoinducer-2.

(a) Fluorescence output is linked to small molecule input, derived from purified or crude sources. Fluorescence from Responders A and B was analysed after exposure to autoinducer-2 (AI-2) in mono and mixed culture environments. (b) Venus expression from in vitro-synthesized AI-2 added to monocultures of A and B. (c) Venus expression from conditioned media (CM) added to monocultures of A and B. CM was isolated from WT W3110 E. coli cultures sampled at indicated OD. Data are averages from triplicate cultures with s.d. indicated. (d) Red and green fluorescence responses to AI-2 during co-incubation of Responders A (pSBP–mCherry+, red) and B (pSBP–eGFP+, green). Representative fluorescence images show colocalization of red and green cells. Scale bar, 10μm. The average cell count per responder cell is plotted against AI-2 concentration, as determined by image analysis in quadruplicate. All data are plotted as averages of at least triplicate samples with s.d.

Specifically, in Fig. 5b, A and B populations were incubated at mid-exponential phase with in vitro-synthesized AI-2 (refs 50, 51) at concentrations: 0, 2, 10, 28 and 75μM. After 12h, samples were observed for fluorescence by confocal microscopy and then quantified by fluorescence-activated cell sorting (FACS; Supplementary Fig. 4c). We found that SBP–Venus expression for responder A cells occurred at the lowest tested level (2μM AI-2), where 56% of the population expressed SBP–Venus and this fraction increased with AI-2 reaching a maximum of 90% at 28μM. For type B, a more gradual trend was found; only ~1% was fluorescent from 0-2μM, and this increased from 9 to 46% as AI-2 was increased to 28μM. Finally, the highest fraction of fluorescing cells was found at the highest concentration tested, 75μM.

We next isolated CM, which contains a dynamic composition of unfiltered metabolites and media components, from W3110 E. coli cultures at intervals during their exponential growth, throughout which AI-2 accumulates (AI-2 levels for the samples are indicated in Supplementary Fig. 7). CM aliquots were mixed with either A or B cells and cultured in triplicate for 12h. Through FACS analysis it was found, again, that a larger subpopulation of A expressed Venus compared with population B at any concentration (Fig. 5c). Statistically relevant expression from B was not apparent until incubated with CM from cultures at an optical density (OD) of 0.23. In all cases, population A recognized AI-2 presence, including from media isolated at a W3110 OD of 0.05, the minimum cell density tested in this study.

The sensitivities of both strains to AI-2-mediated induction corroborate previous literature10, 47. These trends demonstrate that strains engineered for altered sensitivity to molecular cues provide discrimination of concentration level. That is, the identical plasmid expression system was transformed into different hosts, providing robust and distinct levels of expression.

Having developed cell types A and B with differential ability to detect AI-2, we next altered the reporters so that each cell type expressed a unique SBP-fluorescence fusion for colour-coded designation. Cell type A was engineered with pSBP–mCherry and type B with pSBP–eGFP, resulting in red and green fluorescence, respectively. These populations were mixed together in equal proportion at mid-exponential phase, introduced to a range of AI-2 concentrations, and incubated overnight. Populations A and B exhibited equal growth rates when cultured alone and together (Supplementary Fig. 8c); it followed that the cocultures should comprise a 1:1 ratio of each constituent. Fluorescence output is shown by representative images in Fig. 5d. Also in Fig. 5d, the green and red cell count is plotted from a quadruplicate analysis for each input concentration.

Coculturing enables parallel processing as the molecule-rich environment is perceived by each cell, and is processed uniquely per cell type. Yet, since each sensing mechanism is a living and proliferating population, we tested whether the potentially altered dynamics of coculturing would permit the same sensitivities as isolated culturing. We evaluated the Monod-type saturation constant for each population independently and in cocultures. We found, in Fig. 5d, the general trends in response to an increasing AI-2 level were as predicted by modelled response curves (Supplementary Table 4), which were also well-correlated to Fig. 5b data (Supplementary Fig. 8a,b). That is, the saturation constants that describe dependence on AI-2 were unchanged when measured in cocultures. Phenomenologically, as expected, an initial accumulation of red type A responders was found. Then, at higher AI-2 levels, we found an emergence of a green subpopulation (type B). Above 28μM, there was no longer an apparent differential response that would otherwise enable discrimination of AI-2 concentration; based on the consistency with modelled behaviour, coculturing contributed to dampen the response as the maximum percentage of responding cells in cocultures is 50% instead of 100%. However, the overall fluorescence output is enriched by the combination of multiple populations since the ranges of sensitivity overlap and effectively expand that of the master population (Supplementary Fig. 8d). Specifically, because the fluorescence of B is described by a larger saturation constant, its fluorescence continually increases at higher AI-2 concentrations, while the fluorescence of A remains unchanged. Thus, coculturing between A and B enables resolvable output that is lower than the detection limit of B (due to A) yet surpasses the upper limit at which A saturates by the inclusion of B. The choice to fluorescently differentiate A and B was important because the output would otherwise be biased by extracellular components including the existence of non-sensing cells. Due to colour designation of A and B, a colour ‘pattern’ emerges as a feature of the parallel response, which we recognize is independent of the absolute fluorescence of the population.

Consensus feedback through multidimensional processing

We hypothesized that the value of cell-based sensing would be enhanced if the cell output could be collated in an unbiased manner that in turn were easily ‘read’ using optical means. We engaged magnetic processing, which represents an abiotic processing step that enhances the signal by focusing the collective response. Hence, cells were equipped with streptavidin-conjugated mNPs (Fig. 3). The ability of a magnetic field to refine fluorescence output through filtering and focusing is described in the Supplementary Information (Supplementary Fig. 11). Thus, in our combinatorial approach, fluorescence feedback about molecular information within a microbial community entails biotic processing through constituencies of two independent cell types in conjunction with magnetic post-processing that is enabled by guidance at the nanoscale (Fig. 6c). Moreover, since the fluorescence feedback data is provided through two constituencies, consensus from each independently provides an aggregate output; in our example, the output becomes relayed as a distinctive ‘binned’ category due to finite colour-combinations generated from constituencies A and B (Fig. 6c).

Figure 6: Binning molecular information through cell-based parallel processing and magnetically focusing fluorescence into collective consensus output.

Binning molecular information through cell-based parallel processing and magnetically focusing fluorescence into collective consensus output.

(a) A and B cell types were co-incubated with AI-2 levels ranging from 0 to 55μM AI-2 (left axis), then imaged after magnetic nanoparticle coupling and magnetic collation. Fluorescence results (centred directly over the magnet) are shown from high to low input (top left to bottom right). (b) Quantification of red and green fluorescence cell densities per AI-2 level. (c) The process of accessing molecular information begins by distributing Responders A and B within the environment of an AI-2 producer, P. A and B independently express fluorophore fusions and are linked with magnetic nanoparticles on processing autoinducer-2. Magnetic focusing translocates fluorescing responders. Image analysis of the magnetically collated cell aggregate reveals classified fluorescence output, representing the AI-2 composition of the interrogated environment. (d) Bright field (left) and fluorescence (right, red and green filters) images positioned over the edge of a magnet, as indicated by the inset. The sample in the bottom image pair was isolated from an environment of low AI-2 accumulation. The sample in the top image pair was isolated from a high AI-2 environment. (e) Quantification of visual space occupied by collated cells (eGFP and mCherry expressers) while distributed (- magnet) and magnetically focused (+). Scale bars, 50μm.

Again, type A transmits red output (SBP–mCherry+) and type B transmits green (SBP–eGFP+). These were first co-incubated with titred concentrations of AI-2, to obtain results similar to those ofFig. 5d. By coupling mNPs to the responsive parallel populations, we tested for aggregate two-colour output to provide informative feedback within a set of outcomes ranging from no colour, red-only to red+green. After overnight co-incubation and a magnetic sweep with streptavidin-mNPs, fluorescence results are shown in Fig. 6a, where the recovered cells are displayed above a magnet’s center in order from highest to lowest AI-2 level (top left to bottom right). The processing output generated by the range of conditions was quantitatively assessed for contributions from A and B responders. The spatial density of each fluorophore, or the area occupied by fluorescent responders as a percentage of total visible area, was quantified and plotted in Fig. 6b. Here the trend of increasing fluorescence with AI-2 is followed by both A and B cell types; however, red A cells accumulate at a higher rate than green B cells. This relationship between A and B processing is not only consistent with their previous characterizations (Fig. 5) but indicates that the aggregate output is unbiased regardless of assembly with mNPs and magnetic-stimulated redistribution (Supplementary Information, Supplementary Fig. 14a).

Next, A and B cells were added together to probe the QS environment of Listeria innocua, an AI-2-producing cell type that is genetically and ecologically similar to the pathogenic strain L. monocytogenes52. The environment was biased towards low and high cell density conditions by altering nutrient levels to develop contrasting scenarios of AI-2 level. Preliminary characterization in the Supplementary Information indicated that L. innocua proliferation is unperturbed by the presence of E. coli responders (Supplementary Fig. 12) and that type A cells detect AI-2 at lowListeria densities limited by sparse nutrients; then with rich nutrient availability, cell proliferation permits a higher AI-2 level that can be detected by type B (Supplementary Fig. 13). Replicating these conditions, we expected red fluorescence to be observed at low culture density and for green fluorescence to be reported when high (Fig. 6c). Two conditions were tested: L. innocua was proportioned to responder cells at 20:1 in dilute media to establish a low culture density condition or, alternatively, a ratio of 200:1 in rich media for a high culture density condition. After overnight co-incubation and a magnetic sweep (applied directly to the triple strain cultures) with streptavidin-mNPs, the recovered cells are displayed above a magnet’s edge (shown in Fig. 6d). Acute focusing of the fluorescence signals, contributed by each subset population of the processor (A and B), is visually apparent. The magnetic field had a physical effect of repositioning the ‘on’ subsets to be tightly confined within the magnetic field.

The processing output generated by the contrasting culture conditions was again assessed for the respective contributions of A and B, and for changes in spatial signal density due to the magnetic sweep (Fig. 6e). The analysis was based on images provided in Supplementary Fig. 14b. Data inFig. 6e indicate that red type A cells are prevalent regardless of culture condition (except negative controls). However, compared with the low AI-2 condition, the abundance of green cells is 100-fold higher in the high AI-2 condition. In addition, the ratio of green to red was consistent prior to and after magnetic concentration, substantiating observations in the distributed system. Further, data show that magnetic refining increased per-area fluorescence 100-fold or 10-fold in low and high cell culture studies, respectively.

Based on the thresholds established for responder populations A and B, we found colour-coded binning corresponded to AI-2 level, where ‘red-only’ represented less AI-2 than ‘red+green’ (Fig. 5d). Thus, we found a binned output was established via this multidimensional molecular information-processing system and that this matched the expectations. Red feedback (from responder A) indicated dilute AI-2 accumulation occurred in the low density culture. In the dense cultures, high AI-2 accumulation turned on both A and B for combined red and green feedback.

System response patterns defined by parallel populations

Our example demonstrates the concept of an amorphous processing system that utilizes several biotic and abiotic components for multidimensional information processing. Interestingly, a binning effect was enabled: our system yields an index of colour-categorized feedback that characterizes the sampled environment. In Fig. 7, we present a means to extend our approach to multidimensional systems, those with more than one molecule-of-interest and at different concentrations. That is, by appropriate design of the cell responders, we can further enrich the methodology, its depth and breadth of applicability. We depict 10 hypothetical pairs of responses (with defining equations located in Supplementary Table 5)—those that can be driven by appropriately engineering cells to portend altered genetic responses. For example, rows 1 and 3 provide genetic outcomes as a function of analyte (AI-2) concentration. The hypothetical depictions are feasible as ‘designer’ signal transduction and marker expression processes enabled by synthetic biology21, 53, 54. Rows 2 and 4 demonstrate the corresponding visual planes, where red cell numbers (x-axis) are plotted against green (y-axis), illustrated by the first example. If one divides the two-dimensional space into quadrants (no colour, majority red, majority green, and equivalent ratios of red and green), it becomes apparent that the relationship between cell types influences the ‘visual’ or optical output. Thus, the 10 arbitrary response sets yield a variety of pairings that can provide unique visual patterns for categorizing molecular information. We have simplified the analysis by placing dot marker symbols at the various coincident datapoints, revealing visual patterns. In this way, the ability to incorporate unique responses to a multitude of molecular cues, all within a single pair of cells, or through further multiplexing with additional cell populations becomes apparent.

Figure 7: Extension of nano-guided cell networks for hypothetical regulatory structures.

Extension of nano-guided cell networks for hypothetical regulatory structures.

(a) Rows 1 and 3 depict 10 hypothetical genetic responses to molecular inputs for pairs of fluorescence-reporting cell populations (red, R and green, G). Rows 2 and 4 depict genetic responses as phase-plane plots yielding distinct patterns. This establishes a visual field, showing the extent of any population–population bias (illustrated in example case 1). (b) Left panel: a two-population pairing (shown in case 10) defines visual output that inherently bins into three quadrants: Q1, negligible colour; Q2, red bias due to majority red cell output; and Q4, combined red and green output. Right panel: data from Figs 5d and 6bare plotted analogously, where each data point represents an autoinducer-2 input (labelled, μM). As expected, red and green outputs were binned into Q1, Q2 and Q4 as indicated by coloured outlines.

Our AI-2-conveying cell network is similar to example 7 in Fig. 7a and the AI-2 response curves inFig. 5 (characterized by Supplementary Table 4 equations). Example 7 establishes output into three basic quadrants, including Q1 (negligible colour), Q2 (majority red) and Q4 (roughly equal red and green) (Fig. 7b). We recast the data from Figs 5d and 6b as a phase-plane portrait in Fig. 7c. This reveals the mechanisms by which the output is binned and how the originating cell response curves lead to this pattern, which in turn, was unchanged due to magnetic refinement. InSupplementary Fig. 15, we demonstrate a parameterization of the red and green response curves that suggest the methodology is robust, that when cells are appropriately engineered one could ‘tune’ system characteristics to enhance or diminish a binning effect. We suggest that the utility of subcellular genetic tuning extends well beyond per-cell performance. Rather, we suggest such a strategy may be used to guide the dynamics of population architecture for actuation of by-design response patterns at a systems level.

Discussion

While cell-based sensors work well in well-defined assay conditions, extension to complex environments remains a challenge. They grow, they move, they perturb their environs, they report in a time and concentration-dependent manner, small numbers of sensor cells may require signal amplification and so on. Also, increasingly, bacterial cells are engineered for user specified ‘executive’ functions in complex environments55, 56, 57. Their performance depends on their ability to filter out extraneous noise while surveying the molecular landscape, and providing informed actuation.

Our system interrogates the molecular space by focusing on bacterial QS and a widely distributed signal molecule, AI-2. In addition to genetic attributes of the AI-2-responding sensor cells, AI-2 is a chemoattractant for E. coli, and hence E. coli engineered to sense and respond to AI-2 will naturally move towards its sources, enabling full sampling of the prevailing state10, 37. Each strain evaluates AI-2 with a distinct sensitivity. When ‘activated’ in response to a characteristic level, the cells simultaneously expressed a fluorescent marker and a SBP on the outer membrane via AIDAc translocation. SBP provides a means for cell hybridization through its strong affinity to streptavidin, and here, aids in binding mNPs. This enables the non-genetically coded property of cell translocation within a magnetic field through physically stimulated focusing and binning.

By making use of a diversity of biotic and abiotic features, our multidimensional system of ‘responder’ populations exemplifies several key metrics that promote executive performance in such environments: active molecule capture, post-capture refining of the detection output and finally the utilization of multiple feedback thresholds58, 59, 60. Here cells facilitate AI-2 recognition autonomously and actively because, as a distributed network they reside planktonically, chemotaxing to and continually processing signals over time. When AI-2 is detected, a processor cell’s cognate machinery responds by upregulation of the native QS operon, leading to rapid signal uptake and thereby creating an active-capture signal-processing mechanism. To maximize information acquisition and account for a potentially heterogeneous molecular landscape, cells serve as molecular sampling units among a distributed population, which leads to data fed back as a consensus of fluorescent ‘datapoints’. Then, distributed data collection can be selectively reversed via the incorporated abiotic feature: mNPs, fastened externally on the cell through affinity-guided self-assembly. As such, responding cells obtained this extendable feature, thereby becomes sensitized to repositioning within a magnetic field.

The layered nature of the processor here, from the subcellular to multicellular scale, permits a series of selective steps: it commences with the AI-2-triggered expression cascade which releases a tight repressor, surface localization of both the fluorescent protein and SBP tag, and finally nanoparticle binding for recovery. In addition, multiple layers of amplification result in orthogonal fluorescence feedback. The AI-2 detection event leads to whole-cell fluorescence through expression of many protein copies47. Then their physical collection further amplifies the signal, yielding a macroscopic composite of many individual cell units. When utilized as a network of multiple constituencies, responder cell types A and B contribute individual recognition results (off, red or green) to a single consensus output. Finally, due to their overlapping thresholds for recognition of the same molecule, in this case, AI-2, parallel processing by A and B responders can contribute to visual interpretation of information about the molecule. Outcomes are classified into a finite number of states: here output to no fluorescence, red, or red and green, with each addition of colour as a metric of a higher interval of AI-2. In many respects, the elucidation of layered information networks as demonstrated here is analogous to computer information processing via information theory61, 62, 63.

Here, however, interrogation of biological systems requires a reliable means for accessing molecular information—that which is communicated between biological species and that which can be relayed to the end user. The responder cells need not be present in high concentration, nor must they all be collected in the present format. We suggest that engineered biological mechanisms are well-poised to serve at this critical interface between information acquisition and user interaction. Thus, the functional design of components for autonomous self-assembly, decision-making and networking is requisite in the field of micro- and nano-scaled machines. Our combinatorial approach allows for cells to independently assess, yet collectively report, on molecular information. Its processing is enabled through appropriate integration of synthetic biology and nanomaterials design. We suggest this approach provides a rich opportunity to direct many formats of multi-population response through genetic tuning and systems-level engineering. Further development of cellular networks and incorporation of alternate abiotic attributes can expand the depth and breadth of molecular communication for user specified actuation.

Read Full Post »


Cancer Genomics: Multiomic Analysis of Single Cells and Tumor Heterogeneity

Curator: Stephen J. Williams, PhD

 

scTrio-seq identifies colon cancer lineages

Single-cell multiomics sequencing and analyses of human colorectal cancer. Shuhui Bian et al. Science  30 Nov 2018:Vol. 362, Issue 6418, pp. 1060-1063

To better design treatments for cancer, it is important to understand the heterogeneity in tumors and how this contributes to metastasis. To examine this process, Bian et al. used a single-cell triple omics sequencing (scTrio-seq) technique to examine the mutations, transcriptome, and methylome within colorectal cancer tumors and metastases from 10 individual patients. The analysis provided insights into tumor evolution, linked DNA methylation to genetic lineages, and showed that DNA methylation levels are consistent within lineages but can differ substantially among clones.

Science, this issue p. 1060

Abstract

Although genomic instability, epigenetic abnormality, and gene expression dysregulation are hallmarks of colorectal cancer, these features have not been simultaneously analyzed at single-cell resolution. Using optimized single-cell multiomics sequencing together with multiregional sampling of the primary tumor and lymphatic and distant metastases, we developed insights beyond intratumoral heterogeneity. Genome-wide DNA methylation levels were relatively consistent within a single genetic sublineage. The genome-wide DNA demethylation patterns of cancer cells were consistent in all 10 patients whose DNA we sequenced. The cancer cells’ DNA demethylation degrees clearly correlated with the densities of the heterochromatin-associated histone modification H3K9me3 of normal tissue and those of repetitive element long interspersed nuclear element 1. Our work demonstrates the feasibility of reconstructing genetic lineages and tracing their epigenomic and transcriptomic dynamics with single-cell multiomics sequencing.

Fig. 1 Reconstruction of genetic lineages with scTrio-seq2.

Global SCNA patterns (250-kb resolution) of CRC01. Each row represents an individual cell. The subclonal SCNAs used for identifying genetic sublineages were marked and indexed; for details, see fig. S6B. On the top of the heatmap, the amplification or deletion frequency of each genomic bin (250 kb) of the non-hypermutated CRC samples from the TCGA Project and patient CRC01’s cancer cells are shown.

” data-icon-position=”” data-hide-link-title=”0″>

 

Fig. 1 Reconstruction of genetic lineages with scTrio-seq2.

Global SCNA patterns (250-kb resolution) of CRC01. Each row represents an individual cell. The subclonal SCNAs used for identifying genetic sublineages were marked and indexed; for details, see fig. S6B. On the top of the heatmap, the amplification or deletion frequency of each genomic bin (250 kb) of the non-hypermutated CRC samples

 

 

Read Full Post »


Real Time Coverage @BIOConvention #BIO2019: Gene Therapy 2.0: No Longer Science Fiction 1:00-2:15 pm June 3 Philadelphia PA

Reporter: Stephen J. Williams Ph.D. @StephenJWillia2

kkjk

Other Articles on Gene Therapy on this Open Access Journal Include:

Read Full Post »


Human gene editing continues to hold a major fascination within a biomedical and biopharmaceutical industries. It’s extraordinary potential is now being realized but important questions as to who will be the beneficiaries of such breakthrough technologies remained to be answered. The session will discuss whether gene editing technologies can alleviate some of the most challenging unmet medical needs. We will discuss how research advances often never reach minority communities and how diverse patient populations will gain access to such breakthrough technologies. It is widely recognize that there are patient voids in the population and we will explore how community health centers might fill this void to ensure that state-of-the-art technologies can reach the forgotten patient groups . We also will touch ethical questions surrounding germline editing and how such research and development could impact the community at large.

Please follow LIVE on TWITTER using the following @ handles and # hashtags:

@Handles

@pharma_BI

@AVIVA1950

@BIOConvention

# Hashtags

#BIO2019 (official meeting hashtag)

Read Full Post »


Complex rearrangements and oncogene amplification revealed by long-read DNA and RNA sequencing of a breast cancer cell line

Reporter: Stephen J. Williams, PhD

In a Genome Research report by Marie Nattestad et al. [1], the SK-BR-3 breast cancer cell line was sequenced using a long read single molecule sequencing protocol in order to develop one of the most detailed maps of structural variations in a cancer genome to date.  The authors detected over 20,000 variants with this new sequencing modality, whereas most of these variants would have been missed by short read sequencing.  In addition, a complex sequence of nested duplications and translocations occurred surrounding the ERBB2 (HER2) while full-length transcriptomic analysis revealed novel gene fusions within the nested genomic variants.  The authors suggest that combining this long-read genome and transcriptome sequencing results in a more comprehensive coverage of tumor gene variants and “sheds new light on the complex mechanisms involved in cancer genome evolution.”

Genomic instability is a hallmark of cancer [2], which lead to numerous genetic variations such as:

  • Copy number variations
  • Chromosomal alterations
  • Gene fusions
  • Deletions
  • Gene duplications
  • Insertions
  • Translocations

Efforts such as the Cancer Genome Atlas [3], and the International Genome Consortium (2010) use short-read sequencing technology to detect and analyze thousands of commonly occurring mutations however short-read technology has a high false positive and negative rate for detecting less common genetic structural variations {as high as 50% [4]}. In addition, short reads cannot detect variations in close proximity to each other or on the same molecule, therefore underestimating the variation number.

Methods:  The authors used a long-read sequencing technology from Pacific Biosciences (SMRT) to analyze the mutational and structural variation in the SK-BR-3 breast cancer cell line.  A split read and within-read mapping approach was used to detect variants of different types and sizes.  In general, long-reads have better alignment qualities than short reads, resulting in higher quality mapping. Transcriptomic analysis was performed using Iso-Seq.

Results: Using the SMRT long-read sequencing technology from Pacific Biosciences, the authors were able to obtain 71.9% sequencing coverage with average read length of 9.8 kb for the SK-BR-3 genome.

A few notes:

  1. Most amplified regions (33.6 copies) around the locus spanning the ERBB2 oncogene and around MYC locus (38 copies), EGFR locus (7 copies) and BCAS1 (16.8 copies)
  2. The locus 8q24.12 had the most amplifications (this locus contains the SNTB1 gene) at 69.2 copies
  3. Long-read sequencing showed more insertions than deletions and suggests an underestimate of the lengths of low complexity regions in the human reference genome
  4. Found 1,493 long read variants, 603 of which were between different chromosomes
  5. Using Iso-Seq in conjunction with the long-read platform, they detected 1,692,379 isoforms (93%) mapping to the reference genome and 53 putative gene fusions (39 of which they found genomic evidence)

A table modified from the paper on the gene fusions is given below:

Table 1. Gene fusions with RNA evidence from Iso-Seq and DNA evidence from SMRT DNA sequencing where the genomic path is found using SplitThreader from Sniffles variant calls. Note link in table is  GeneCard for each gene.

SplitThreader path

 

# Genes Distance
(bp)
Number
of variants
Chromosomes
in path
Previously observed in references
1 KLHDC2 SNTB1 9837 3 14|17|8 Asmann et al. (2011) as only a 2-hop fusion
2 CYTH1 EIF3H 8654 2 17|8 Edgren et al. (2011); Kim and Salzberg
(2011); RNA only, not observed as 2-hop
3 CPNE1 PREX1 1777 2 20 Found and validated as 2-hop by Chen et al. 2013
4 GSDMB TATDN1 0 1 17|8 Edgren et al. (2011); Kim and Salzberg
(2011); Chen et al. (2013); validated by
Edgren et al. (2011)
5 LINC00536 PVT1 0 1 8 No
6 MTBP SAMD12 0 1 8 Validated by Edgren et al. (2011)
7 LRRFIP2 SUMF1 0 1 3 Edgren et al. (2011); Kim and Salzberg
(2011); Chen et al. (2013); validated by
Edgren et al. (2011)
8 FBXL7 TRIO 0 1 5 No
9 ATAD5 TLK2 0 1 17 No
10 DHX35 ITCH 0 1 20 Validated by Edgren et al. (2011)
11 LMCD1-AS1 MECOM 0 1 3 No
12 PHF20 RP4-723E3.1 0 1 20 No
13 RAD51B SEMA6D 0 1 14|15 No
14 STAU1 TOX2 0 1 20 No
15 TBC1D31 ZNF704 0 1 8 Edgren et al. (2011); Kim and Salzberg
(2011); Chen et al. (2013); validated by
Edgren et al. (2011); Chen et al. (2013)

 

SplitThreader found two different paths for the RAD51B-SEMA6D gene fusion and for the LINC00536-PVT1 gene fusion. Number of Iso-Seq reads refers to full-length HQ-filtered reads. Alignments of SMRT DNA sequence reads supporting each of these gene fusions are shown in Supplemental Note S2.

 

 References

 

  1. Nattestad M, Goodwin S, Ng K, Baslan T, Sedlazeck FJ, Rescheneder P, Garvin T, Fang H, Gurtowski J, Hutton E et al: Complex rearrangements and oncogene amplifications revealed by long-read DNA and RNA sequencing of a breast cancer cell line. Genome research 2018, 28(8):1126-1135.
  2. Hanahan D, Weinberg RA: The hallmarks of cancer. Cell 2000, 100(1):57-70.
  3. Kandoth C, McLellan MD, Vandin F, Ye K, Niu B, Lu C, Xie M, Zhang Q, McMichael JF, Wyczalkowski MA et al: Mutational landscape and significance across 12 major cancer types. Nature 2013, 502(7471):333-339.
  4. Sudmant PH, Rausch T, Gardner EJ, Handsaker RE, Abyzov A, Huddleston J, Zhang Y, Ye K, Jun G, Fritz MH et al: An integrated map of structural variation in 2,504 human genomes. Nature 2015, 526(7571):75-81.

 

Other articles on Cancer Genome Sequencing in this Open Access Journal Include:

 

International Cancer Genome Consortium Website has 71 Committed Cancer Genome Projects Ongoing

Loss of Gene Islands May Promote a Cancer Genome’s Evolution: A new Hypothesis on Oncogenesis

Identifying Aggressive Breast Cancers by Interpreting the Mathematical Patterns in the Cancer Genome

CancerBase.org – The Global HUB for Diagnoses, Genomes, Pathology Images: A Real-time Diagnosis and Therapy Mapping Service for Cancer Patients – Anonymized Medical Records accessible to

 

Read Full Post »

Older Posts »