In late January, the Salk Institute hosted a rather powerhouse conference on the state of synthetic biology: the Tenth Annual Symposium on Biological Complexity, along with Fondation IPSEN and Science. The speakers pretty well ranged the gamut of syn bio, and they presented some impressive work, from expanding the genetic code, to developing an “acoustic GFP”, to making a programming language and compiler for cells. To our category-happy ears, there seemed to be six major themes afield, so we’ll try to sum up what we saw in:
- Basic Principles (e.g., computing sense from cross-talk madness)
- Biological Logic and Computing (e.g., cellular programming language)
- New Tools (e.g., the GFP of ultrasound)
- Therapy (e.g., programmed cancer killers)
- Stuff Life Ain’t Never Seen Before (e.g., expanding the genetic code)
- Bio-Production (e.g., biofuels, medicines, and nanomaterials)
One branch of synthetic biology appropriates as its motto “What I cannot create, I do not understand” from Richard Feynmann. In this vein, some speakers were elucidating basic biological principles by putting things together, like Michael Elowitz from Cal Tech, who had two basic principles to share – how BMP receptors perform complex computations, and how chromatin stores epigenetic memories.
BMP receptors come in many forms and interact with many different signals, and all the cross-talk makes it hard to see how cells can have any idea what’s going on. Extracting sense from the madness, Elowitz showed how this system can compute the sum or ratio of two different ligands or respond only when the signals are either balanced or unbalanced, and that cells can change their calculation type by simply changing the ratio of two receptors.
Moving on to epigenetic memory, Elowitz proposed a three-state model for gene silencing, with genes shifting between active and reversibly silenced, and occasionally becoming permanently silenced. Different chromatin modifications had different effects, with some being easily reversible and some silencing genes for good.
Ned Wingreen from Princeton University emphasized a principle important for metabolic pathways: enzyme clustering. Most biological pathways involve several steps and multiple enzymes, and Wingreen showed that clustering the enzymes within microdomains helps maximize the flux through the pathway. His group even redirected metabolic flux inside cells just by clustering enzymes, which suggests a potential new strategy for optimizing yield of biofuels or other biochemicals.
Daniel Fletcher from UC Berkeley answered a long-standing question of how early Xenopus embryos match the sizes of their mitotic spindles to their cells – a problem because the cells get smaller with each division. By making different size droplets of egg extract, he showed that the droplet sizes directly controlled the spindle size, with no need for a DNA-encoded development program. Presumably, the droplet size limits the number of spindle building blocks, which sets the spindle size. In another story, he showed how dimerizing membrane proteins accumulate where two vesicles touch each other, and how this can even crowd out other proteins.
One basic principle we’ve all had drummed into us is that correlation does not imply causation, but Gerald Pao from the Salk Institute pointed out that nor does causation always imply correlation. Using the convergent cross-mapping method developed by George Sugihara, he was able to find causation between yeast genes from RNAseq data, even when the genes had no classical correlation. And speaking of algorithms…
Biological Logic and Computing
Another branch of synthetic biology is populated by computer scientists and electrical engineers trying to make biology as easily programmable as computers. This may not sound like it could ever work for something as squishy as biology, but Chris Voigt of MIT put a significant dent into the naysaying with his cellular programming language, CELLO (still in Beta). Instead of electrons, these cellular circuits control the flow of RNA polymerase, and the language is based on “NOR gates”, really promoters that are expressed only when neither of two repressors is present.
NOR gates can be combined to make every other Boolean logic gate, so in principle you could make any digital electronic circuit. After developing a set of ~30 orthogonal repressors and ~40 sensors, the lab made a compiler that lets users define how a set of outputs should respond to a set of inputs (i.e., a truth table), and outputs a DNA sequence to implement the circuit – all a user has to do is order the DNA!
Out of 60 desired circuits, the compiler designed a sequence that worked on the first try 75% of the time. CELLO is limited to transcriptional regulation and steady-state outputs for now, but it looks like a huge step forward for abstracted biological computing.
Meanwhile, another Bostonian, Timothy Lu of MIT, demonstrated biological “state machines”, which have memory of previous signals and respond in different ways depending on the order of signals. For example, a biosensor cell in your yogurt might do one thing if it senses digestion problems in the stomach, and another if it senses them after it has seen intestine environment. The design uses phage recombinases to write events to DNA, and it worked on up to a 3-input system with 16 distinct states. The DNA state can be read with PCR, and like CELLO, they are working on an algorithm to aid design.
Wendell Lim, at UC San Francisco, demonstrated the power of integrating multiple signals like this in killer T cells. He first showed how to make synthetic Notch cell surface receptors, and then combined them in T cells that would only activate killing if they detected two particular antigens in the correct order. In mice, this allowed the designer T cells to specifically target one particular type of cancer cell!
Also sticking it to cancer with designer immune cells was Ron Weiss of MIT. In addition, he’s working on a homeostasis system to keep pumping out new insulin-secreting beta cells in Type I diabetes.
Another seeker of homeostasis was Martin Fusenegger from ETH Zurich, who envisions a future with “metabolic prostheses”. Biosensor cells could be implanted in a protected matrix, allowing them to measure metabolites and secrete an appropriate response at the first sign of trouble. And speaking of treatments…
Underlying essentially all biological research is the hope that discoveries will eventually turn into treatments. A good example of that was Jeff Hasty of UC San Diego, who explored genetic oscillators. First his lab made a simple cellular oscillator that made cells blink on and off, then they scaled it up by synchronizing cells first within a colony and then across colonies on a macroscopic scale. When transferred into tumor-homing Salmonella and armed with a cancer toxin, an oscillating gene circuit delivered pulses of toxin directly to tumors, which, when combined with chemotherapy, significantly extended the lives of mice.
On the viral end of the disease spectrum, Rino Rappouli of GSK Vaccines talked about how synthetic biology helps develop vaccines against recalcitrant pathogens by designing optimal antigens to target.
Mark Davis of Cal Tech took a more chemical approach to therapy, designing self-assembling nanoparticles that can deliver drugs or genetic therapies like siRNA to specific target cells or even across the blood-brain barrier.
Of course, treatment begins with diagnosis, so Orystya Stus from UC San Diego helped develop a cheap and easy HIV sensor by simply printing cell extract with a genetic sensor circuit onto filter paper. The result is cheap, sensitive, and robust, and it could dramatically expand testing into remote areas. And speaking of handy tools…
Tools and Techniques
Science and engineering are only as good as the tools we have to do them with – witness the revolution wrought by GFP. As we begin trying to understand and hack the bacteria living in us, though, GFP isn’t going to cut it. Light can’t get into our guts and back, so we have no way to see where different bugs are or what they’re doing in real time. We do, however, peer inside ourselves with ultrasound pretty regularly, so Raymond Bourdeau of Cal Tech developed what he hopes will be the “acoustic GFP”.
Cyanobacteria have tiny gas vesicles – actually hollow protein shells – that fill with gas. These increase contrast in ultrasound imaging, just like conventional microbubbles. Bourdeau optimized a set of gas vesicle genes that he could transfer between species, making them show up in ultrasounds. What’s more, the gas vesicles collapse under strong acoustic pulses, akin to photobleaching GFP, with a tunable threshold. That means different bugs could be labeled with different collapse thresholds, allowing multiplex ultrasound imaging.
Nicholas Butzin from Virginia Tech showed off another cool technique – translational “gates” made up of strings of rare codons. When ribosomes hit these gates, they stall, collide, and fall off, preventing expression of the downstream gene. The gates can be opened by inducing the truant tRNAs that match the rare codons. And speaking of opening new gates…
Stuff Life Ain’t Never Seen Before
In addition to rearranging existing biological parts, part of synthetic biology is trying to make brand new things. For example, for all known history of all life everywhere ever, DNA has contained 4 nucleotides making 2 base pairs. Floyd Romesberg from The Scripps Research Institute gave a special lecture about adding a new base pair. After a lot of hard work, his lab found two nucleotides, dNaM and d5SICS, that worked almost as well as standard DNA in vitro (i.e., PCR).
To avoid wobble pairing with natural bases, these new bases pair using hydrophobic interactions, instead of hydrogen bonding. This means long runs of them would destabilize the DNA double helix, but Romesberg doesn’t care, because all he needs is one site per codon to expand the genetic code, thus allowing incorporation of new amino acids and hugely expanding the space of possible protein functions. By endowing E. coli with a nucleotide transporter, Romesberg’s lab even got the non-natural base pairs to replicate in vivo! Transcription into RNA works too, and he seemed coyly hopeful that translation would be working soon as well.
The rest of the “new space” bunch were designing new proteins, headlined by Frances Arnold of Cal Tech. Arnold gave another special lecture, firmly anchored by the belief that evolution is the best engineer the world has ever seen. Based on that philosophy, she showed that directed evolution can generate new functions starting with just a tiny bit of enzyme promiscuity. Harnessing this process, and injecting quite a bit of rational tweaking, she evolved enzymes with new abilities and new specificities that chemists would never be able to achieve.
Wilfred van der Donk, from the University of Illinois at Urbana-Champaign, took a different approach to novelty, branching out from the standard linear amino acid chain. Using natural dehydratase and cyclase enzymes, he explored the space of cyclic peptides, including genome mining, engineering new peptide loops, and developing cyclic peptides with new, potentially therapeutic activities.
With all these new proteins floating around, what might they actually do? Katie Digianantonio from Princeton University screened a library of brand new, but well-folded, proteins for ability to rescue essential gene knockouts in E. coli. Interestingly, twice she found that the rescuing proteins did not replace the missing enzyme activity, but instead upregulated native enzymes that had some catalytic promiscuity. This suggests it may be easier for de novo proteins to change gene expression than to have essential enzymatic activity.
Another protein-designer was David Baker from the University of Washington, Seattle, who actually starts with his desired protein structure, and then designs an amino acid sequence to fold into it. This lets him design proteins to tightly bind other proteins or ligands, and he can also make proteins that self-assemble into larger complexes.
Finally, James Wells of UC San Francisco engineered proteins ranging from enzyme inhibitors to new biomarkers for intracellular signaling events like protein phosphorylation.
The last cluster of research we saw focused on harnessing biology to make useful things, whether chemicals or fuel. Sarah O’Connor, from the John Innes Centre, described some nice work mapping and hacking metabolic pathways in plants like Madagascar Periwinkle, which makes many medicinal compounds. By rewiring these pathways, she can make new compounds that could become even better medicines.
Michelle Chang of UC Berkeley did some nice metabolic rewiring herself, studying how enzymes work, optimizing them both rationally and through directed evolution, and mixing and matching them to create desired pathways. One key achievement was highly efficient production of butanol, a potential biofuel.
Angela Belcher came from MIT to describe her work using engineered phage to template highly precise, self-assembled nanomaterials. This approach has grown such wonder materials as better batteries, fuel cells, and solar cells.
Of course, for any of these fledgling bio-products to make it to store shelves, they have to make it through the scale-up valley of death. Gregory Stephanopoulos from MIT shared some insight into how to make bio-products at a volume and cost to really help the world.
Walking out of this symposium, it was clear that synthetic biology is really truly maturing. People on all its branches aare making impressive progress, whether that be a true programming language for genetic circuits, acoustically imaging the microbiome in vivo, expanding the genetic code, programming natural cancer killers, or making amazing new materials. And all this even though we missed the closing talk by Craig Venter (J. Craig Venter Institute)!