Chemistry in Living Systems, by the Numbers

This essay was prompted by Carolyn Bertozzi’s lecture “Chemistry in Living Systems”, given at UC Irvine in January 2014 as the ISIS Pharmaceuticals Lecture and posted by UCI Media (watch it here). It is a talk for chemists. What a quantitative reader will notice is how often she stops to give a number: a rate constant, an order of magnitude, a count of molecular species in a drug. Those numbers are the subject here. Bioorthogonal chemistry is usually presented as a set of clever reactions. Read as measurement, it is the layer that makes the sugars on cells and proteins visible and countable, even though they have no gene and no sequence. This piece is part of a series on how molecular biology gets turned into mathematics and measurement.

What bioorthogonal means, and why the constraint is so tight

Bertozzi opens with a contrast that defines the whole field. A synthetic chemist in a fume hood controls the solvent and the temperature, the atmosphere, and the order and rate of addition. Every one of those controls is lost inside a living cell, where the solvent is water at roughly neutral pH and 37 °C and the reagents must be nontoxic. In a whole animal you also inherit metabolism, circulation, and serum half-lives. Her definition of the term, which her lab coined, is that bioorthogonal chemistry is “chemistry among functional groups that neither interact with nor interfere with a biological system” (from the lecture). The 2005 review by Prescher and Bertozzi lays out the two-step strategy that follows from that definition. A small chemical reporter, often a single functional group, is built into a target biomolecule by the cell’s own biosynthetic machinery. A probe delivered from outside then reacts with that reporter and nothing else ¹. The 2009 review by Sletten and Bertozzi adds the kinetic requirement: the two partners must react rapidly and selectively under physiological conditions in the presence of everything a cell contains ².

Why glycans in particular need this approach is worth spelling out. A protein can be tagged by fusing its gene to green fluorescent protein, so the cell makes the tag for you. There is no gene for a glycan. Sugars are assembled by competing enzymes without a template, and the same protein site can carry many different glycan structures at once ³. In the lecture Bertozzi describes cell-surface glycans as an integrated data set that reflects gene expression, nutrient availability, and central metabolism. She notes that the glycome of a stem cell differs from that of differentiated cells and that malignant transformation produces stereotyped changes in glycosylation. Her summary of the state of the field is blunt: “we don’t really know how to read it” (from the lecture). The chemical reporter was her lab’s answer to the reading problem.

The azide as the reporter

The reporter that carried most of the glycan work is the azide, a group of three nitrogen atoms. Bertozzi’s case for it in the lecture is a list of design constraints. It is abiotic, so nothing in the body carries one. It is small, so the enzymes of a sugar pathway will tolerate it. And it has two modes of reactivity that are rare in biology: it is a soft electrophile and a 1,3-dipole, meaning it can undergo cycloadditions with alkynes and related partners. The metabolic entry point was the sialic acid pathway. Feeding cells an azide-bearing analogue of N-acetylmannosamine, which the lab calls ManNAz, lets six consecutive enzymes carry the azide through to sialic acid on the cell surface. She describes that tolerance as a lucky break.

The first reaction to exploit the azide on living cells was the Staudinger ligation, reported by Saxon and Bertozzi in 2000. An engineered triarylphosphine reacts with the azide and rearranges to form a stable amide bond, and azides installed in cell-surface glycans by metabolism of an azidosugar were captured with a biotinylated phosphine ⁴. In the lecture she recalls that the reaction also works when both reagents are injected into a mouse and left to find each other. The problem was speed.

A ladder of rate constants

Everything that follows depends on one number per reaction, the second-order rate constant k, in units of M⁻¹ s⁻¹. For a bioorthogonal reaction in an animal, the reporter on a glycan sits at a fixed and usually low concentration. The probe is at whatever concentration you can safely inject. Under those conditions the reaction behaves as pseudo-first-order in the reporter, and the half-life of unreacted reporter is ln 2 divided by k times the probe concentration. That formula is the whole argument of this section. A slow reaction can be rescued only by raising the probe concentration, and probe concentration is capped by toxicity and clearance.

Sletten and Bertozzi state the consequence directly. The bioorthogonal reactions available in 2009 spanned rate constants from about 10⁻⁴ to 10³ M⁻¹ s⁻¹, and those values force the use of high micromolar to millimolar concentrations of the secondary reagent in vivo ². The rungs of the ladder are these:

• The Staudinger ligation: about 10⁻³ M⁻¹ s⁻¹ with alkyl azides ². Bertozzi’s verdict in the lecture is that it takes many hours to reach completion at the concentrations an animal permits. Metabolism can outrun it and nothing fast can be studied.
• The copper-catalyzed azide-alkyne cycloaddition, the click reaction of Sharpless and Meldal: fast, but the copper(I) catalyst is toxic to cells and so unusable for live imaging ⁵. She notes in the talk that ligand engineering to sequester the copper was being worked on.
• The strain-promoted cycloaddition of Agard, Prescher, and Bertozzi in 2004: an alkyne bent into an eight-membered ring reacts with azides without any catalyst and with no apparent toxicity on living cells ⁵. In the lecture she puts the ring strain of a cyclooctyne in the range of 20 kcal/mol and credits the physical organic literature of Huisgen and his contemporaries for the idea. The first-generation compound her lab made reacted at about 1 × 10⁻³ M⁻¹ s⁻¹ in a model reaction. She describes that as on par with the Staudinger ligation and therefore not enough on its own.
• DIFO, a difluorinated cyclooctyne reported by Baskin and colleagues in 2007: fluorine atoms at the propargylic position raise the rate to 7.6 × 10⁻² M⁻¹ s⁻¹ with benzyl azide. That is 17 to 63 times faster than the Staudinger ligation or earlier cyclooctynes ⁶. In the lecture this is the compound the lab first trusted for in vivo imaging.
• BARAC, the biarylazacyclooctynone of Jewett and colleagues in 2010: two fused benzene rings and an amide in the ring add strain. The reagent gives usable signal at nanomolar concentrations without washing steps ⁷. She puts its rate constant at almost a thousand times the first-generation compound and says the cyclooctynes hit a ceiling there, since more reactive versions became unstable.
• The tetrazine ligation of Blackman, Royzen, and Fox in 2008: a tetrazine reacts with trans-cyclooctene by an inverse-electron-demand Diels-Alder reaction with k₂ of about 2,000 M⁻¹ s⁻¹. It runs in water and cell lysate as well as in organic solvent ⁸. Bertozzi closes the lecture by saying these rate constants exceed everything she presented by orders of magnitude.

Putting those numbers through the half-life formula makes the ladder concrete. At a probe concentration of 100 µM, a reaction at 10⁻³ M⁻¹ s⁻¹ has a reporter half-life of about 80 days. DIFO at 7.6 × 10⁻² M⁻¹ s⁻¹ brings that to about 25 hours. A tetrazine ligation at 2,000 M⁻¹ s⁻¹ brings it to about 3.5 seconds, and even at 1 µM it is under six minutes. These are model-reaction values and the constants shift with structure and medium, so treat them as orders of magnitude. Two further points keep the arithmetic straight. First, an image does not require completion. A cell surface carries an enormous number of azido glycans, so converting a small fraction with a bright dye is already visible. This is why Baskin and colleagues could image azide-labeled glycans on cells after one minute with 100 µM DIFO-488 ⁶. Second, signal to background improves as the probe concentration falls, because unreacted probe is what you have to wash away. That is the practical meaning of BARAC’s nanomolar labeling.

The same logic appears in the first story of the lecture, which is about proteins rather than sugars. Her lab wanted an irreversible reaction with an aldehyde installed site-specifically in a protein. The classic Pictet-Spengler reaction does that but with a rate constant she gives as around 10⁻⁴ M⁻¹ s⁻¹. Her student’s redesign, the Pictet-Spengler ligation, replaced the amine with an N-methyl aminooxy group and moved the nucleophilic carbon of the indole. She reports it as four to five orders of magnitude faster, enough to run in aqueous buffer at room temperature in a few hours. The motivation was also numerical. She estimates that an approved antibody-drug conjugate made by random lysine coupling, with around 100 accessible lysines and about 3.5 drugs per antibody on average, is a mixture of upwards of 10,000 distinct molecular species.

Reading glycans in a living animal

With a fast enough copper-free reaction in hand, the lab moved to zebrafish, whose embryos are translucent and develop in a dish. Laughlin and colleagues added an azido sugar to the water around fertilized embryos and later labeled the azides the fish had built into their glycans with DIFO dyes. At 60 hours after fertilization glycan biosynthesis was elevated in the jaw, the pectoral fins, and the olfactory organs. A multicolor scheme let them separate glycans made at different times ⁹. Bertozzi walks through the multicolor version in the lecture. A five-day-old larva is labeled with three dyes at three time points. The red dye marks glycoproteins made just before imaging, and the green and blue mark populations made 10 and 12 hours earlier. Under a confocal microscope the red population sits on the cell surface while the older populations have already been internalized, so the same image reports membrane turnover cell by cell. A later student saw mucin-type glycoproteins concentrate at the cleavage furrow every time an early embryonic cell divided. Her point about method is the one to keep: open a window onto a class of biomolecules that had no imaging method and you see things that were invisible before.

The lecture ends with the same idea applied to bacteria. Bacteria use D-amino acids in their cell wall peptidoglycan and mammals do not, so azide- and alkyne-bearing D-alanine analogues label bacteria orthogonally to a host. In a pulse-chase experiment on mycobacteria, an undergraduate wrote an image-processing routine that integrates fluorescence along the long axis of each cell, and the resulting profiles reproduced the polar asymmetric growth another group had reported by a different method. That is the general pattern in this series: a chemical trick creates a signal, and software turns the signal into a measurement.

The 2022 Nobel Prize

The Nobel Prize in Chemistry 2022 was awarded jointly to Carolyn R. Bertozzi, Morten Meldal, and K. Barry Sharpless “for the development of click chemistry and bioorthogonal chemistry” ¹⁰. The two halves of that citation map onto the ladder above. Sharpless and Meldal’s copper-catalyzed reaction is the fast, general coupling that works wherever a catalyst is allowed. Bertozzi’s contribution is the version that works where it is not, inside cells and animals, together with the metabolic labeling that gives the reaction something to find.

Glycans as a missing data layer

Personal molecular profiles today typically stack a genome and a transcriptome on top of a proteome and blood biomarkers. The glycome is usually absent, and the reason is the same one that made Bertozzi’s chemistry necessary. There is no template to read. Ruhaak and colleagues, in their 2018 review of mass spectrometry for glycomics and glycoproteomics, describe glycans as arguably harder to analyze than any other biopolymer. The reasons they give are the diversity of structures, the range of polarities among sugar residues, and poor ionization efficiency. The number of glycoforms at a single protein site makes it harder to characterise than other post-translational modifications, and the analyses remain largely the province of specialized laboratories ³.

What a quantitative glycomics measurement looks like is quite different from sequencing. In the released-glycan approach, sugars are enzymatically cleaved from the proteins in a sample such as plasma or purified immunoglobulin G. They are then separated by liquid chromatography or capillary electrophoresis and detected by mass spectrometry or fluorescence. The output is a vector of relative abundances over a few dozen glycan structures, normalized to total signal, rather than a list of discrete variants. Glycoproteomics goes one level deeper and keeps the peptide attached, so each measurement is a peptide-plus-glycan pair at a known site ³. That vector is a slowly varying and biologically integrated summary. Krištić and colleagues measured the IgG glycome in 5,117 people from four European populations. Three glycans alone explained up to 58 percent of the variance in chronological age, and the residual variance correlated with physiological markers of biological age ¹¹. That is the kind of number a longitudinal profile is built to track.

The two measurement modes give different things. Bioorthogonal labeling reports where a class of sugar is and how it moves, in a living system and over time. It reads one metabolic entry point at a time and does not return a composition. Mass spectrometry returns a composition, but from a homogenized sample with no spatial or dynamic information. Neither one is a sequence.

What is established and what is still hard

Established: the chemical reporter strategy and azide labeling of sialic acids and other sugars, together with copper-free cycloadditions and tetrazine ligations whose rate constants are reproducible laboratory measurements. Live imaging of glycans in cells and in zebrafish is established too. The ladder from 10⁻³ to 10³ M⁻¹ s⁻¹ is a settled account of what each generation of reagent bought.

Still hard: reading the glycome as a composition rather than an image. Isomers with the same mass but different linkages are difficult to resolve. Site-specific glycoproteomics across a whole proteome remains a specialist task, and comparability across laboratories and platforms is an open problem ³. On the chemistry side, the cyclooctyne series reached the stability ceiling Bertozzi describes, and in vivo use still depends on the pharmacokinetics of the probe. Her own closing remark stands: compared with the shelves of conventional organic synthesis, the compendium of bioorthogonal reactions is short.

Questions people also ask

What does bioorthogonal mean? A reaction whose partners do not react with anything in a living system and are not interfered with by it, so it can be run inside cells or animals under physiological conditions ^{1 2}.

Why does the rate constant matter so much for imaging in a living animal? Because the reporter is at low concentration and the probe concentration is capped by toxicity and clearance. The half-life of the reaction scales as 1 divided by k times the probe concentration, so a thousand-fold change in k is the difference between days and seconds ².

What is the difference between click chemistry and bioorthogonal chemistry? Click chemistry in the sense of Sharpless and Meldal is a fast copper-catalyzed coupling of azides and alkynes. Bioorthogonal chemistry drops the catalyst so the reaction can run in living systems, using strained alkynes or tetrazines instead ^{5 8 10}.

Can glycans be measured in a personal molecular profile? Yes, by released-glycan or glycoproteomic mass spectrometry, though these assays are still mostly done in specialized laboratories ³. The IgG glycome is the best-studied case and tracks age closely ¹¹.

Woolf Software builds longitudinal molecular profiles of individuals: whole-genome sequencing, RNA sequencing, proteomics, blood biomarkers, and continuous glucose data, integrated into one model of you. Build your profile.

Footnotes

1. Jennifer A. Prescher, Carolyn R. Bertozzi. Chemistry in living systems. Nature Chemical Biology, 2005. https://doi.org/10.1038/nchembio0605-13 ↩ ↩²

2. Ellen M. Sletten, Carolyn R. Bertozzi. Bioorthogonal Chemistry: Fishing for Selectivity in a Sea of Functionality. Angewandte Chemie International Edition, 2009. https://doi.org/10.1002/anie.200900942 ↩ ↩² ↩³ ↩⁴ ↩⁵

3. L. Renee Ruhaak, Gege Xu, Qiongyu Li, et al. Mass Spectrometry Approaches to Glycomic and Glycoproteomic Analyses. Chemical Reviews, 2018. https://doi.org/10.1021/acs.chemrev.7b00732 ↩ ↩² ↩³ ↩⁴ ↩⁵

4. Eliana Saxon, Carolyn R. Bertozzi. Cell Surface Engineering by a Modified Staudinger Reaction. Science, 2000. https://doi.org/10.1126/science.287.5460.2007 ↩

5. Nicholas J. Agard, Jennifer A. Prescher, Carolyn R. Bertozzi. A Strain-Promoted [3 + 2] Azide−Alkyne Cycloaddition for Covalent Modification of Biomolecules in Living Systems. Journal of the American Chemical Society, 2004. https://doi.org/10.1021/ja044996f ↩ ↩² ↩³

6. Jeremy M. Baskin, Jennifer A. Prescher, Scott T. Laughlin, et al. Copper-free click chemistry for dynamic in vivo imaging. Proceedings of the National Academy of Sciences, 2007. https://doi.org/10.1073/pnas.0707090104 ↩ ↩²

7. John C. Jewett, Ellen M. Sletten, Carolyn R. Bertozzi. Rapid Cu-Free Click Chemistry with Readily Synthesized Biarylazacyclooctynones. Journal of the American Chemical Society, 2010. https://doi.org/10.1021/ja100014q ↩

8. Melissa L. Blackman, Maksim Royzen, Joseph M. Fox. Tetrazine Ligation: Fast Bioconjugation Based on Inverse-Electron-Demand Diels−Alder Reactivity. Journal of the American Chemical Society, 2008. https://doi.org/10.1021/ja8053805 ↩ ↩²

9. Scott T. Laughlin, Jeremy M. Baskin, Sharon L. Amacher, Carolyn R. Bertozzi. In Vivo Imaging of Membrane-Associated Glycans in Developing Zebrafish. Science, 2008. https://doi.org/10.1126/science.1155106 ↩

10. The Nobel Foundation. The Nobel Prize in Chemistry 2022. NobelPrize.org, 2022. https://www.nobelprize.org/prizes/chemistry/2022/summary/ ↩ ↩²

11. Jasminka Krištić, Frano Vučković, Cristina Menni, et al. Glycans Are a Novel Biomarker of Chronological and Biological Ages. The Journals of Gerontology: Series A, 2014. https://doi.org/10.1093/gerona/glt190 ↩ ↩²