For the most up-to-date information, please see our external website here

____________________________________________________________________________

Technology developement

Rapid phenotypic Antibiotic Susceptibility Testing

Antibiotic resistance is one of our largest global health challenges. We develop technologies to guide rapid, effective antibiotic treatment and to limit the use of broad-spectrum antibiotics. Our tools allow us to trap and observe individual bacteria in controlled environments and quantify their growth dynamics in response to antibiotics in real time. This makes it possible to determine which antibiotic can be used in less than an hour, rather than over days, for many bacteria.

In our first diagnostics paper (Baltekin et al., PNAS 2017), we measured the average growth-rate impact across many individual bacteria and obtained robust results across bacterial isolates by normalizing to untreated control. Recognizing the clinical potential, members of the lab founded Astrego Diagnostics in 2017 to translate the method into a practical system. Through engineering advances in nanofluidic chip design, automated imaging, and data analysis, the concept was developed into a robust diagnostic platform. This ultimately led to the PA‑100 system, which is used at European primary care facilities to determine antibiotic susceptibility for urinary tract infections in 45 minutes. The system was also awarded the first Longitude Prize in 300 years.

Currently, we are developing methods to perform sepsis AST directly from patient blood (Miguélez et al. npj Digital Medicine 2025) and tools for rapid tuberculosis (TB) susceptibility testing (Tran et al. Nature Commun. 2025). Sepsis is a common cause of death across the globe, where patients deteriorate quickly, often within hours. A challenge for rapid AST is that the bacterial concentration in the blood of septic patients is very low. We are therefore investigating methods to perform AST at the ultimate sensitivity level: one single cell.

Drug-resistant TB is a major threat to human health and a contributor to antimicrobial resistance worldwide. The bacteria that cause TB grow slowly, and testing the bacteria’s sensitivity to different antibiotics is incredibly time-consuming. Using microfluidic solutions to trap bacteria and monitor drug response at the single-cell level in real time, we can shorten antibiotic sensitivity testing for TB from several weeks to less than 24 hours.

Finally, we have developed a method for combining AST and species identification at the single-cell level, enabling rapid AST of mixed samples (Kandavalli et al. Nat. Commun. 2022).

Optical Pooled Screening

We first described and implemented the method that is now known as Optical Pooled Screening. Here, a pooled library of cell strains is imaged and phenotyped before they are fixed, and the individual strain’s identity is identified by in situ genotyping of a genetic barcode. The method was first described in a patent application that was made public in January 2016 (Elf et al. 2016). Our proof-of-principle publication (Lawson et al. MSB 2017) described live-cell single-molecule phenotyping capability in a minimal CRISPR library. Later, we published a larger CRISPRi screen for replication initiation accuracy (Camsund et al. Nature Methods 2020). In collaboration with the Mats Nilsson lab, we have extended the method to work with chromosomally expressed barcodes (Soares et al. 2025).​

Single-molecule tracking in living cells

Together with Gene-Wei Li, Johan Elf first demonstrated single-molecule tracking in living cells using stroboscopic illumination in the Xie lab (Elf et al. Science 2007). We also characterized the time it takes for a transcription factor to find and bind its chromosomal operator. We later used the stroboscopic method (English et al. PNAS 2011) to track freely diffusing proteins for the first time, and to introduce the concept of monitoring in vivo binding states based on diffusion constants. The paper’s results regarding RelA biology have, for good reason, been questioned (Elf and Barkefors, Annual Review of Biochemistry, 2019). In 2013, we published a method to track individual molecules binding and unbinding and determine their association and dissociation rates using hidden Markov model analysis (Persson et al. 2013). We have since used the single-molecule method to characterize intracellular search kinetics for LacI (Hammar et al. Science 2012, Nature Genetics 2014), Cas9 (Jones et al. Science 2017), and in homologous recombination (Wiktor et al. Nature 2021) after a double-stranded break. Together with the Hell group, we also developed MINFLUX (Balzarotti et al. Science 2017) and extended it to rapid single-molecule tracking in 3D (Amselem et al. Nature Communications 2023).

The next subvolume method

The next subvolume method is an algorithm for performing correct simulations of stochastic chemistry in three spatial dimensions (Elf and Ehrenberg, Systems Biology 2004), as described by the reaction-diffusion master equation. In this algorithm, space is divided into N voxels, and the time for the next reaction in each voxel is sampled and ordered in a data structure that supports fast search. This enabled reaction-diffusion simulations that scale with log(N) rather than N in a naive Gillespie implementation. This makes it practically possible to simulate stochastic 3D systems.

In extensions, we have developed simulation software (Hattne et al. Bioinformatics 2005), corrections to the RDME that apply when the voxel size approaches the reaction radius of the molecules (Fange et al. PNAS 2010), and tools to simulate single-molecule microscopy experiments (Lindén et al. Nature Communications 2016).

Popular science presentation

Our research explores some of the most fundamental questions in biology: how genes are turned on and off, how chromosomes are organized, and how bacterial cells grow, divide, and evolve. To answer these questions, we combine physics, biology, and cutting-edge technology. Along the way, we also develop new groundbreaking tools when current methods fall short. Somewhat unexpectedly, some of these advances are now opening the door to faster, more powerful diagnostics for infectious diseases.

We focus on a few fundamental questions:

How do molecules find their targets?

Expressing the right proteins at the right time is crucial for every living cell. The proteins that regulate gene expression face a daunting task; they must locate specific sequences within the vast genome, much like finding a single word in a massive library. How they solve this search problem is one of biology’s great puzzles, and we have spent over a decade laying it. We’ve discovered that the protein LacI, which controls lactose metabolism, combines free movement through the cell with short “sliding” motions along DNA to speed up the search. Even though it scans the entire genome, it typically finds its target in just a few minutes.

LacI can scan the sequence from the outside of the DNA double helix. Other DNA-probing molecules, such as the CRISPR-Cas9 gene-editing tool, face a more challenging task. Cas9 is a “programmable” DNA binder, but to compare its RNA program against the genome sequence, it must unwind DNA to check the sequence, which slows the search considerably. This highlights a trade-off: what Cas9 gains in flexibility, it loses in speed.

How does the cell decide when it’s time to divide?

Even when grown under identical conditions, bacteria of a certain species can vary widely in size and growth rate. Yet, we have shown that E. coli bacteria usually start preparing for division when they reach a certain size relative to their DNA content, thereby maintaining the ratio of one chromosome replication per cell division. The mystery lies in the lack of a mechanism for the cell to know when it’s large enough to begin replicating the chromosome. Several previously proposed explanations fail to account for this size-sensing mechanism. We predict that an unknown regulator plays a key role, and we are actively searching for it using advanced genetic, optical, and microfluidic tools.

When bacteria grow in optimal conditions, their generation time can be shorter than the time required to make a chromosome, which means they have to start producing genetic material for their offspring's offspring. How they manage to do this and, at the same time, continue to perform all the functions required for survival is fascinating. If we had a dynamic map of bacterial DNA that allowed us to track the various events as the cell cycle progresses, we would be much wiser. For human cells and other eukaryotes, this map is available, but for bacteria, we are still left guessing. To change this, we are using advanced microscopy and computational modeling to create the first 4-dimensional map of the bacterial chromosome.

How does bacterial resistance emerge and how do we stop it?

Antibiotic resistance is one of the most urgent global health challenges. By studying millions of individual cells in microfluidic chips, we can identify and characterize the rare events that drive emerging resistance.
For example, we have discovered that a small fraction of bacteria can continue growing even after exposure to lethal antibiotic levels. These “survivors” receive additional opportunities to acquire resistance mutations, underscoring the importance of adhering to prescribed treatment regimens even when symptoms have resolved.

To mitigate the consequences of resistance and slow future development, we are developing new tools that can test how bacteria respond to antibiotics in real time. This technology is the foundation of the PA-100, a diagnostics system that can detect resistant urinary tract infection in < 60 minutes. The PA-100 is available at care facilities across Europe, already making a difference in the fight against resistant bacteria. We are now adapting the technology for other infections, including sepsis and tuberculosis, two global killers made worse by antibiotic

Intracellular search problems

How do molecules search for and identify specific genetic information in the genome? This question has been the theme for several different projects in the lab. We have concluded that the transcription factor (TF) lacI, which acts as a repressor of the lactose digestion machinery in the cell, searches for its binding site using a combination of 3D diffusion and 1D sliding on the chromosomal DNA. The TF slides about 45 bases before detaching after 1 ms, and although it searches the entire genome, it takes only a few minutes for the protein to find and bind the operator (Hammar et al. Science 2012; Hammar et al. Nature Genetics 2014). We have also measured how the TF interrogates DNA at the microsecond time scale (Marklund et al. Nature 2020; Marklund et al. Science 2022).

In contrast to LacI, which searches the DNA sequence via interactions with the DNA grooves, Cas9 must unwind the double helix and interrogate the DNA sequence to discriminate right from wrong. Although Cas9’s search problem is somewhat reduced by the fact that potential targets are defined by the presence of a PAM sequence, it remains daunting. We have found that, on average, the dCas9 molecule spends 6 hours in search for its target sequence. Compared to the search time of transcription factors, it’s an eternity, but it is the price the CRISPR/Cas9 system has to pay to be reprogrammable to target any sequence (Jones et al. Science 2017).

We have also contributed to answering the long-standing question of how a broken chromosome can search for and find its homologous sister chromosome within minutes following a double-stranded DNA break. By incorporating the broken strands into a stiff, transcellular filament, the cell reduces the search problem from 3D to 2D, thereby considerably speeding up the process (Wiktor et al., Nature 2021).

The bacterial cell cycle and the chromosome structure

A central question in our research has been how bacterial cells coordinate DNA replication and gene expression to maintain homeostasis while remaining responsive to environmental change. We have approached this problem using single-cell and single-molecule methods, combining microfluidics with high-resolution microscopy to directly observe growth dynamics in living bacteria. This work has revealed that gene expression and growth are tightly coupled at the level of individual cells, yet inherently noisy, requiring robust regulatory strategies (Wallden et al. Cell 2016; Knöppel et al. PNAS 2023).

Another aspect of replication that remains elusive is the significance of the 3D genome structure and its contingent relationship with the cell cycle. We now use advanced microscopy and sequencing-based methods to create a time-resolved, high-resolution E. coli chromosome structure that would allow us to ask questions like (1) how reproducible is chromosome organization from cell-to-cell at each point in the cell cycle, (2) does the 4D position of a segment depend on its sequence or its 1D position in the chromosome, (3) how does the structure change in response to large sequence rearrangements, (4) how does the 3D localization of a gene influence its expression, (5) how does the cell manage the different chromosome copy numbers at different growth rates and (6) how is the structure altered to allow homologous recombination after a specific double-stranded break?

Although E. coli is a useful model organism, we also need to look into other species to understand, e.g., pathogenicity and interspecies differences. The chromosome structure of Mycobacterium tuberculosis (M. tb.) plays a central role in the biology of the disease and potentially, in the development of new diagnostic approaches. Chromosome structure rearrangements are closely linked to the pathogen’s ability to persist in hostile environments by regulating metabolic adaptation and antibiotic tolerance. We use the methods developed for E. coli to shed light on these mechanisms.