In August, Dr Barbara Ghiglione of the Institute of Research in Bacteriology and Molecular Virology (IBaViM) at the University of Buenos Aires (UBA) in Argentina presented a poster at the Environmental Dimensions of Antimicrobial Resistance Conference 2026 (EDAR8) in Australia on how she implemented the PHA4GE Wastewater Contextual Data Specification in an antimicrobial resistance (AMR) surveillance pilot at a hospital in Buenos Aires. Here, she paints a backstory to that poster, and outlines the adoption of the specification in her study.
How it started
The project was born in late 2017, at a turning point for me: I was closing my post-doc and taking my first steps as an independent researcher. Together with Milena Dropa, from University of São Paulo (USP), we had been kinetically characterising a beta-lactamase of the CTX-M family, and the good dynamic of that collaboration led us to build a joint project that would strengthen us both.
That year I applied to one of the first grant programmes for young investigators in Argentina and, together with Dr. Dropa, I wrote the project and we started the sampling. I reached out to two hospitals in the City of Buenos Aires and, with a great deal of effort and sterile jars, we began collecting samples from the sewage chambers. Nobody was doing this in those hospitals: there were few precedents in the country, isolated efforts, and no network of researchers in contact with one another.
The underlying question was whether Klebsiella spp. from wastewater could serve as a bioindicator of the environmental dissemination of antimicrobial resistance – whether the clones reaching the treatment plants are the same ones circulating in the hospital, and whether the genes, genetic platforms and clones found in the sewage represent those colonising the population. Using classical microbiology, we built a collection of around 100 Klebsiella spp. isolates. Back then all my data management lived in spreadsheets: metadata and susceptibility profiles guided by CLSI standards and by in-house protocols adapted from clinical samples.
How it grew
When we returned to the lab in 2021, still under restrictions, we gradually began characterising the isolates: we ran the antibiograms and, step by step, extracted the DNA and sequenced by Illumina. In 2024 we applied to the PHA4GE call together with Dr. Maria Sol Haim, an expert in bioinformatics.
In the meantime I was invited to join a project collecting E. coli from the Parana river. By the time the PHA4GE opportunity came up, I was already bringing together experiences from different matrices and settings: Santa Fe (INALI), with a much more environmental focus (river and lagoon), and myself in Buenos Aires, in an urban setting with hospital sewage effluent. That, together with seeing the specification and everything it makes possible to measure, opened my mind.
In discussing the specification we also saw that the definition of a “sample” can vary from group to group: for us, from a single sample we would keep the gram-negative bacilli and, from there, the Klebsiella – but we could equally keep the total DNA.
With the specification and the protocol in place, in 2025 and 2026 we ran four sampling campaigns – one per quarter, in both hospitals – together with Micaela Asiner, a physician and toxicologist, with the aim of measuring antimicrobials. Dr Asiner is now in Neuquen, where she can set up a large-scale project like the one in Santa Fe: stream, lake, treatment plant and hospital effluent.
The network kept growing: we brought in an Argentine PhD student, now a post-doc at the University of Dresden working on environmental settings, and we began collaborating with the virology group at our own institute, who joined the hospital sampling to look for viromes; our first joint sampling was in July of this year.
Using the specification, and how we ‘updated it’
More than modifying the specification, in the beginning the specification updated us. Using the pathogen-agnostic template in DataHarmonizer, we populated the core of required and recommended fields (about 29 of that template’s 212) ,and left the rest empty where it did not yet apply – a tiered use, not an incomplete dataset. Seeing everything the specification allows you to measure organized our methodology and showed us what else to record: it was metadata as a feedback loop, not paperwork.
Over time, as the project grew and we added matrices (hospital effluent, river, lagoon) and collaborators, we filled in more “columns”. And the other sense of “updating” appeared: today we generate data that the specification does not yet fully accommodate. We measure gram-negative bacilli counts and phenotypic resistance percentages to cefotaxime and meropenem; the specification has fields for counts (fecal coliforms) and for diagnostic results, but none tailored to a gram-negative bacilli count. That is precisely where a user can propose a new term through the formal mechanism (the New Term Request) – and that, to me, is the most concrete way our field experience can give something back to the specification.
At the same time, much of what we are adding did not require any change to the tool at all – it was already there, waiting. We chose the pathogen-agnostic template because we worked with Klebsiella isolates recovered from wastewater; in our most recent sampling we took a further step and, from the same sample, we now also perform total-DNA extraction, so that a single sample yields two products – the isolate and metagenomic DNA – both documented within that same template. And as we move to run PCR for specific resistance genes on those same samples, the fields for it are already provided (gene symbol, target presence, and quantitative result under pathogen diagnostic testing), just as the fields for in-silico resistance detection are. In other words, the pathogen-agnostic template already accommodates our whole workflow – isolation, phenotypic resistance, gene PCR, in-silico detection and metagenomics – and our part is simply to populate categories that used to stay empty as the project matures.
The structure was ready before our capacity to generate the data was. Our contribution, then, is above all a use case that is still under-represented: culture-based AMR surveillance, in a hospital and in a resource-limited setting. The structure is already in place, waiting for the larger project; this year, with the difficulties facing science in Argentina, we could not advance as much as we wanted, but the specification grows step by step, together with us. (On the current situation of science in Argentina: Nature 636, 528-529 (2024), “‘There will be nothing left’: researchers fear collapse of science in Argentina“).
Beyond the specification itself, joining this effort has been my first – and so far only – experience working within an international group, and I am learning a great deal from it. PHA4GE invited me to present our progress at several meetings, and people like Emma Griffiths gave me the confidence to train, to work in English and to feel capable of carrying this forward. It has also let me get to know very different realities across other continents and exchange with colleagues around the world – and that, as much as the data structure, is what makes a network like this possible.
What I value most about the specification, as a researcher, is how every member of the community can contribute to improving the tool from their own place and perspective. Seeing the PHA4GE community at work was, for me, a much-needed paradigm shift: it is a form of constant, lifelong learning – one that is communicable, shareable and respectful, and that works to make access to science more democratic and to have a real impact on improving public health.
Dr Barbara Ghiglione
