NEW YORK – Last summer, Isaac Nijman, a core lab director at University Medical Center Utrecht in the Netherlands, was getting ready to party.
His facility, which got its start offering next-generation sequencing mostly to other Dutch researchers, was to turn 10 years old in October. It had also recently acquired an instrument that Nijman saw as the future of his lab: the CosMx Spatial Molecular Imager, a high-resolution spatial transcriptomics and proteomics instrument from NanoString Technologies. His lab was one of the first in the Netherlands to get one and had been an early adopter of NanoString's lower-resolution, whole-transcriptome GeoMx instrument, as well.
The event was going to be big, with guests from across the country. As part of the festivities, NanoString CSO Joe Beechem, the brains behind the firm's chemistries, was set to attend.
But in September, everything began to fall apart. Nijman learned that he would not be able to use his new CosMx, which had only run two experiments so far, due to a decree from a new patent court that had jurisdiction over most of Europe. 10x Genomics, NanoString's chief competitor in the spatial transcriptomics market, had sued for patent infringement in June and just a few months later had convinced the court that it was likely to win.
Beecham, one of two key guests, pulled out citing a busy schedule, and Nijman cancelled the celebration. "We didn't want to have this overshadow the party," Nijman said. "We felt awkward, as well." He had spent months convincing potential collaborators in the Netherlands and elsewhere to write proposals for studies on that CosMx. "That's their time they invested. You build a relationship, and then it feels a little bit like you overpromised," he said.
The lab was able to hold a small celebration for themselves, Nijman said, "but we didn't make a huge national thing out of it."
Nijman is just one of many spatial researchers feeling squeezed as 10x, NanoString, and other companies offering high-resolution spatial transcriptomics instruments are duking it out for dominance in the market.
After a concerted push in 2022 to complete development of its Xenium in situ analysis platform, 10x Genomics has spent much of 2023 aggressively staking out its place in labs around the world. Sales have taken off, but the firm has not only ridden the strength of its technology, it has also lashed out at competitors, namely NanoString and its CosMx but also Vizgen and Curio Bioscience.
The legal strategy appears to be working, scaring off some researchers who were considering using the CosMx platform to run studies; for those who had already committed, it has caused chaos and sleepless nights.
"It really felt like a bad dream for us when we were told about [the injunction]," said Lasse Kristensen, a researcher at Aarhus University in Denmark, who obtained a CosMx with the help of a government grant. "I could see that they can sue each other, and maybe one has to pay some money. But to suddenly close down all research related to this technology, that was really unimaginable to me." His lab is considering using 10x's offerings, "but we already have some interesting data that we cannot really follow up on until we can use CosMx again."
Generally speaking, researchers do not hold NanoString blameless, however, with questions swirling about whether it could have done more to avoid the IP battle with 10x and also whether it was too quick to market with CosMx.
And new studies, issued as preprints late last year, are suggesting that CosMx simply doesn't compare to Xenium where it counts the most: detecting transcripts. However, these studies have also been flashpoints. NanoString's Beechem said that some of the CosMx runs included in the studies were not a good representation of what the platform could produce.
Cognizant that these technologies are new and cutting edge, researchers still feel that some products in the spatial biology market are not where they need to be in terms of reliability or even in terms of matching the advertised specifications.
Unlike Nijman, Ioannis Vlachos, who directs the Spatial Technologies Unit, a center dedicated to spatial tissue profiling at Beth Israel Deaconess Medical Center, has been able to celebrate the use of his platforms. However, often the celebrations seem disproportionate to the achievement.
"If you're going out for a good dinner because the assay worked, that's not a good assay," Vlachos said. "For some assays, that's how it was initially. I'm not saying we did this every time. But a good assay is an assay where you don't focus on the assay, you focus on the findings."
The list of issues that researchers are dealing with is long, and seemingly no platform is without them. Tissue sections are prone to fall off glass slides, burning the sample and the expensive reagents.
Runs may not pass quality control. Some assays will work easily with some sample types, but not with others. Cell segmentation algorithms aren't yet optimized. Data storage needs are immense and not always straightforward. Bioinformatics pipelines have not yet caught up to these data types, and there is little commonality between the data types.
Frustrations then action
For Luciano Martelotto, a single-cell and spatial sequencing expert at the University of Adelaide's Center of Epigenetics in Australia, frustrations peaked early last year. "After witnessing my team's struggles and the absence of tangible support from some of the spatial platforms providers, I was prepared to take on everyone by myself," he said. Conversations with Vlachos and Jasmine Plummer, founding director of the Center for Spatial Omics at St. Jude Children's Research Hospital, prompted him to "channel this energy into something more productive."
In mid-2023, the trio launched an organization they're calling GESTALT, short for the Global alliancE for SpaTiAL Technologies. The word comes from German — literally, "shape" — and is intended to invoke the idea of "an organized whole that is perceived as more than the sum of its parts," Martelotto said.
The response was immediate and immense: In just a matter of months, it has grown to more than 600 people with interest from around the world. Already, working groups have been formed to tackle data analysis, standardization, and the drive to the clinic. There's even talk of a group to be the official liaisons between the technology developers and the GESTALT community.
"This has evolved to be more like preventative maintenance for the field," Martelotto said. "We keep making decisions based on fear of missing out, and it's hard to tell what's real and what isn't. There is a disconnect between expectations and where the field is, where we are versus where we want to be. It's happened multiple times in the past. We all want to be part of the revolution."
Behind the frenzy in the field, there's a belief that spatial technologies will bring clarity to genomics.
"I can see anything in a spatial data set. That's why people are excited," Plummer said. "It's beautiful to see biology in a picture. But what will be signal and what's noise?"
All three of the GESTALT cofounders said they shared a passion for the technology. Vlachos noted that when these methods first came on the scene around 2019, he dropped everything to reallocate resources toward incorporating spatial profiling into his work. "My research has benefited tremendously," he said. "It's very easy to realize it's a paradigm shift. … If we nurture it appropriately, it will become bigger than bulk [analysis] and much larger than single-cell."
More importantly, these new spatial technologies can more easily slide into the existing clinical pathology framework, he said. "I really do believe that it can be a tool that can enable us to have a significant impact on society through research innovation and clinical practice."
"It has the ability to replace standard pathology," Martelotto added.
The companies that have commercialized spatial transcriptomics believe the same, especially at the high-resolution end of the spectrum. NanoString launched first with CosMx in late 2022, and 10x fast-tracked development of Xenium in 2022 in order to catch up.
Vizgen, a spinout of Harvard University, also offers a platform with sub-cellular resolution while companies like Curio Bioscience offer resolution up to 10 micrometers. And the cofounders of Curio have recently published a method that adds spatial information to single-nucleus sequencing experiments.
At 10x, sales of Xenium have been strong, and in 2023, the firm twice raised its full-year revenue guidance because of them. "That [Xenium] works out of the gate and the data is high-quality is resonating" with customers, 10x Cofounder and CEO Serge Saxonov said on a conference call with investors following the release of the firm's second quarter results in August. "All signs are showing there's huge potential."
10x officials noted that they will be prioritizing the commercial push to sell Xenium instruments and are also launching a higher-definition version of their Visium platform.
NanoString, too, believes firmly in CosMx, though the last six months have been trying for the Seattle-based firm. Despite the legal attacks from 10x and multiple rounds of layoffs, NanoString CEO Brad Gray told investors at the annual JP Morgan Healthcare Conference earlier this month that he sees growth in this sector, and the firm recently announced a new whole-transcriptome assay for CosMx slated for release in 2025.
Data disputed
But it could be getting even tougher for NanoString. Last month, two independent teams published preprints showing head-to-head comparisons of CosMx and Xenium. One study from researchers at the Broad Institute also compared the two platforms with Vizgen's Merscope.
"It was an eye-opener about the importance of assay sensitivity," said David Cook, first author of one of the studies, done in collaboration with Martelotto, Plummer, and other researchers. Cook was finishing up his postdoc at Canada's Lunenfeld-Tanenbaum Research Institute and has recently started his own lab at Ottawa Hospital Research Institute, where he plans to use spatial technologies to study ovarian cancer.
His study found that while CosMx offers a bigger panel — over 1,000 genes at the time versus 377 for Xenium — on a per-cell basis, transcript detection was almost twice as low. About a quarter of all the transcripts came from three very highly expressed genes (MALAT1, XB1, and TPT1), "which aren't very informative biologically," he said. When the team looked at the approximately 120 genes found in both panels, transcript counts per cell were similar, though slightly higher with Xenium.
"The biggest concern was the low sensitivity of CosMx data, the inability to get robust detection of marker transcripts where they should be," Cook said. "Combining that with inflated noise, it becomes very difficult to establish the difference between true signal and noise."
By comparison, Xenium produced clean populations in cell typing experiments and between five- and tenfold better detection of individual genes.
Samouil Farhi, senior author of the Broad-led study, declined to discuss it, citing its status as not yet peer reviewed, however, he noted that Cook's study "reaches similar conclusions in a similar way."
NanoString's Beechem said that the different runs on CosMx and Xenium for the Broad-led study are not fit for comparison. He pointed to a data supplement that showed the samples run with CosMx were processed 44 days after being sliced off the tissue block. One run of the Xenium lung panel was processed at a similar time after — 43 days — but the other two Xenium runs, a multi-tissue panel and a breast panel, were run only 21 and six days after slicing, respectively. Some Vizgen Merscope runs took place 74 days after slicing.
"This is unheard of," he said, noting that NanoString's manuals do not recommend letting samples sit on slides for that length of time. "It'd be problematic for all the technologies," he said. Farhi and the Broad Institute did not return a request for comment on these potential issues in the sample preparation.
Beechem said Cook's CosMx runs would also be "outside the range of what we'd consider a normal distribution of transcripts. I think that's also an anomalously low CosMx signal."
Unreviewed and unpublished data on breast cancer samples from researchers at the University of Tokyo led by Yutaka Suzuki showed results more favorable to CosMx. CosMx delivered up to five times as many transcripts per cell, Suzuki claimed, and more than 10 times more transcripts per square micrometer compared to Xenium. CosMx provided data on 890 unique genes detected compared to 160 for Xenium, with 120 detected by both, according to a web page published by Suzuki.
"What concerns me as a user is that while CosMx data can be good, it can also be quite bad, and there's no clear explanation for why that occurs," Cook said. "I've been told that scan time is a big contributor, and therefore full slide scans like the one we did are more likely to result in worse data. … Regardless, even considering the higher quality CosMx datasets out there [such as the Suzuki breast cancer data], Xenium outperforms it."
Farhi's study included an example of the frustrations with spatial platforms that have plagued many others. The Vizgen Merscope protocol called for an imaging depth of 10 micrometers, but because of human error, it was set to 5 micrometers.
The study included the data as an example of a failed run, but on second try that run failed, as well.
Vlachos, the spatial core lab director and GESTALT cofounder, said that he has seen platforms that never worked, though often field application scientists are able to work out a solution with his lab. Still, these issues are disruptive and costly. As head of a center that provides production-grade services, Vlachos said he's able to take the time to make sure they work. For other labs, who may be part of a consortium or have grant applications riding on new data produced by these platforms, such delays would be even more costly.
"I think a lot of things have been launched prematurely or have made it so that the onus has been on scientists to figure it out," Plummer said. "But that was true of single-cell back in the day. We're in the beginning [with spatial]. I would hope that companies, prior to all the marketing, would have made it clear what the caveats are, but that's a speed bump amongst all emerging techs."
Cook suggested the issues were detracting, or at least distracting, from talk about the science. "I want to talk about the biology, but instead we're talking about how we should invest our money and what's going to be acceptable by reviewers in two years," he said.
Like many working with spatial technologies, Cook has joined the ranks of GESTALT. Among the core benefits are collaboration and knowledge sharing, Martelotto said.
"With GESTALT, we are committed to making a meaningful contribution to science in a manner that is equitable for all," he said. "Our goal is to ensure that taxpayer money is invested wisely in instruments that are truly valuable and functional and won't become mere paperweights."
Interoperability and standardization are key goals. "We will champion the creation and adoption of standards that will become best practices," he said. This will help get everyone on the same page. Ideally, he said, someone running a spatial technology could provide a minimal amount of information so that someone on the other side of the world could do the same.
While GESTALT is rich in enthusiasm, it does not have any funding at the moment. However, Martelotto said there are plans to actively seek financial support for its initiatives.
Will it get worse before it gets better? That's an open question for the field, and 10x has a real chance to box out its competition. NanoString has received a delisting warning from the Nasdaq and has laid off about 25 percent of its workforce in the last half-year. Vizgen has reportedly laid off employees as well, while 10x laid off approximately 8 percent of its employees in August 2022.
Moreover, Europe is not the only place where 10x is winning its legal battles. In November, 10x won $31 million in damages after a US jury found that NanoString's lower-resolution, whole-transcriptome GeoMx platform infringed patents held by 10x. 10x said it is seeking an injunction to stop sales of that platform.
Nijman, the Dutch core lab director, is trying to keep up what momentum he has left with spatial. He's engaged with the researchers he had convinced to use CosMx, except now the conversations are about running them in the US. Many are not willing to send their samples so far away. Rumor has it there's a CosMx running in a lab somewhere in Spain, which is not party to the Unified Patent Court.
He's frustrated with NanoString for not settling the legal issues upfront and with 10x for pursuing its claims so aggressively. Part of his frustration comes from having seen this before with Illumina's aggressive prosecution of its sequencing-by-synthesis patents.
"It's disruptive for science and not good if there's no competition in the field," he said. "We definitely don't want to see 10x completely dominating and all the other competition wiped out. Companies have to exist, but this is the darker side of companies the scientific community doesn't like as much."