Analysis of gene expression and chemoresistance of CD133+ cancer stem cells in glioblastoma

If you've ever wondered why glioblastoma so often comes roaring back after appearing defeated on a scan, here’s the unsettling idea that has taken root: a tiny, stubborn subpopulation of cells in these tumors behaves like stem cells. These cells divide slowly, repair damage effectively, and shrug off drugs. You can kill the bulk of the tumor, but these cells can survive, regroup, and rebuild. That's the story a body of work has been establishing, with a particularly clear chapter coming from Liu and colleagues in Molecular Cancer back in two thousand six. They focused on a marker called CD133, which tags a neural stem cell-like subset in the adult brain and appears on glioblastoma cells that exhibit similar behavior. In three primary glioblastoma cultures—referred to as No. 66, No. 377, and No. 1049—they measured how many cells were CD133-positive and then isolated those from their CD133-negative neighbors. The fractions varied wildly from line to line: about ten percent in No. 66, nearly seventy percent in No. 377, and roughly twenty-eight percent in No. 1049. This variability already provides a clue about glioblastoma's heterogeneity. But the more important question was what those CD133-positive cells could actually do. Functionally, they could perform the hallmark stem cell function: self-renew. Take a single CD133-positive cell, place it in conditions that support neural stem cells, and it forms a free-floating neurosphere. This is not a cluster from many cells combined; it’s a sphere traced back to one cell. If you keep them in that neural stem cell medium, they maintain CD133 expression. Even in more standard serum conditions, the CD133-positive populations could be sustained as colonies for dozens of passages—thirty-six in their experiments—after which, if switched back to a stem cell medium, they would migrate off the dish into floating spheres once again. Other teams have already shown that these CD133-positive glioblastoma cells can differentiate into various neural lineages—neuronal, astrocytic, oligodendroglial—so the functional picture aligns with their name: stem-like. Now, the obvious next step is to ask what's going on under the hood. Liu and colleagues compared gene expression in the sorted CD133-positive and CD133-negative fractions. They used real-time polymerase chain reaction and a standard way to interpret those measurements. If you've seen the delta C T method, this is it: measure how many cycles it takes to detect a target gene and how many it takes to detect a control like beta-actin, take the difference, and compute two to the power of minus that difference. This provides a relative quantity. It's a tidy equation, but what matters is how significant the gaps were. On the stemness side, some differences were enormous. MELK, a kinase linked to progenitor cell proliferation, shot up more than a thousand-fold in one line. It was one thousand three hundred fifty-one times higher in CD133-positive cells from No. 66, and hundreds of times higher in the other two. CXCR4, a chemokine receptor that guides cell migration, was also significantly elevated: roughly three hundred thirty-eight-fold in No. 66 and about two hundred fifty to two hundred sixty-five-fold in the others. Markers associated with neural precursors, such as Nestin and Musashi-1, increased as well. Nestin saw an increase of around twenty-fold, while Musashi rose approximately fifty- to eighty-fold across the lines. Components of the Hedgehog pathway, GLI1 and PTCH, increased by ten-fold or more too. Moreover, there was a set of developmental players—Bmi-1, Sonic hedgehog itself, OCT4, PSP, Snail—that were simply undetectable in the CD133-negative cells but present in the CD133-positive ones. Different lines, same pattern: a consistent stemness transcript signature. These transcripts hint at behavior, but the team wanted to connect the signature to the clinical challenge of drug resistance. They conducted a straightforward viability assay. They sorted the cells, plated equal numbers, treated them with the typical glioma chemotherapies—temozolomide, carboplatin, etoposide, known as VP sixteen, or paclitaxel, also called Taxol—and measured the living cells after forty-eight hours using a WST-1 readout. Across all three lines, the CD133-positive populations were much more difficult to kill. The differences were statistically significant compared to their autologous CD133-negative counterparts. And for temozolomide, the resistance corresponded with a familiar culprit. MGMT, the DNA repair enzyme that removes the drug’s key lesion, was elevated by about thirty to fifty-six-fold in CD133-positive cells relative to their neighbors. Chemoresistance is rarely a single issue; it’s a combination of factors. In those CD133-positive cells, multiple mechanisms were at play. The efflux pump BCRP1—the type of protein cells use to pump drugs back out—increased four- to six-fold. Anti-apoptotic pathways were highly activated: FLIP, which blocks a death signal at the membrane, was up by roughly one hundred fifty to nearly three hundred-fold, depending on the line. BCL-2, the classic mitochondrial protector, increased up to around fourteen-fold, while its relative BCL-XL rose a few-fold as well. Members of the inhibitor of apoptosis family also showed notable increases. XIAP jumped about ten- to twenty-two-fold, cIAP1 climbed as high as thirty-nine-fold, with NAIP and cIAP2 also elevated. On the contrary, the pro-apoptotic gene BAX was down-regulated—decreasing to about a third to half of the level seen in CD133-negative cells. String these factors together, and you get a comprehensive picture: drugs enter, more of them are pumped out, DNA damage is repaired faster, and the internal triggers for cell death become harder to activate. There was another telling signal hiding in plain sight within that signature. CXCR4, the migration receptor that was so dramatically elevated, isn’t just an identity marker. In many cancers, it directs cells toward gradients of SDF-1 and aids in their movement. So it's not just that the CD133-positive fraction can survive treatment; the transcript profile suggests they may also be primed to invade. In glioblastoma, where the distinction between “recurrent” and “residual” disease often revolves around cells that diffuse into healthy brain, this is significant. All of this would be an academic pattern if it didn't match what happens in patients. Liu's group tested this too. They analyzed paired tumor samples—primary glioblastoma at diagnosis, and the recurrent tumor from the same patient after treatment. In all five cases they examined, CD133 messenger RNA was higher in the recurrent tumor than in the original. The paper doesn’t simplify those to a single fold change for the cohort, but the direction was consistent across patients and significant in their assays. This kind of within-patient comparison is valuable. It indicates we are likely not just observing a marker of a particular tumor type; we are witnessing therapy winnow the population and leave behind a more CD133-enriched, stem-like residue. If you zoom out for a moment, clinicians have long observed the pattern on scans: a glioblastoma shrinks with radiation and chemotherapy, and then, months later, new growth emerges with alarming speed. What Liu and others are providing is a mechanistic bridge from that clinical pattern back to cellular behavior. This bridge has three main points. First, there is a functionally definable stem-like subpopulation in these tumors—CD133-positive cells that can self-renew from a single cell and sustain a stemness program. Second, those cells possess a coordinated gene-expression profile that promotes survival under stress: MGMT for repair, BCRP1 for efflux, a suite of anti-apoptotic mechanisms, and migration machinery like CXCR4. Third, the clinical outcomes—recurrent tumors—show higher levels of the CD133 signal than the primary tumors that preceded them. You don’t need a complicated model to connect those dots. Of course, there are caveats. This work is based on three primary cell lines, and the initial fraction of CD133-positive cells ranged from a tenth to more than two-thirds, indicating how uneven the terrain is across patients. Not every stemness or anti-apoptotic gene was detectable in the CD133-negative cells, which highlights the contrast but also serves as a reminder that absence can sometimes be due to assay depth. Furthermore, CD133 serves as a useful marker, but it is not an absolute definition of what a tumor-initiating cell is. Some tumor-propagating cells may not fit neatly under that label. There's another caution inherent in the biology itself. Normal neural stem cells share aspects of this playbook. They can be relatively quiescent. They resist toxins. They express some of the same markers. So when considering therapy, “target the CD133-positive cells” is not a complete plan; rather, it's a starting point with inherent risks. The objective must be to identify the distinctions where cancerous stem-like cells diverge from their healthy counterparts, the molecular differences that create a therapeutic opportunity. Still, the data provide us with valuable insights. MGMT is not only correlated with temozolomide resistance; it’s mechanistically involved. This is why strategies to inhibit MGMT or bypass its repair pathway have garnered interest. Efflux pumps like BCRP1 can potentially be targeted, although the blood-brain barrier presents its own challenges. If a high-CXCR4 state enables these cells to migrate, that chemokine axis becomes more than just a biomarker; it's a potential way to limit their spread. These are hypotheses to explore, not definitive conclusions. The takeaway is clear enough to remember during your commute. Within a glioblastoma, a small CD133-positive compartment appears to function as a command center for survival. It renews itself. It carries a toolkit for evading chemotherapy—repairing DNA, expelling toxins, and reducing apoptosis. It seems more prevalent after treatment than before. And as Liu and colleagues demonstrated, you can read that story through both the actions of these cells and their expression profiles: a thousand-fold increase in a stemness kinase here, a three-fold decrease in a pro-death gene there, along with the chemoresistance we fear emerging in a simple viability assay after just two days of drug exposure. If there’s a silver lining, it’s that this isn’t a mystery without direction. We have markers to sort by, pathways to measure, and patient-matched samples indicating where to direct our attention. The challenge now is precision—removing the cancer’s stem-like shield without affecting the brain’s own stem cells. That’s a tall order. But it’s a tangible one, and it’s a better place to be than standing in the dark, wondering why the tumor returned. This lecture was created by ennepō. Go to ennepo dot A I to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.