Adaptive therapy achieves long-term control of chemotherapy resistance in high grade ovarian cancer

Drug resistance results in poor outcomes for most patients with metastatic cancer. Adaptive Therapy (AT) proposes to address this by exploiting presumed fitness costs incurred by drug-resistant cells when drug is absent, and prescribing dose reductions to allow fitter, sensitive cells to re-grow and re-sensitise the tumour. However, empirical evidence for treatment-induced fitness change is lacking. We show that fitness costs in chemotherapy-resistant ovarian cancer cause selective decline and apoptosis of resistant populations in low-resource conditions. Moreover, carboplatin AT caused fluctuations in sensitive/resistant tumour population size in vitro and significantly extended survival of tumour-bearing mice. In sequential blood-derived cell-free DNA and tumour samples obtained longitudinally from ovarian cancer patients during treatment, we inferred resistant cancer cell population size through therapy and observed it correlated strongly with disease burden. These data have enabled us to launch a multicentre, phase 2 randomised controlled trial (ACTOv) to evaluate AT in ovarian cancer.


14
Systemic cancer treatment has been based on the principle that delivery of high drug doses will 15 increase the likelihood of eradicating all malignant cells and achieve cure. Unfortunately this paradigm 16 frequently fails, especially in the management of advanced and metastatic solid cancers, regardless of 17 tumour type or specific drug therapy 1 . This is likely a direct consequence of therapy selecting for pre-18 existing drug-resistant subclones that arise stochastically due to the large number of somatic 19 mutations that accumulate during tumour expansion 2 or could be due to the inherent plasticity in 20 cancer cell phenotype 3 . The paucity of available anti-cancer drug therapies means that emergence of 21 a sufficiently large resistant cancer cell population will eventually result in treatment-resistant 22 cancers 1 . 23 24 Evolutionary theory states that treatment-resistance should come at a fitness cost 4 that becomes 25 apparent when the cancer cell is in an environment that exposes the cost. Trait evolution is usually 26 subject to tradeoffs 5 ; if a cancer clone evolves to become optimal at a particular trait, such as 27 maintaining a resistant phenotype, it will inevitably come at the price of being less good at another, 28 for example proliferation, metastasis or invasion 6,7 . It follows that the relative fitness of drug-sensitive 29 and -resistant cells is reversed by drug therapy, such that in the presence of drug, resistant cells are 30 fitter, whereas in the absence of drug sensitive cells have higher fitness 8 . Similar to other biological 31 systems, the tradeoff between reproduction and survival is expected to be most apparent in low-32 resource settings where the limited resources expose "suboptimal" phenotypes. Adaptive therapy 33 (AT) is a new treatment paradigm that exploits these competitive interactions between sensitive and 34 resistant subclones 7 , aiming to maintain a sufficient population of sensitive cells to suppress the 35 proliferation of 'less fit' resistant cells 9 . This approach accepts that within the palliative setting, cancer 36 cannot be eradicated and aims to control rather than cure 8,10 . 37 38 Adaptive therapy is predicted to be beneficial in mathematical models 11,12 and has been shown to 39 prolong survival in preclinical in vivo models 4,13 . The first AT trial to report outcomes used the oral 40 CYP17A1 androgen synthesis inhibitor abiraterone in metastatic castration-resistant prostate 41 cancer 14,15 . Abiraterone AT was directed by changes in the serum tumour marker PSA (prostate-42 specific antigen) as a proxy for total tumour burden. Men who received abiraterone AT experienced a 43 prolonged median time to progression of 33.5 months compared to 14 [18][19][20][21][22] , and is now standard of care for all patients with advanced disease following 59 platinum response 17 . Despite this intensive treatment, the majority of HGSC cases recur and patients 60 are subsequently treated with multiple lines of platinum-containing chemotherapy. At each sequential 61 relapse, chemotherapy is less effective, ultimately leading to treatment failure 23 . 62 63 This relapsing-remitting nature and clear evolution of resistance to platinum-based therapy makes 64 HGSC an exemplary disease in which to study AT. Here we show that platinum-resistant HGSC cells 65 exhibit reduced fitness in the absence of drug that is exposed by low resource conditions. This results 66 in resistant population decline that is mediated by apoptosis rather than secreted factors. Our 67 experiments reveal that this reduced fitness of resistant cells is reversed by platinum treatment such 68 that the growth dynamics of sensitive/resistant populations fluctuate during drug therapy.

69
Importantly, we demonstrate a significant advantage of carboplatin AT compared to standard dosing 70 in mice with established tumours. These discoveries have led to the multicentre, randomised ACTOv 71 clinical trial (Adaptive ChemoTherapy in Ovarian Cancer) 24 that opened to recruitment in March 2023.

73
Like the abiraterone trial, AT dosing in ACTOv will be directed by changes in tumour burden, indicated 74 by the serum tumour marker, cancer antigen 125 (CA125). Ideally, AT would be directed by the size of 75 the emergent resistant population but platinum resistance is not associated with easily trackable 76 markers such as recurrent single nucleotide variants 25,26 or common copy number drivers 27 . However, 77 post-treatment HGSCs do carry new copy number alterations (CNAs) in addition to their already highly 78 altered genomes 28 that appear to be patient-specific. We recently developed a new bioinformatics 79 pipeline, liquidCNA (LiqCNA) 29 , which exploits these copy number changes to quantify the emergence 80 of treatment-resistance. Thus in situations like platinum chemotherapy, where there is no known 81 recurrent mutational driver of resistance, LiqCNA could track the resistant population and decipher 82 whether AT does indeed work by controlling its growth. In the current study, we use sequential blood 83 and biopsy samples to demonstrate that LiqCNA correlates with tumour burden in HGSC patients. This 84 crucial development is expected to improve AT by enabling the evolution of therapy resistance, rather 85 than proxy measures of tumour burden, to direct adaptive drug dosing for future cancer patients.

87
Drug-resistant HGSC cell populations exhibit a proliferative fitness cost in low resource conditions 88 Cisplatin-and carboplatin-resistant HGSC cells were created as we previously described 26 S1). All cells grew exponentially with similar growth rates for all seeding ratios (0.31 ± 0.05 for 107 sensitive cells and 0.43 ± 0.04 for resistant cells, mean ± SD) and by day 7, no cell line had reached 108 logistic growth (Fig.S2). By using the difference between these two growth rates, 'g', estimated from 109 the co-culture data of one seeding ratio (5:95), we fitted all measures starting from other ratios with 110 an accuracy across four orders of magnitude (Fig.1Di). This gives strong evidence of independent 111 growth without competition in high resource conditions. Since resources are constrained in human cancer, co-culture experiments were repeated in constant 113 low resource conditions by exchanging medium for fresh 0.5% FBS-containing medium every 24 hours.

114
This time, although the overall population increased, the size of the resistant cell population 115 decreased over time relative to the total population (Fig.1C). In these low resource conditions, growth 116 rates of mixed sensitive and resistant populations were logistic indicating that populations were 117 competing for resources and that carrying capacity constrained total population size. Resistant cell 118 growth rates were lower than for sensitive cells ( − = 0.08 doublings/day) and remained 119 independent of the initial ratio of sensitive:resistant cells (Fig.1Dii). This demonstrates the fitness cost We next examined mechanisms of decreased resistant cell fitness in low-resource conditions. Cells 138 were grown in low resource conditions (constant 0.5% FBS as before) either as 100% mono-cultures 139 or in co-culture at a starting ratio of 85% sensitive OVCAR4-GFP:15% resistant Ov4Cis and cell cycle 140 profiles were obtained by FACS for up to 13 days ( Fig.2A). There was no significant change in the 141 proportion of OVCAR4-GFP cells in any phase of the cell cycle in co-culture compared to mono-culture.

142
In contrast, compared to mono-culture, more Ov4Cis cells in co-culture appeared in sub-G0 (    We then compared the change in ER calculated by LiqCNA with the change in CA125 for each patient 287 over time (Fig.5B). Clinical details are provided in (Fig.5C) Fig.1E). The slope of this fitted line measures g=g s -g r , the difference in 417 growth rate between sensitive and resistant cells, and is reported in Fig.1E. We then aligned the 418 datasets obtained at other seeding ratios so that the first time-point of each dataset fell on the line.

419
We found that all time-points in all datasets closely followed the linear fit, despite these data points 420 not used for fitting, confirming that the dynamics of sensitive to resistant population ratio is 421 independent of the initial seeding ratio.          Week 12 Pixels (normalised to total p53+ tumour) Pixels (normalised to total p53+ tumour) i.

ii.
Input of sensitive and resistant cells

FIGURE 4
Duration of treatment with Vehicle and Standard Therapy