4.3. Phage Training 4.3.噬菌体训练
A remarkable advantage of phage therapy relative to traditional chemical antibiotics is phage capability to evolve to overcome bacterial resistance. Phage training, also known as phage (pre)adaptation, exploits the natural potency for phage populations to quickly evolve to overcome bacterial defense mechanisms. In essence, the virus population is trained to anticipate how the target bacteria in the treated patient will evolve phage resistance. This approach allows the virus population to accumulate random point mutations or gene insertions/deletions to counter bacterial resistance. This way, phages can be evolutionarily trained to acquire expanded host ranges and to evolve traits that minimize the rise of phage-resistant bacteria during therapy.
与传统的化学抗生素相比，噬菌体疗法的一个显著优势是噬菌体能够进化以克服细菌的抗药性。噬菌体训练（也称为噬菌体（预）适应）利用了噬菌体种群快速进化以克服细菌防御机制的天然能力。从本质上讲，病毒群经过训练，能够预测接受治疗的患者体内的目标细菌将如何进化出噬菌体抗药性。这种方法允许病毒群积累随机点突变或基因插入/缺失，以对抗细菌的抗药性。这样，噬菌体就能在进化过程中得到训练，从而扩大宿主范围，并进化出最大程度减少治疗过程中噬菌体抗性细菌增加的特性。
One popular approach in phage training is to propagate viruses in vitro for serial rounds of infection on a nonevolving host strain. Morello and coworkers (69) employed this method for the first preclinical study demonstrating the benefits of trained phages in vivo. After five consecutive passages in liquid culture with a clinical Pseudomonas aeruginosa host strain, a trained phage population evolved to kill the host bacteria with a tenfold greater efficiency than the ancestral naturally isolated phage. This higher efficiency also translated into improved phage killing efficacy in a mouse model. The trained phages were able to kill an independent set of 20 P. aeruginosa clinical strains more efficiently than the ancestral phage. Therefore, phage training was successful in improving both growth rate on a target host and expansion of phage host range to encompass additional unselected bacterial strains (69).
噬菌体训练的一种常用方法是在体外繁殖病毒，对非进化宿主菌株进行连续感染。莫雷洛和同事（69）采用这种方法进行了首次临床前研究，证明了训练有素的噬菌体在体内的益处。经过与临床铜绿假单胞菌宿主菌株在液体培养液中连续五次传代后，训练有素的噬菌体群体杀灭宿主细菌的效率比自然分离的祖先噬菌体高出十倍。在小鼠模型中，噬菌体的杀菌效率也得到了提高。经过训练的噬菌体能够比祖先噬菌体更有效地杀死一组独立的 20 种铜绿假单胞菌临床菌株。因此，噬菌体训练成功地提高了噬菌体在目标宿主上的生长率，并扩大了噬菌体的宿主范围，包括更多未经选择的细菌菌株 ( 69)。
Another phage training approach is based on allowing the host to coevolve with the phages during serial passage in the lab. For example, Borin and colleagues (70) coevolved a lytic phage with its E. coli host for 28 daily serial transfers. Initially, the host evolved to reduce the expression of the phage’s receptor LamB. In turn, phages coevolved to gain the ability to infect through a secondary phage receptor, OmpF. This trained phage was found to suppress bacterial growth and minimize evolved phage resistance more efficiently than the ancestral phage strain (70). Because these trained phages recognize two cell-surface receptors, bacteria would likely have to acquire multiple mutations to evolve complete phage resistance. This outcome is less probabilistic than acquiring a single mutation and may impose a bacterial fitness cost, thus limiting the likelihood of bacteria to develop resistance against the trained phages. Similar results have been reported for trained phages targeting other pathogens such as P. aeruginosa (71).
另一种噬菌体训练方法是让宿主与噬菌体在实验室中连续传递的过程中共同进化。例如，Borin及其同事（70）在每天28次的连续传送过程中，让噬菌体与大肠杆菌宿主共同进化。最初，宿主的进化减少了噬菌体受体 LamB 的表达。反过来，噬菌体通过共同进化获得了通过次级噬菌体受体 OmpF 感染的能力。研究发现，这种训练有素的噬菌体能比祖先的噬菌体菌株更有效地抑制细菌生长，并最大限度地降低进化出的噬菌体抗药性（70）。由于这些训练有素的噬菌体能识别两种细胞表面受体，细菌很可能需要获得多次突变才能进化出完全的噬菌体抗性。这种结果比获得单个突变的概率要低，可能会造成细菌的健康成本，从而限制细菌对训练有素的噬菌体产生抗性的可能性。针对其他病原体（如铜绿假单胞菌）的训练有素的噬菌体也有类似的结果 ( 71)。
Already, phage training has been leveraged successfully in the clinic to update commercial phage cocktails against emerging epidemiological strains of bacteria. For example, Ujmajuridze et al. (72) were able to increase the activity of a phage cocktail targeting multiple species of uropathogenic bacteria (Staphylococcus aureus, E. coli, Streptococcus spp., P. aeruginosa, Proteus mirabilis), from 41% to 75% coverage of clinical strains. The trained cocktail was then used to treat nine patients with phage-susceptible bacterial infections, and the observed bacterial loads decreased in six of the nine subjects (72). In principle, leveraging phage training to expand host range can reduce the number of phage types that must be combined in a cocktail to achieve broad killing and inhibit evolution of bacterial resistance, thus greatly simplifying the production process (73).
噬菌体训练已成功应用于临床，针对新出现的流行病菌株更新了商业噬菌体鸡尾酒。例如，Ujmajuridze 等人（72）能够提高针对多种尿路致病菌（金黄色葡萄球菌、大肠杆菌、链球菌属、绿脓杆菌、奇异变形杆菌）的噬菌体鸡尾酒的活性，临床菌株的覆盖率从 41% 提高到 75%。然后用训练好的鸡尾酒治疗九名患有噬菌体易感细菌感染的患者，九名受试者中有六名的细菌量有所下降 ( 72)。原则上，利用噬菌体训练来扩大宿主范围可以减少鸡尾酒中必须组合的噬菌体类型数量，以达到广泛杀灭和抑制细菌耐药性进化的目的，从而大大简化生产过程 ( 73)。
Despite promising lab studies and clinical relevance of phage training approaches, there are some notable caveats. Previous work warns that trade-offs between increased phage growth rate and decreased host range may occur during phage training (74, 75), and alternative approaches that address this problem are already emerging (75). The efficacy of phage training might be specific to phage and host strains, suggesting the need to determine the efficacy for each phage-host pair. Importantly, the outcomes of long-term coevolution assays are not necessarily predictable, owing to many possible evolutionary trajectories (76). After coevolving phages and bacteria for ten passages, Betts et al. (77) reported variable outcomes, with some phages losing infectivity against their Pseudomonas hosts, which is certainly not the preferred outcome when harnessing phage training for therapeutic applications. In a separate study, four trained phages had similar variable effects on bacteria, which differed from strain to strain (71). This variability in outcomes highlights the importance of using well-designed evolution experiments to train phages and careful evaluations of the trained phage populations, in order to elucidate the extent to which (co)evolutionary trajectories of trained viruses are truly predictable versus highly stochastic, and whether this differs across certain phage-bacteria combinations.
尽管噬菌体训练方法的实验室研究和临床实用性前景广阔，但也存在一些值得注意的问题。以前的研究警告说，在噬菌体训练过程中，可能会出现噬菌体生长速度增加与宿主范围缩小之间的权衡（74、75），解决这一问题的替代方法已经出现（75）。噬菌体训练的效果可能取决于噬菌体和宿主菌株，这表明有必要确定每对噬菌体-宿主的效果。重要的是，由于存在多种可能的进化轨迹，长期协同进化试验的结果并不一定可以预测（76）。在噬菌体和细菌共同进化十次之后，Betts 等人 ( 77) 报告了不同的结果，一些噬菌体失去了对假单胞菌宿主的感染力，这当然不是利用噬菌体训练进行治疗的理想结果。在另一项研究中，四种训练有素的噬菌体对细菌的作用也有类似的变化，不同菌株的效果也不同（71）。这种结果上的差异凸显了使用精心设计的进化实验来训练噬菌体和仔细评估训练过的噬菌体群体的重要性，以便阐明训练过的病毒的（共）进化轨迹在多大程度上是真正可预测的，而不是高度随机的，以及这在某些噬菌体-细菌组合中是否有所不同。