Occurrence, genotoxicity, and carcinogenicity of regulated and emerging disinfection by-products in drinking water: a review and roadmap for research

Disinfection by-products (DBPs) are formed when disinfectants (chlorine, ozone, chlorine dioxide, or chloramines) react with naturally occurring organic matter, anthropogenic contaminants, bromide, and iodide during the production of drinking water. Here we review 30 years of research on the occurrence, genotoxicity, and carcinogenicity of 85 DBPs, 11 of which are currently regulated by the U.S., and 74 of which are considered emerging DBPs due to their moderate occurrence levels and/or toxicological properties. These 74 include halonitromethanes, iodo-acids and other unregulated halo-acids, iodo-trihalomethanes (THMs), and other unregulated halomethanes, halofuranones (MX [3-chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone] and brominated MX DBPs), haloamides, haloacetonitriles, tribromopyrrole, aldehydes, and N-nitrosodimethylamine (NDMA) and other nitrosamines. Alternative disinfection practices result in drinking water from which extracted organic material is less mutagenic than extracts of chlorinated water. However, the levels of many emerging DBPs are increased by alternative disinfectants (primarily ozone or chloramines) compared to chlorination, and many emerging DBPs are more genotoxic than some of the regulated DBPs. Our analysis identified three categories of DBPs of particular interest. Category 1 contains eight DBPs with some or all of the toxicologic characteristics of human carcinogens: four regulated (bromodichloromethane, dichloroacetic acid, dibromoacetic acid, and bromate) and four unregulated DBPs (formaldehyde, acetaldehyde, MX, and NDMA). Categories 2 and 3 contain 43 emerging DBPs that are present at moderate levels (sub- to low-mug/L): category 2 contains 29 of these that are genotoxic (including chloral hydrate and chloroacetaldehyde, which are also a rodent carcinogens); category 3 contains the remaining 14 for which little or no toxicological data are available. In general, the brominated DBPs are both more genotoxic and carcinogenic than are chlorinated compounds, and iodinated DBPs were the most genotoxic of all but have not been tested for carcinogenicity. There were toxicological data gaps for even some of the 11 regulated DBPs, as well as for most of the 74 emerging DBPs. A systematic assessment of DBPs for genotoxicity has been performed for approximately 60 DBPs for DNA damage in mammalian cells and 16 for mutagenicity in Salmonella. A recent epidemiologic study found that much of the risk for bladder cancer associated with drinking water was associated with three factors: THM levels, showering/bathing/swimming (i.e., dermal/inhalation exposure), and genotype (having the GSTT1-1 gene). This finding, along with mechanistic studies, highlights the emerging importance of dermal/inhalation exposure to the THMs, or possibly other DBPs, and the role of genotype for risk for drinking-water-associated bladder cancer. More than 50% of the total organic halogen (TOX) formed by chlorination and more than 50% of the assimilable organic carbon (AOC) formed by ozonation has not been identified chemically. The potential interactions among the 600 identified DBPs in the complex mixture of drinking water to which we are exposed by various routes is not reflected in any of the toxicology studies of individual DBPs. The categories of DBPs described here, the identified data gaps, and the emerging role of dermal/inhalation exposure provide guidance for drinking water and public health research.     

Effects of swimming training on three superoxide dismutase isoenzymes in mouse tissues.

The purpose of the present study was to investigate the effects of swimming training on the changes in three superoxide dismutase (SOD) isoenzymes in mice. The trained mice underwent a 6-wk swimming program (1 h/day, 5 days/wk) in water at 35-36 degrees C. Immunoreactive extracellular SOD (EC-SOD), copper- and zinc-containing SOD (CuZn-SOD), and manganese-containing SOD (Mn-SOD) contents and their mRNA abundance were determined in serum, heart, lung, liver, kidney, and gastrocnemius muscle. EC-SOD content in liver and kidney was significantly increased with training. After training, CuZn-SOD content rose significantly only in kidney but decreased significantly in heart, lung, and liver. Mn-SOD content showed a significant increase in lung, kidney, and skeletal muscle but a significant decrease in liver. In most tissues, however, the changes in SOD isoenzyme contents were not concomitant with those in their mRNA levels. The results obtained thus suggest that, except for kidney, the responses in mouse tissues of three SOD isoenzymes (protein levels and mRNA abundance) to swimming training are different and that kidney may be one of the most sensitive organs to adapt to oxidative stress during physical training, although the mechanism remains vague.     

Predicting chlorine decay and THM formation in water supply systems

In 1976, the US Environmental Protection Agency (USEPA) published the results of a national survey that showed that chloroform and other trihalomethanes (THMs) were ubiquitous in chlorinated drinking water. Also in 1976, the US National Cancer Institute published results linking chloroform to cancer in laboratory animals, thus giving rise to an important public health issue. Although numerous disinfection by-products (DBPs) have been reported in the literature since that time, with more than 700 confirmed species to date, only a small number have been addressed in either quantitative or health effects studies. The DBPs that have been quantified in drinking water are generally present at low to mid μg/l levels or below. Approximately 50% of the total organic halide (TOX) formed during the chlorination of drinking water and more than 50% of the assimilable organic carbon (AOC) formed during ozonation of drinking water is still not accounted for and little is known about the potential toxicity of many of the vast number of DBPs present in drinking water. The presence of free chlorine is a prerequisite to THM formation. Therefore, a robust understanding of the mechanisms of both chlorine decay and THM formation are fundamental to the management of THMs in water supply systems. This paper presents a review of work undertaken to improve our understanding of these key phenomena and highlights areas of vulnerability in our knowledge and so recommends areas of future research.     

Factors Influencing the Occurrence of High Numbers of Iodine-Resistant Bacteria in Iodinated Swimming Pools

It has been shown that, although iodinated swimming-pool waters are usually free from coliform bacteria and enterococci, the total counts frequently become relatively high. Pseudomonas alcaligenes and Alcaligenes faecalis have been shown to account for most of these high counts. It was of interest, therefore, to compare the microbial flora of four alternately chlorinated and iodinated swimming pools. By means of the membrane filter method and suitable selective media, examinations were made for total viable counts, coliform bacteria, enterococci, staphylococci, Streptococcus salivarius, and P. aeruginosa. Colonies also were picked from membrane filters incubated on standard plate count agar and identified. The results showed that, although viable counts were significantly higher during the iodinated periods, the specific types of bacteria determined were either fewer than or the same as in chlorinated periods. During chlorination, the predominant microbial flora consisted of staphylococci and members of the genus Bacillus. During iodination, however, the P. alcaligenes-A. faecalis group accounted for 92 to 99% of the microbial flora. The accumulation of high numbers of these bacteria was shown to be due to their iodine resistance and their ability to grow rapidly in pool water in the absence of free iodine.     

Impacts of drinking water pretreatments on the formation of nitrogenous disinfection by-products

The formation of disinfection by-products (DBPs), including both nitrogenous DBPs (N-DBPs) and carbonaceous DBPs (C-DBPs), was investigated by analyzing chlorinated water samples following the application of three pretreatment processes: (i) powdered activated carbon (PAC) adsorption; (ii) KMnO(4) oxidation and (iii) biological contact oxidation (BCO), coupled with conventional water treatment processes. PAC adsorption can remove effectively the precursors of chloroform (42.7%), dichloroacetonitrile (28.6%), dichloroacetamide (DCAcAm) (27.2%) and trichloronitromethane (35.7%), which were higher than that pretreated by KMnO(4) oxidation and/or BCO process. The removal efficiency of dissolved organic carbon by BCO process (76.5%)--was superior to that by PAC adsorption (69.9%) and KMnO(4) oxidation (61.4%). However, BCO increased the dissolved organic nitrogen (DON) concentration which caused more N-DBPs to be formed during subsequent chlorination. Soluble microbial products including numerous DON compounds were produced in the BCO process and were observed to play an essential role in the formation of DCAcAm in particular.     

Algicidal and Sanitizing Properties of Armazide

Algicidal and sanitizing properties of Armazide, a new swimming pool additive, consisting of 12% w/v each of dodecylamine hydrochloride, trimethyl alkyl ammonium chloride, and methyl alkyl dipolyoxypropylene ammonium methyl sulfate in solution, were evaluated by laboratory techniques against algae and sewage. Results indicated it to be highly effective in low concentrations especially in conjunction with low concentrations of chlorine. Swimming pool field tests were found to confirm the laboratory findings. Various treatment levels and methods are described for swimming pools using treatments based upon the actual condition of the pool and water. The use of the product permitted a reduction in chlorine residual in pools resulting in greatly reduced requirements for chlorinating chemicals along with absence of irritation, odor, and other undesirable results usually associated with standard pool chlorination methods.     

Determination of haloacetic acids in human urine by headspace gas chromatography–mass spectrometry

Haloacetic acids (HAAs) are water disinfection byproducts (DBPs) formed by the reaction of chlorine oxidizing compounds with natural organic matter in water containing bromine. HAAs are second to trihalomethanes as the most commonly detected DBPs in surface drinking water and swimming pools. After oral exposure (drinking, showering, bathing and swimming), HAAs are rapidly absorbed from the gastrointestinal tract and excreted in urine. Typical methods used to determine these compounds in urine (mainly from rodents) only deal with one or two HAAs and their sensitivity is inadequate to determine HAA levels in human urine, even those manual sample preparation protocols which are complex, costly, and neither handy nor amenable to automation. In the present communication, we report on a sensitive and straightforward method to determine the nine HAAs in human urine using static headspace (HS) coupled with GC–MS. Important parameters controlling derivatisation and HS extraction were optimised to obtain the highest sensitivity: 120 μl of dimethylsulphate and 100 μl of tetrabutylammonium hydrogen sulphate (derivatisation regents) were selected, along with an excess of Na₂SO₄ (6 g per 12 ml of urine), an oven temperature of 70 °C and an equilibration time of 20 min. The method developed renders an efficient tool for the precise and sensitive determination of the nine HAAs in human urine (RSDs ranging from 6 to 11%, whereas LODs ranged from 0.01 to 0.1 μg/l). The method was applied in the determination of HAAs in urine from swimmers in an indoor swimming pool, as well as in that of non-swimmers. HAAs were not detected in the urine samples from non-swimmers and those of volunteers before their swims; therefore, the concentrations found after exposure were directly related to the swimming activity. The amounts of MCAA, DCAA and TCAA excreted from all swimmers are related to the highest levels in the swimming pool water.     

Occurrence of Regulated and Emerging Iodinated DBPs in the Shanghai Drinking Water

Drinking water chlorination plays a pivotal role in preventing pathogen contamination against water-borne disease. However, chemical disinfection leads to the formation of halogenated disinfection by products (DBPs). Many DBPs are highly toxic and are of health concern. In this study, we conducted a comprehensive measurements of DBPs, including iodoacetic acid (IAA), iodoform (IF), nine haloacetic acids and four trihalomethanes in drinking waters from 13 water plants in Shanghai, China. The results suggested that IAA and IF were found in all the water treatment plants, with maximum levels of 1.66 µg/L and 1.25 µg/L for IAA and IF, respectively. Owing to deterioration of water quality, the Huangpu River has higher IAA and IF than the Yangtze River. Our results also demonstrated that low pH, high natural organic matter, ammonia nitrogen, and iodide in source waters increased IAA and IF formation. Compared to chlorine, chloramines resulted in higher concentration of iodinated DBP, but reduced the levels of trihalomethanes. This is the first study to reveal the widespread occurrence of IAA and IF in drinking water in China. The data provide a better understanding on the formation of iodinated disinfection byproducts and the findings should be useful for treatment process improvement and disinfection byproducts controls.     

Comparison of DNA damage in human-derived hepatoma line (HepG2) exposed to the fifteen drinking water disinfection byproducts using the single cell gel electrophoresis assay

Disinfection of drinking water reduces pathogenic infection, but generates disinfection by-products (DBPs) in drinking water. In this study, the effect of fifteen DBPs on DNA damage in human-derived hepatoma line (HepG2) was investigated by the single cell gel electrophoresis (SCGE) assay. These fifteen DBPs are: four trihalomethanes (THMs), six haloacetic acides (HAAs), three haloacetonitriles (HANs), 3-chloro-4-(dichloromethyl)-5-hydroxy-2(5H)-furanone (MX), and chloral hydrate (CH). Based on the minimal effective concentration (MEC) at which DBPs induced significant increase in olive tail moment (OTM), the rank order of DNA-damaging potency is: bromodichloromethane (BDCM)>dibromochloromethane (DBCM)>tribromomethane (TBM)>trichloromethane (TCM) of the four THMs; iodoacetic acid (IA)>bromoacetic acid (BA)>dibromoacetic acid (DBA)>dichloracetic acid (DCA)>trichloroacetic acid (TCA) of the five HAAs; dibromoacetonitrile (DBN)approximately dichloroacetonitrile (DCN)>trichloroacetonitrile (TCN) of the three HANs. The DNA damaging potency of MX and CH is similar to TCA and DCA, respectively. IA is the most genotoxic DBP in the fifteen DBPs, followed by BA. Chloroacetic acid (CA) is not genotoxic in this assay. Our findings indicated that HepG2/SCGE is a sensitive tool to evaluate the genotoxicity of DBPs and iodinated DBPs are more genotoxic than brominated DBPs, but chlorinated DBPs are less genotoxic than brominated DBPs.     

Identification of mutagenic compounds formed during chlorination of humic acid

Humic acid chlorination products are being studied in an effort to identify the chemicals responsible for the mutagenicity formed during water chlorination. In the present report, 19 chlorinated organic compounds have been identified and quantified in ether extracts of chlorinated humic acid solutions. 10 of these compounds, including a number of chlorinated propanones and chlorinated propenals, are direct-acting mutagens in the Salmonella/microsome mutagenicity assay. The position of the chlorine substituent has been found to be an important factor in the mutagenic activity of these two classes of compounds. The total mutagenicity of the compounds identified thus far, when tested either individually or as a composite, accounts for only 7-8% of the total TA100 mutagenicity, and less than 2% of the TA98 mutagenicity formed during humic acid chlorination. The addition of bromide to the humic acid chlorination reaction results in up to a 2-fold increase in the level of mutagenicity formed.     

Comparative assessment of different biocides in swimming pool water

For disinfection of swimming pool water chlorine of chlorine-based products are normally used. In practice, these products have proven their worth regarding killing of pathogenic micro-organisms. Detailed values of their biocidal activity in swimming pool water were not found in literature. In the given study the efficacy of sodium hypochlorite (NaOCl) versus five micro-organisms was investigated.It is known that chlorination of swimming pool water may lead to formation of specific unwanted products like haloform. Nowadays, the concentration of those by-products in swimming pool water is limited and specific measures exist to minimize their formation. Nevertheless, there is increasing interest in alternative methods without by-product formation like e.g. hydrogen peroxide (H₂O₂) treatment.In the given study the antimicrobial activity of sodium hypochlorite was compared with that of different hydrogen peroxide-based products. The test procedure used was specifically designed to simulate practical conditions in a swimming pool but at the same time to lead to adequate reproducibility. Five test organisms were selected being relevant for the swimming pool area: Pseudomonas aeruginosa, Escherichia coli, Legionella pneumophila, Staphylococcus aureus and Candida albicans.The swimming pool water for the test was artificially prepared. Water hardness, temperature and pH value were adjusted to a defined level. Regarding simulation of organic load it was found that a mixture of urea, creatinine and several amino acids was most appropriate.Addition of the test organisms was done in three portions: one big in the beginning and two smaller after 10 and 20 min to simulate recontamination by bathers. Total test period was 30 min. The number of surviving cells was determined after 30 s as well as after 10, 20 and 30 min.Sodium hypochlorite was tested at a concentration of 1 ppm active chlorine. Compared to that three products based on hydrogen peroxide were investigated: pure hydrogen peroxide, hydrogen peroxide + silver nitrate and a trade product based on hydrogen peroxide.Sodium hypochlorite resulted in total kill of the inoculated organisms after 10, 20 and 30 min corresponding to a log 4 reduction. In contrast to that the biocidal effect achieved by the hydrogen peroxide-based products was significantly lower than one log cycle notwithstanding a very high concentration of up to 150 ppm.The test results confirm the very good killing activity of sodium hypochlorite versus micro-organisms relevant for the swimming pool area. Products based on hydrogen peroxide, with or without silver ions, are from a microbiological point of view no real alternative to chlorine disinfection in swimming pools.     

Viruses in recreational water‐borne disease outbreaks: a review

Viruses are believed to be a significant cause of recreationally associated water‐borne disease. However, they have been difficult to document because of the wide variety of illnesses that they cause and the limitations in previous detection methods. Noroviruses are believed to be the single largest cause of outbreaks, which have been documented in the published literature 45% (n = 25), followed by adenovirus (24%), echovirus (18%), hepatitis A virus (7%) and coxsackieviruses (5%). Just under half of the outbreaks occurred in swimming pools (49%), while the second largest outbreak occurred in lakes or ponds (40%). The number of reported outbreaks associated with noroviruses has increased significantly in recent years probably because of better methods for virus detection. Inadequate disinfection was related to 69% (n = 18) of swimming pool outbreaks. A lack of required reporting and nonuniform water quality and chlorination/disinfection standards continues to contribute to water‐borne recreational disease outbreaks.     

Nutritional, cardiac and adrenergic changes induced by swimming training in rats (author's transl)

The effects of training were investigated in male Sprague Dawley rats group (N), fed ad libitum, by measuring the weight increase and food intake of the animals, biochemical parameters (myocardial triglycerides and glycogen levels), mechanical and metabolic properties of the heart, and adrenergic reactions to swimming stress. An other group of rats remaining sedentary served as control (T). Conditioned rats had been submitted to a one hour test swim 6 days a week for 9 weeks. Gradually additional weights were fixed to the thorax. At the end of training, the load reached 6% of the body weight. Both groups were sacrified by decapitation at the same time, thirty hours after the last swimming session. The study of mechanical performances and metabolic properties was achieved with isolated working heart preparation. Adrenergic reactions of swimming stress were evaluated from plasma, heart and adrenal catecholamine concentrations.     

Effects of exercise stress and cold stress on glutathione and gamma-glutamyltransferase in rat liver

Effects of acute and chronic stress (exercise and cold) on glutathione and gamma-glutamyltransferase (gamma GT) in the rat liver were investigated. Such stress, except for in the case of acute exercise, had no definite influence on the glutathione level. On the other hand, gamma GT activity in both the extramicrosomal and microsomal fractions varied substantially, suggesting that acute exercise increases the release ability of the microsomal membrane of the rat liver, and that swimming training and long-term cold exposure stabilize the membrane. Immunoreactive gamma GT, however, did not always correlate with the enzyme activity, especially in the extramicrosomal fraction. Cross-adaptation appeared to exist between swimming training and chronic cold exposure.     

Physical training under the influence of beta blockade in rats: effect on adrenergic responses.

Rats were trained by daily swimming or running exercises with and without daily propranolol injections. Both training methods resulted in cardiac enlargement, but only swimming exercise caused hypertrophy of the brown adipose tissue. These changes were antagonized by beta blockade. The size of the adrenals reflected the stress of the treatments, but other known stress parameters, such as the size of the thymus or sexual organs dit not. Only swimming training without beta blockade sensitized the rats to the calorigenic action of noradrenaline. The cooling rate of the rats in water, when taking into account the insulative capacity of the body, was decreased in swimming-trained as well as in propranolol-treated rats but increased in running-trained rats. The latter two changes may be due to circulatory alterations, while the delayed body cooling in swimming-trained rats probably results from increased heat production capacity. Training-induced resting bradycardia and enhanced tachycardic response to isoprenaline were observable only in the animal groups trained without beta blockade. The pressor response to noradrenaline tended to be higher in the trained groups and the propranolol-treated group than in the controls and was smaller in the animal groups trained under the influence of beta blockade. On the other hand, the hypotonic response to isoprenaline was smaller in the propranolol-treated and running-trained animals. The results emphasize the importance of the sympathetic nervous system in the adaptation of an organism to physical training.     

Morphological and physiological responses to altitude in deer mice Peromyscus maniculatus.

Individuals within a species, living across a wide range of habitats, often display a great deal of phenotypic plasticity for organ mass and function. We investigated the extent to which changes in organ mass are variable, corresponding to environmental demand, across an altitudinal gradient. Are there changes in the mass of oxygen delivery organs (heart and lungs) and other central processing organs (gut, liver, kidney) associated with an increased sustainable metabolic rate that results from decreased ambient temperatures and decreased oxygen availability along an altitudinal gradient? We measured food intake, resting metabolic rate (RMR), and organ mass in captive deer mice (Peromyscus maniculatus bairdii) at three sites from 1,200 to 3,800 m above sea level to determine whether energy demand was correlated with organ mass. We found that food intake, gut mass, and cardiopulmonary organ mass increased in mice living at high altitudes. RMR was not correlated with organ mass differences along the altitudinal gradient. While the conditions in this study were by no means extreme, these results show that mice living at high altitudes have higher levels of energy demand and possess larger cardiopulmonary and digestive organs than mice living at lower altitudes.     

Recovery and diversity of heterotrophic bacteria from chlorinated drinking waters.

Heterotrophic bacteria were enumerated from the Seattle drinking water catchment basins and distribution system. The highest bacterial recoveries were obtained by using a very dilute medium containing 0.01% peptone as the primary carbon source. Other factors favoring high recovery were the use of incubation temperatures close to that of the habitat and an extended incubation (28 days or longer provided the highest counts). Total bacterial counts were determined by using acridine orange staining. With one exception, all acridine orange counts in chlorinated samples were lower than those in prechlorinated reservoir water, indicating that chlorination often reduces the number of acridine orange-detectable bacteria. Source waters had higher diversity index values than did samples examined following chlorination and storage in reservoirs. Shannon index values based upon colony morphology were in excess of 4.0 for prechlorinated source waters, whereas the values for final chlorinated tap waters were lower than 2.9. It is not known whether the reduction in diversity was due solely to chlorination or in part to other factors in the water treatment and distribution system. Based upon the results of this investigation, we provide a list of recommendations for changes in the procedures used for the enumeration of heterotrophic bacteria from drinking waters.     

Recovery and diversity of heterotrophic bacteria from chlorinated drinking waters

Heterotrophic bacteria were enumerated from the Seattle drinking water catchment basins and distribution system. The highest bacterial recoveries were obtained by using a very dilute medium containing 0.01% peptone as the primary carbon source. Other factors favoring high recovery were the use of incubation temperatures close to that of the habitat and an extended incubation (28 days or longer provided the highest counts). Total bacterial counts were determined by using acridine orange staining. With one exception, all acridine orange counts in chlorinated samples were lower than those in prechlorinated reservoir water, indicating that chlorination often reduces the number of acridine orange-detectable bacteria. Source waters had higher diversity index values than did samples examined following chlorination and storage in reservoirs. Shannon index values based upon colony morphology were in excess of 4.0 for prechlorinated source waters, whereas the values for final chlorinated tap waters were lower than 2.9. It is not known whether the reduction in diversity was due solely to chlorination or in part to other factors in the water treatment and distribution system. Based upon the results of this investigation, we provide a list of recommendations for changes in the procedures used for the enumeration of heterotrophic bacteria from drinking waters.     

Congener Level PCB Desorption Kinetics of Field-Contaminated Sediments

The purpose of this research was to study congener level PCB desorption kinetics of field-contaminated sediments and develop a simple methodology to analyze the desorption behavior. Batch desorption kinetic studies were conducted using XAD-4 resin. Two-phase desorption kinetics were observed for most PCB congeners, consisting of an initial fast rate followed by an extended period of slow rate. A dual first-order rate model was fitted to the PCB desorption data to estimate PCB concentrations in the fast and slow desorbing pools. The fast and slow desorption rates were found to decrease with increasing chlorina-tion of PCB congeners, decreasing ortho chlorination, and decreasing temperature. Estimated first-order desorption rate constants for the fast pools were found to be two orders of magnitude higher than those for the corresponding slow pools. The log of first-order rate constants for the different PCB congeners were found to be linearly related to the log of octanol-water partition coefficients. Therefore, desorption rate constants of higher chlorinated PCB congeners, which typically require long-term desorption tests, could be extrapolated using measured rate constants of lower-chlorinated PCB congeners from short-term desorption tests. This finding may provide a process to significantly reduce the duration of experiments required for estimating desorption kinetics.     

The complex response of free and bound amino acids to water stress during the seed setting stage in Arabidopsis

Free amino acids (FAAs) and protein‐bound amino acids (PBAAs) in seeds play an important role in seed desiccation, longevity, and germination. However, the effect that water stress has on these two functional pools, especially when imposed during the crucial seed setting stage is unclear. To better understand these effects, we exposed Arabidopsis plants at the seed setting stage to a range of water limitation and water deprivation conditions and then evaluated physiological, metabolic, and proteomic parameters, with special focus on FAAs and PBAAs. We found that in response to severe water limitation, seed yield decreased, while seed weight, FAA, and PBAA content per seed increased. Nevertheless, the composition of FAAs and PBAAs remained unaltered. In response to severe water deprivation, however, both seed yield and weight were reduced. In addition, major alterations were observed in both FAA and proteome compositions, which indicated that both osmotic adjustment and proteomic reprogramming occurred in these naturally desiccation‐tolerant organs. However, despite the major proteomic alteration, the PBAA composition did not change, suggesting that the proteomic reprogramming was followed by a proteomic rebalancing. Proteomic rebalancing has not been observed previously in response to stress, but its occurrence under stress strongly suggests its natural function. Together, our data show that the dry seed PBAA composition plays a key role in seed fitness and therefore is rigorously maintained even under severe water stress, while the FAA composition is more plastic and adaptable to changing environments, and that both functional pools are distinctly regulated.